PDA

View Full Version : The Ultimate Defense Thread



Pages : [1] 2

nate
06-14-2010, 03:20 PM
I think we're bogging down some other threads with a tangential discussion of defense so perhaps this thread can serve as a catch-all for the defensive diversions. I'm not entirely sure but I think these two statements summarize where we're at in the other threads:


POINT: Gomes is a terrible defender who gives away a ton of his bat's value with his glove. He's probably worth the money being paid to him. Chris Heisey is likely to outproduce him given the chance.

COUNTERPOINT: You overvalue defense


POINT: Orlando Cabrera is a poor defender who would give back his offensive value if he had any offensive value. He's way overpaid for his production. Paul Janish has similar value for a lot less money.

COUNTERPOINT: You overvalue defense and you're fat

OK, discuss!

:cool:

Chip R
06-14-2010, 03:21 PM
Oh, Lord.

nate
06-14-2010, 03:24 PM
Oh, Lord.

I hope you're wearing a seatbelt!

Chip R
06-14-2010, 03:26 PM
I hope you're wearing a seatbelt!


The over/under of posts on this thread is 450. I'm taking the over.

nate
06-14-2010, 03:28 PM
The over/under of posts on this thread is 450. I'm taking the over.

We'll need some serious odds for the undertakers.

Lol, get it?

westofyou
06-14-2010, 03:29 PM
Defense is like love, you know what's good, what's bad, what you can live with and what you can live without, everyone can see it in some form, but its payment to that person is likely to be processed differently then it's processed by the next person.

lollipopcurve
06-14-2010, 03:33 PM
Defense is like love, you know what's good, what's bad, what you can live with and what you can live without, everyone can see it in some form, but its payment to that person is likely to be processed differently then it's processed by the next person.

In it for the money, or the metrics?

Chip R
06-14-2010, 04:09 PM
We'll need some serious odds for the undertakers.

Lol, get it?

PuffyPig
06-14-2010, 04:15 PM
There is a huge difference between defensive value at SS than LF.

There is a huge difference in offensive value (right now anyway) between Gomes (the NL RBI leader) and Cabrerra.

You can easiliy make an argument that Cabrerra's offensive value over Janish doesn't justify the defenisive difference between them.

I don't believe you can make an argument that Gomes' offensive value doesn't offset his defensive deficiencies at his curent rate of production, given the lesss value LF is in the defensive scheme of things.

nate
06-14-2010, 04:45 PM
There is a huge difference between defensive value at SS than LF.

There is! It's accounted for in WAR.

High five!


There is a huge difference in offensive value (right now anyway) between Gomes (the NL RBI leader) and Cabrerra.

A number of metrics that measure individual performance better than RBI verify that this is indeed, true.

High five!


You can easiliy make an argument that Cabrerra's offensive value over Janish doesn't justify the defenisive difference between them.

OK, go ahead!


I don't believe you can make an argument that Gomes' offensive value doesn't offset his defensive deficiencies at his curent rate of production, given the lesss value LF is in the defensive scheme of things.

It's the other way around. His defense gives back the majority of his offense.

I'll tell you one of the stats that tells me this. It's one not related to UZR, WAR, wOBA or smiles per inning.

His salary.

dougdirt
06-14-2010, 04:54 PM
I don't believe you can make an argument that Gomes' offensive value doesn't offset his defensive deficiencies at his curent rate of production, given the lesss value LF is in the defensive scheme of things.

Except that there are multiple things you aren't addressing. There is a baseline of 'average' at each position. Average in left field is the 'expected' for all left fielders. They get about 325 balls they are reasonably expected to 'turn into outs' in a full season (line drives, fly balls) based on the 2009 qualified players in terms of their balls in zone and average 'out of zone plays made' over a full season. If the average baseline of turning those balls in a left fielders zone is 89%, and your guy is only doing it at say 82% and they also have 20 fewer 'out of zone plays made' we are now looking at an additional 35 hits given up by the 82% guy (one ever 4-5 games). Left field or shortstop, 35 hits is 35 hits (and I would actually argue 35 hits given up to left is a lot more hurtful than 35 hits by the shortstop given the extra base hits mixed in on those to left).

kaldaniels
06-14-2010, 04:59 PM
To a layperson like me, defensive stats are nice but not a be-all end-all. I view them as somewhat-subjective, and I don't think they consider defensive positioning.

Look at that play BP made the other day where he went paralell to catch the ball...you can't tell me that play got its true due in his defensive stats.

dougdirt
06-14-2010, 05:04 PM
To a layperson like me, defensive stats are nice but not a be-all end-all. I view them as somewhat-subjective, and I don't think they consider defensive positioning.

Look at that play BP made the other day where he went paralell to catch the ball...you can't tell me that play got its true due in his defensive stats.

Positioning is part of defense. Range matters, but if a guy only has to go 8 feet to get the ball because he positions himself better than the guy who can go 12 feet to get it, doesn't matter. Both players got to the ball right?

Sounds like you are indeed a layperson because the right defensive stats do indeed take into consideration positioning because positioning helps or hurts a player in terms of turning a ball into an out.

Brutus
06-14-2010, 05:10 PM
Positioning is part of defense. Range matters, but if a guy only has to go 8 feet to get the ball because he positions himself better than the guy who can go 12 feet to get it, doesn't matter. Both players got to the ball right?

Sounds like you are indeed a layperson because the right defensive stats do indeed take into consideration positioning because positioning helps or hurts a player in terms of turning a ball into an out.

I don't worry about positioning as a whole because I believe it largely evens out. However, when you're trying to evaluate the talent of a player on defense, positioning is a big deal because a better scouting report can give a mediocre player a chance to get to many more balls than he otherwise would. Consequently, more balls could get by a better defender who is victimized by poor positioning of his team.

kaldaniels
06-14-2010, 05:13 PM
What about managierial shifts that dont allow a fielder to get to a ball that he normally would have gotten.

dougdirt
06-14-2010, 05:16 PM
What about managierial shifts that dont allow a fielder to get to a ball that he normally would have gotten.

Unfortunately its part of the game. Defensive stats aren't measuring skill (which is what you are speaking to), they are measuring what happened. Its much like when a player has a 'career year' because his BABIP was .390. It isn't that guys skill level, but it did indeed happen.

nate
06-14-2010, 05:24 PM
What about managierial shifts that dont allow a fielder to get to a ball that he normally would have gotten.

A question: is that something that causes a fielder to miss a great deal more balls than he would've if he'd been positioned "properly?" If so, roughly what percentage more? How do you know?

Also, wouldn't that typically only a very few times as managers and coaches "caught on" and repositioned their players better?

nate
06-14-2010, 05:27 PM
To a layperson like me, defensive stats are nice but not a be-all end-all. I view them as somewhat-subjective, and I don't think they consider defensive positioning.

Look at that play BP made the other day where he went paralell to catch the ball...you can't tell me that play got its true due in his defensive stats.

I'm sure it did. The more range exhibited, the more credit he gets.

I'm not sure why one would even be skeptical of such a thing. Can you explain why he wouldn't get hit "true due?" What would his "true due" be in that situation?

dougdirt
06-14-2010, 05:28 PM
I'm sure it did. The more range exhibited, the more credit he gets.

I'm not sure why one would even be skeptical of such a thing. Can you explain why he wouldn't get hit "true due?" What would his "true due" be in that situation?

Credit for 2 outs?

kaldaniels
06-14-2010, 05:33 PM
I'm sure it did. The more range exhibited, the more credit he gets.

I'm not sure why one would even be skeptical of such a thing. Can you explain why he wouldn't get hit "true due?" What would his "true due" be in that situation?

True due would encompass the speed of the ball hit and the height at which it was caught. The degree of difficulty on that play was great.

dougdirt
06-14-2010, 05:35 PM
True due would encompass the speed of the ball hit and the height at which it was caught. The degree of difficulty on that play was great.

Either it was a hit or it wasn't, right? The spectacular play is still only worth one out (or hit if it isn't made). If Dan Uggla doesn't make that play, its still just a single to CF. Brandon Phillips made it. He got credit for turning that into an out.

kaldaniels
06-14-2010, 05:37 PM
Either it was a hit or it wasn't, right? The spectacular play is still only worth one out (or hit if it isn't made). If Dan Uggla doesn't make that play, its still just a single to CF. Brandon Phillips made it. He got credit for turning that into an out.

But if that play is worth the same in terms of defensive metrics as a softer, higher hit line drive...thats a problem to me.

PuffyPig
06-14-2010, 05:43 PM
Except that there are multiple things you aren't addressing. There is a baseline of 'average' at each position. Average in left field is the 'expected' for all left fielders. They get about 325 balls they are reasonably expected to 'turn into outs' in a full season (line drives, fly balls) based on the 2009 qualified players in terms of their balls in zone and average 'out of zone plays made' over a full season. If the average baseline of turning those balls in a left fielders zone is 89%, and your guy is only doing it at say 82% and they also have 20 fewer 'out of zone plays made' we are now looking at an additional 35 hits given up by the 82% guy (one ever 4-5 games). Left field or shortstop, 35 hits is 35 hits (and I would actually argue 35 hits given up to left is a lot more hurtful than 35 hits by the shortstop given the extra base hits mixed in on those to left).


There is one think you are not addressing.

His actual numbers.

How many basehits is he actually giving up on defense?

edabbs44
06-14-2010, 05:44 PM
A question: is that something that causes a fielder to miss a great deal more balls than he would've if he'd been positioned "properly?" If so, roughly what percentage more? How do you know?

Also, wouldn't that typically only a very few times as managers and coaches "caught on" and repositioned their players better?

Maybe, what if a fielder is positioned due to the guy playing next to him. For example, if the Reds coaching staff knows that Stubbs is going to get to a ball in the gap quicker than Gomes will, would it make sense to play Gomes a step or two closer to the line to take away plays in the corner? The same could be said for a SS who has amazing range, allowing a 3B to play a half step closer to the line.

nate
06-14-2010, 05:50 PM
True due would encompass the speed of the ball hit and the height at which it was caught. The degree of difficulty on that play was great.

That technology is coming. That it doesn't exist now doesn't make UZR useless. It's still infinitely more useful than WoTV.

UZR takes the emotion out of the play. You think that play is good because it was exciting and memorable and you'll remember it until the next one (or tomorrow if your memory is like mine.) But UZR remembers it as being in zone X and can compare that play to every other play made by every other 2B when hit by a certain handed batter who has a certain level of speed off a certain handed pitcher in a certain ballpark.

To the point, UZR typically finds BP to be a good defender if not excellent. Is that wrong?

nate
06-14-2010, 05:58 PM
There is one think you are not addressing.

His actual numbers.

How many basehits is he actually giving up on defense?

It looks like he's had 61 balls hit into his zone and made 50 plays. He has made 10 plays out of his zone. So, 11 hits.

If I'm reading this (http://www.fangraphs.com/leaders.aspx?pos=lf&stats=fld&lg=all&qual=y&type=1&season=2010&month=0) right.

nate
06-14-2010, 06:00 PM
But if that play is worth the same in terms of defensive metrics as a softer, higher hit line drive...thats a problem to me.

A softer, higher ball might be fielded by someone else. Or not at all.

dougdirt
06-14-2010, 06:15 PM
There is one think you are not addressing.

His actual numbers.

How many basehits is he actually giving up on defense?

It is impossible to truly tell. Just like its impossible to truly tell how much a 'hit' is truly worth. Players get credit in offensive stats if they get a single with no one on and the next batter makes the final out of the inning. Sure, they got a hit, but it literally led to nothing. But they get a fraction of a 'run' for that in offensive metrics and everyone seems to be just fine with that.

dougdirt
06-14-2010, 06:18 PM
But if that play is worth the same in terms of defensive metrics as a softer, higher hit line drive...thats a problem to me.

Sounds like you would be a fan of John Dewan's +/- system, in which they watch and chart every play in baseball via video and note the type of ball it was.

Scrap Irony
06-14-2010, 06:25 PM
It looks like he's had 61 balls hit into his zone and made 50 plays. He has made 10 plays out of his zone. So, 11 hits.

If I'm reading this (http://www.fangraphs.com/leaders.aspx?pos=lf&stats=fld&lg=all&qual=y&type=1&season=2010&month=0) right.

So, according to UZR, he's missed eleven balls that an average defender would get to.

And, again, according to UZR, those 11 balls are worth five runs.

Doesn't that seem excessive to anyone else?

Scrap Irony
06-14-2010, 06:28 PM
The other major problem I have with UZR is that a player is compared to other players. It shouldn't affect Gomes' numbers if, say, Carl Crawford is playing or not.

But it does.

A great deal, in fact.

WMR
06-14-2010, 06:28 PM
There might be a good defensive stat 5-10 years from now but the ones being used right now to assign run values are beyond kooky.

nate
06-14-2010, 06:32 PM
So, according to UZR, he's missed eleven balls that an average defender would get to.

And, again, according to UZR, those 11 balls are worth five runs.

Doesn't that seem excessive to anyone else?

Well without even looking at a run value chart, if I think about how many runs can score on 11 hits...no?

In a game situation, how many runs would 11 singles in a row score?

Now consider some of the misplays turn singles into doubles and perhaps even doubles into triples.

Is it totally outside the realm of possibility that 5 runs could score on 11 hits?

nate
06-14-2010, 06:33 PM
The other major problem I have with UZR is that a player is compared to other players. It shouldn't affect Gomes' numbers if, say, Carl Crawford is playing or not.

But it does.

A great deal, in fact.

Explain how.

nate
06-14-2010, 06:34 PM
There might be a good defensive stat 5-10 years from now but the ones being used right now to assign run values are beyond kooky.

I believe they're the same ones being used for offense.

Scrap Irony
06-14-2010, 06:44 PM
Explain how.

UZR attempts to find the average defensive player. Carl Crawford plays LF. His numbers are averaged in with Gomes (and the other LFers) to get that average.

So, if Crawford gets hurt (God forbid), Gomes' numbers improve. (This is part of the reason Jeter's numbers improved last season; four or five defense-first SSs got hurt and/or lost time to other players, so Jeter's numbers got a lot better.)

dougdirt
06-14-2010, 06:45 PM
The other major problem I have with UZR is that a player is compared to other players. It shouldn't affect Gomes' numbers if, say, Carl Crawford is playing or not.

But it does.

A great deal, in fact.

Gomes certainly gets offensive credit when Crawford plays left..... so why doesn't it go both ways?

Falls City Beer
06-14-2010, 06:45 PM
Doesn't that seem excessive to anyone else?

Definitely.

dougdirt
06-14-2010, 06:46 PM
UZR attempts to find the average defensive player. Carl Crawford plays LF. His numbers are averaged in with Gomes (and the other LFers) to get that average.

So, if Crawford gets hurt (God forbid), Gomes' numbers improve. (This is part of the reason Jeter's numbers improved last season; four or five defense-first SSs got hurt and/or lost time to other players, so Jeter's numbers got a lot better.)

And hitting stats try to find the average offensive player too. When Crawford plays, Gomes gets better because Crawford brings that average down.

dougdirt
06-14-2010, 06:47 PM
Definitely.

Except it wasn't 11 plays. It was 11 plays + below average out of range plays, likely to the tune of 4 or 5 more. 14 or 15 hits sounds like 5 runs to me....

Scrap Irony
06-14-2010, 06:47 PM
Well without even looking at a run value chart, if I think about how many runs can score on 11 hits...no?

In a game situation, how many runs would 11 singles in a row score?

Now consider some of the misplays turn singles into doubles and perhaps even doubles into triples.

Is it totally outside the realm of possibility that 5 runs could score on 11 hits?

11 singles in a row? Sure, it's probable five runs would score in that one situation. How many times has that happened in the history of the game, though?

How many runs have Gomes' gaffes cost the Reds? That's what I want to know. Anything else is just noise and ranking.

It certainly shouldn't be considered as a measuring statistic. Ever. You can't add two and two and get orange.

Falls City Beer
06-14-2010, 06:48 PM
Is it totally outside the realm of possibility that 5 runs could score on 11 hits?

Of course not, if they were strung together back to back to...

Scrap Irony
06-14-2010, 06:50 PM
Except it wasn't 11 plays. It was 11 plays + below average out of range plays, likely to the tune of 4 or 5 more. 14 or 15 hits sounds like 5 runs to me....

No, he said Gomes missed 11 balls an average LFer would get to.

Eleven.

Scrap Irony
06-14-2010, 06:52 PM
Gomes certainly gets offensive credit when Crawford plays left..... so why doesn't it go both ways?

OPS measures Gomes against other players?

Not according to my math, doug.

Nor does BA, OBP, HR, RBI, BBs, Ks, or any other counting, measurable stat.

RedsManRick
06-14-2010, 06:55 PM
The other major problem I have with UZR is that a player is compared to other players. It shouldn't affect Gomes' numbers if, say, Carl Crawford is playing or not.

But it does.

A great deal, in fact.

It absolutely should matter. On both sides of the ball, there is no such thing as a 0 baseline. The meaningful analysis is always against replacement and the only way to determine replacement is to look at the other players who are playing out there. It's not like we can say "How many balls does Gomes get to compared to how many balls a non-existent LF gets to?"

Even when looking at simple counting stats, there is an implied context of comparison. If I tell you player X has hit 30 HR, the only way for you to make sense of that is if I tell you how many PA it took him and how hard it is to hit HR.

dougdirt
06-14-2010, 06:56 PM
No, he said Gomes missed 11 balls an average LFer would get to.

Eleven.

Except that statement was not correct at all. Gomes is 50 of 61. No fielder has a 1.00 in terms of BIZ and Outs made. Right now, the league's left fielders are carrying a UZR of .873 (or turning 87.3% of balls in their zone into outs). Gomes is at .820. That means he is missing about 5 balls in the zone per 100 that the 'average' gets to, or roughly 15 over a full season.

Then there are the out of zone plays. The 'averrage' left fielder right now makes 1 OOZ play per 20.3 innings. That means in the given innings Gomes has played, he should have made 19. He actually has made 10.

So at this point, Gomes has cost himself 3 plays in the zone and roughly 9 out of the zone, for a total of 12 plays this season versus the average.

RedsManRick
06-14-2010, 06:56 PM
OPS measures Gomes against other players?

Not according to my math, doug.

Nor does BA, OBP, HR, RBI, BBs, Ks, or any other counting, measurable stat.

Not explicitly. But tell me what Gomes' OPS means in terms of how good of a hitter he guy is without comparing him to other players. No matter what stat you use, you're usually using it in the context of other players.

Is a .300 average good? In the abstract there is no way to answer that question. We only know it's good because we know how few players are able to do it.

dougdirt
06-14-2010, 06:58 PM
OPS measures Gomes against other players?

Not according to my math, doug.

Nor does BA, OBP, HR, RBI, BBs, Ks, or any other counting, measurable stat.

All advanced offensive metrics do weigh them against each other. Sure, not the simple counting stats. But if you want to try and figure out how good of a hitter Gomes is for a left fielder, you better compare him to other players even if you are using your counting stats or you are going to have absolutely no clue how good or bad he actually is because there is literally no baseline.

RedsManRick
06-14-2010, 06:59 PM
So at this point, Gomes has cost himself 3 plays in the zone and roughly 9 out of the zone, for a total of 12 plays this season versus the average.

I wonder how many runs each of those plays cost the Reds compared to the value provided by making an out? Could it conceivably be in the .4 run ballpark?

nate
06-14-2010, 06:59 PM
UZR attempts to find the average defensive player. Carl Crawford plays LF. His numbers are averaged in with Gomes (and the other LFers) to get that average.

So, if Crawford gets hurt (God forbid), Gomes' numbers improve. (This is part of the reason Jeter's numbers improved last season; four or five defense-first SSs got hurt and/or lost time to other players, so Jeter's numbers got a lot better.)

If we don't compare to every other player, how do we know what average is?

kaldaniels
06-14-2010, 07:05 PM
That technology is coming. That it doesn't exist now doesn't make UZR useless. It's still infinitely more useful than WoTV.

UZR takes the emotion out of the play. You think that play is good because it was exciting and memorable and you'll remember it until the next one (or tomorrow if your memory is like mine.) But UZR remembers it as being in zone X and can compare that play to every other play made by every other 2B when hit by a certain handed batter who has a certain level of speed off a certain handed pitcher in a certain ballpark.

To the point, UZR typically finds BP to be a good defender if not excellent. Is that wrong?

So if that technology is not here yet (which would be pretty cool) how in the world, just by using zones, can we tell how many runs a player is worth on defense. I mean, jump over to the Gomes thread and you got people saying exactly how many runs his defense takes away from his offense...that is my issue with UZR/fielding metrics...the results that people are extrapolating from them. As a comparative thing, I don't mind them comparing player x to player y...no problem whatsoever.

kaldaniels
06-14-2010, 07:06 PM
Sounds like you would be a fan of John Dewan's +/- system, in which they watch and chart every play in baseball via video and note the type of ball it was.

Absolutely, the more data used, the better. With the data used now, it is a start...but there is a long way to go.

nate
06-14-2010, 07:07 PM
11 singles in a row? Sure, it's probable five runs would score in that one situation. How many times has that happened in the history of the game, though?

How do you measure him turning a single into a double followed by a "legit" single that leads to a run?


How many runs have Gomes' gaffes cost the Reds? That's what I want to know. Anything else is just noise and ranking.

Just as an example, the Reds have 612 hits and 327 runs. A ratio of 1.87 hits required to make 1 run.

11 hits to 5 runs = 2.2:1

Is it still just a total "whoa, dude!" mindblower that 11 hits could lead to 5 runs?


It certainly shouldn't be considered as a measuring statistic.


But it is.


Ever.

Right now!


You can't add two and two and get orange.

I'm not.

RedsManRick
06-14-2010, 07:14 PM
So if that technology is not here yet (which would be pretty cool) how in the world, just by using zones, can we tell how many runs a player is worth on defense. I mean, jump over to the Gomes thread and you got people saying exactly how many runs his defense takes away from his offense...that is my issue with UZR/fielding metrics...the results that people are extrapolating from them. As a comparative thing, I don't mind them comparing player x to player y...no problem whatsoever.

You look at balls hit in to each zone and what types of hits they tend to become. You then apply the run values for each hit type compared to an out to each play. Multiply it all out and add it up. There's a ton of great stuff in the UZR primer (http://www.fangraphs.com/blogs/index.php/the-fangraphs-uzr-primer/) on fangraphs as well as at Tango's blog on both the run value calculations for each hit type and the UZR math.

It's not a measure of actual runs prevented/allowed. It's a measure of the average run value of the typical result of balls hit to that zone. The actual impact of that performance is going to be highly dependent on the circumstances. But since those circumstances are completely unpredictable, because it's not the fielders fault/credit for creating that circumstance, and because there's no evidence that fielders are better/worse in certain circumstances, using the average is the way to get the most consistent results.

For example, while a diving play in the bottom of the ninth with 2 outs and the bases loaded to save a game is huge in that specific circumstance, both preventing real runs and in win probability, it doesn't mean that the defender is any better of a defender than had he made the same play with the bases empty in the 3rd. Just like we don't give hitters extra credit for hitting 3 run homers versus solo shots on the run production side of things.

Conceptually speaking, UZR is just like +/- with run values attached.

Falls City Beer
06-14-2010, 07:18 PM
Shouldn't the "damage" that one player's defense causes be taken in light of the relative strength or weakness of the rest of the team's defense? Gomes on a bad overall defensive team is likely to be more costly to that team's run prevention than it would to a team that could in effect minimize a Gomes' gaffe. It's a question.

PuffyPig
06-14-2010, 07:28 PM
Shouldn't the "damage" that one player's defense causes be taken in light of the relative strength or weakness of the rest of the team's defense? Gomes on a bad overall defensive team is likely to be more costly to that team's run prevention than it would to a team that could in effect minimize a Gomes' gaffe. It's a question.

I think that is likley true, that an error costs a poor defensive team (whether it being poor becuase of defense or poor pitching) more than a team good at preventing runs.

But, an extra hit helps a great offensive team more I would assume too.

I would assume the the % of hits that are turned into runs increases the more hits you have.

jojo
06-14-2010, 07:30 PM
I don't believe you can make an argument that Gomes' offensive value doesn't offset his defensive deficiencies at his curent rate of production, given the lesss value LF is in the defensive scheme of things.

Gomes' WAR is actually suggesting his marginal offensive value is significantly greater than the magnitude of his marginal defensive value (the absolute value of his bat being worth roughly three times the absolute value of his leather as his glove is minus -4.9 runs relative to replacement and his bat is worth +8 runs above average and another +7 for the difference between average and replacement level given his playing time).

I think these kinds of devils in the details often get lost in these discussions.

Falls City Beer
06-14-2010, 07:33 PM
But, an extra hit helps a great offensive team more I would assume too.

I would assume the the % of hits that are turned into runs increases the more hits you have.

This makes sense.

dougdirt
06-14-2010, 07:34 PM
Shouldn't the "damage" that one player's defense causes be taken in light of the relative strength or weakness of the rest of the team's defense? Gomes on a bad overall defensive team is likely to be more costly to that team's run prevention than it would to a team that could in effect minimize a Gomes' gaffe. It's a question.

Why should it? He is either turning balls into outs or he isn't. The ones he isn't turning into outs lead to hits for the other team does it not? Regardless of how good everyone else on the field is, Gomes is responsible for the balls to left field.

kaldaniels
06-14-2010, 07:36 PM
Why should it? He is either turning balls into outs or he isn't. The ones he isn't turning into outs lead to hits for the other team does it not? Regardless of how good everyone else on the field is, Gomes is responsible for the balls to left field.

You keep making remarks like that. Why not just track the actual runs that missed zone balls lead to. That would be a helpful tool. As you say, it either turned into a run or it didn't.

nate
06-14-2010, 07:37 PM
I think that is likley true, that an error costs a poor defensive team (whether it being poor becuase of defense or poor pitching) more than a team good at preventing runs.

Indirectly in that a poor defensive team gives up more hits.


But, an extra hit helps a great offensive team more I would assume too.


Extra hits are like extra money, always helpful.


I would assume the the % of hits that are turned into runs increases the more hits you have.

Hmm...I dunno. I would think the kinds of hits matter a lot. Lots of singles probably doesn't trump lots of doubles and HR.

VR
06-14-2010, 07:39 PM
It would be interesting to see UZR, even E's....with runners in scoring postion etc.

dougdirt
06-14-2010, 07:39 PM
You keep making remarks like that. Why not just track the actual runs that missed zone balls lead to. That would be a helpful tool. As you say, it either turned into a run or it didn't.

Because advanced offensive metrics don't work that way either, which is where the run translations come from. Advanced offensive metrics will tell you that a single with no one on and two outs that is followed by a strikeout to end the inning is actually worth a fraction of a run despite not leading to any runs at all. The reason is, that the player himself is not entirely responsible for the run. In the defensive scenario, the pitcher takes some of the fault, as does perhaps other fielders on previous or subsequent plays that also contributed to that run.

Falls City Beer
06-14-2010, 07:41 PM
In the defensive scenario, the pitcher takes some of the fault, as does perhaps other fielders on previous or subsequent plays that also contributed to that run.

Another problem with assigning a definite/unchanging run value to plays spread out over time/innings.

dougdirt
06-14-2010, 07:45 PM
Another problem with assigning a definite/unchanging run value to plays spread out over time/innings.

Not really. It is like assigning a player 3 runs for hitting a 3 run HR. That player isn't solely responsible for all of those runs scoring.

Falls City Beer
06-14-2010, 07:47 PM
Not really. It is like assigning a player 3 runs for hitting a 3 run HR. That player isn't solely responsible for all of those runs scoring.

But OPS doesn't take RBI into account.

OUReds
06-14-2010, 07:49 PM
Conceptually speaking, UZR is just like +/- with run values attached.

And they often come up with not insignificant differences, while using the same data. That seems to be a problem to me.

jojo
06-14-2010, 07:53 PM
Another problem with assigning a definite/unchanging run value to plays spread out over time/innings.

There's nothing arbitrary about the linear weights. They're derived from years of data and thousands of observations.

In other words, the standard errors associated with a run value just aren't big so really the potential to have bias introduced via this proposed mechanism (assigning run values to an event based upon the average number of runs that score from the event given the base/out state till the end of an inning over many years versus actually tallying the true runs) really isn't great provided enough defensive innings.

Here's another consideration. Assigning true run values actually confounds things because it adds the element of chance which is completely out of the defender's control (i.e. the same problem associated with rbis for instance).

Falls City Beer
06-14-2010, 07:58 PM
There's nothing arbitrary about the linear weights. They're derived from years of data and thousands of observations.

In other words, the standard errors associated with a run value just aren't big so really the potential to have bias introduced via this proposed mechanism (assigning run values to an event based upon the average number of runs that score from the event given the base/out state till the end of an inning over many years versus actually tallying the true runs) really isn't great provided enough defensive innings.

Here's another consideration. Assigning true run values actually confounds things because it adds the element of chance which is completely out of the defender's control (i.e. the same problem associated with rbis for instance).

Linear weights aren't arbitrary, no, but again, too much context is ignored with respect to the rest of the roster's defensive acumen, pitchers, etc.

Scrap Irony
06-14-2010, 08:02 PM
If we don't compare to every other player, how do we know what average is?

Why should I care about average, nate? I want to know how many runs Gomes has cost the team with his glove and how many runs his bat has added to the team.

That's all.

UZR doesn't do that. It ranks him (subjectively) against other LFers, then puts a number to that ranking.

Scrap Irony
06-14-2010, 08:13 PM
Linear weights aren't arbitrary, no, but again, too much context is ignored with respect to the rest of the roster's defensive acumen, pitchers, etc.

A look at Cincinnati's OF is a case in point. Stubbs' speed and defensive ability on balls hit out of zone (31 so far this year) likely means Gomes has allowed Stubbs (his CF and the man who's "responsible" for any ball he can get to) to catch balls he could have gotten, thereby helping his own UZR. Just three or four of those lazy fly balls Gomes could have caught but allowed Stubbs to grab would put Gomes in the average category.

By comparison, St. Louis' CF, Colby Rasmus, has a poor UZR largely because of his poor OOZ numbers (21 so far this season). Meanwhile, Matt Holliday (who has 24 OOZ plays) is positive largely because he takes those 'tweener balls.

Makes little sense to penalize a defensive player when an out is made, regardless of who makes it.

RedsManRick
06-14-2010, 08:25 PM
You keep making remarks like that. Why not just track the actual runs that missed zone balls lead to. That would be a helpful tool. As you say, it either turned into a run or it didn't.

If individual plays led to runs, they would. But the reality of baseball is that runs are scored through the combined efforts of multiple players over multiple plays. If you want to try and assign credit to individuals you have to start looking at fractions of runs.

nate
06-14-2010, 08:39 PM
Why should I care about average, nate?

So you know who's good? I mean, we have to know who's average in any other stat. Certainly, no one is capable of watching a baseball season with no record of what happened and tell you who's good over who's exciting.


I want to know how many runs Gomes has cost the team with his glove and how many runs his bat has added to the team.

Because sometimes misplays lead to actual runs and sometimes they don't. They idea is to isolate the potential impact of the misplay based on a huge amount of run expectations.

1. misplays a single into a double, the man on second scores on a hit by the next batter
2. misplays a ball with a man on second who scores as a result
3. misplays a ball that allows a man on base, the next batter grounds into a DP
4. misplays a ball that turns into a hit, the next batter hits a HR
5. misplays a ball that turns into a hit, the next batter also reaches on error a third batter hits a single

I mean, these are plausible situations in which he contributes to the run expectancy of the other team. It's not all just "1" or "0" because baseball is a team sport. He was _part_ of what allowed the possibility of a run scoring but not always _all_ of it.

Francisco Cordero got a pitching "win" the other night by blowing a save and being fortunate enough to be the pitcher of record when the Reds came back and won it. How has that measured anything useful about what he did?

Again, I gave the example of how many hits it's taken to make a run. It's not so implausible to think that 11 hits has led to 5 runs as you were trying to make it out to be.


That's all.

UZR doesn't do that. It ranks him (subjectively) against other LFers, then puts a number to that ranking.

It's not really that simple (the "puts a number to that ranking"). Simplistic is RBI, pitching wins and saves which rank players subjectively based largely on what teammates have done and puts a number to that ranking.

I think the run values end up correlating to the actual runs rather highly.

nate
06-14-2010, 08:39 PM
And they often come up with not insignificant differences, while using the same data. That seems to be a problem to me.

For example?

Remember, this thread is ULTIMATE!

Scrap Irony
06-14-2010, 08:52 PM
So you know who's good? I mean, we have to know who's average in any other stat. Certainly, no one is capable of watching a baseball season with no record of what happened and tell you who's good over who's exciting.

OPS doesn't rank anyone. Neither does obp, slugging, BBs, whathaveyou. Give me a number based on the runs he's cost and I can judge for myself.



Because sometimes misplays lead to actual runs and sometimes they don't. They idea is to isolate the potential impact of the misplay based on a huge amount of run expectations.


But it doesn't just count misplays, nate. It also counts OOZ plays. In fact, that's the largest portion of why Gomes is so bad defensively according to UZR. If it were misplays, I'd be okay with that.

It's not misplays.

It's plays that he doesn't make in zones that aren't his. (As compared to other LFer who make plays in zones that aren't there's, but without his CF.)

I think it might be more effective a stat if you added all OF plays together and got a total OF "score".

OUReds
06-14-2010, 09:19 PM
For example?

Remember, this thread is ULTIMATE!

There are plenty. Using everyone's favorite left fielder Adam Dunn, there was a 20 run difference between his 2008/2009 UZR rating (-64.3) and his +/- (-45).

Source (http://sports.yahoo.com/mlb/blog/big_league_stew/post/Everything-you-always-wanted-to-know-about-UZR?urn=mlb,212311#remaining-content)

These are systems using similar methodologies and the same data mind you.

edabbs44
06-14-2010, 09:36 PM
OPS doesn't rank anyone. Neither does obp, slugging, BBs, whathaveyou. Give me a number based on the runs he's cost and I can judge for myself.



But it doesn't just count misplays, nate. It also counts OOZ plays. In fact, that's the largest portion of why Gomes is so bad defensively according to UZR. If it were misplays, I'd be okay with that.

It's not misplays.

It's plays that he doesn't make in zones that aren't his. (As compared to other LFer who make plays in zones that aren't there's, but without his CF.)

I think it might be more effective a stat if you added all OF plays together and got a total OF "score".

Or if you looked at those OOZ plays and figured out why the plays weren't made.

RedsManRick
06-14-2010, 09:45 PM
OPS doesn't rank anyone. Neither does obp, slugging, BBs, whathaveyou. Give me a number based on the runs he's cost and I can judge for myself.



But it doesn't just count misplays, nate. It also counts OOZ plays. In fact, that's the largest portion of why Gomes is so bad defensively according to UZR. If it were misplays, I'd be okay with that.

It's not misplays.

It's plays that he doesn't make in zones that aren't his. (As compared to other LFer who make plays in zones that aren't there's, but without his CF.)

I think it might be more effective a stat if you added all OF plays together and got a total OF "score".

Add up all the team's OF and you're basically there, right? Of course, then you wouldn't know if you had 3 average guys or 2 great guys a schmuck. That doesn't help much when you're trying to decide who gets to start or how to improve your team.

nate
06-14-2010, 10:43 PM
There are plenty. Using everyone's favorite left fielder Adam Dunn, there was a 20 run difference between his 2008/2009 UZR rating (-64.3) and his +/- (-45).

Source (http://sports.yahoo.com/mlb/blog/big_league_stew/post/Everything-you-always-wanted-to-know-about-UZR?urn=mlb,212311#remaining-content)

These are systems using similar methodologies and the same data mind you.

Both say Dunn was a terrible fielder those years. So they don't agree, are they wrong in saying Dunn was a dreadful defender? Is one of them way off or are they both off?

Again, I'm not arguing UZR is perfect. I'm arguing that it's good and way better than WoTV.

mth123
06-14-2010, 10:51 PM
Both say Dunn was a terrible fielder those years. So they don't agree, are they wrong in saying Dunn was a dreadful defender? Is one of them way off or are they both off?

Again, I'm not arguing UZR is perfect. I'm arguing that it's good and way better than WoTV.

But if all they reliably tell us is that Dunn was a trerrible defender, I don't need UZR to tell me. I know that much from your dreaded WoTV system. Seems like your saying UZR is really no improvement over that. If so, I'd agree completely.

kaldaniels
06-15-2010, 01:13 AM
WoTV...is there one person on here who thinks that is an adequate way to truly overall judge a player's defense? If so, I stand corrected. If not, it is your typical strawman.

edabbs44
06-15-2010, 06:56 AM
But if all they reliably tell us is that Dunn was a trerrible defender, I don't need UZR to tell me. I know that much from your dreaded WoTV system. Seems like your saying UZR is really no improvement over that. If so, I'd agree completely.

Agreed. Taking a flawed metric like this and forcing its use into an equation that is trying to find out something as exact as run value is highly questionable, at best.

edabbs44
06-15-2010, 07:08 AM
Does anyone have Dewan's fielding values for the Reds this year? I'd be curious to see the differences between the two.

jojo
06-15-2010, 07:53 AM
Does anyone have Dewan's fielding values for the Reds this year? I'd be curious to see the differences between the two.

Here's a partial list examining only the primary starters (i.e. guys who have logged the most innings at each position) and including Dunn as a first baseman:



Name Pos Inn UZR Dewans
Joey Votto 1B 511 3.1 5
B Phillips 2B 561 3 1
Jay Bruce RF 549 0.2 2
Scott Rolen 3B 472 -0.5 -2
Orlando Cabrera SS 521 -1.7 -2
Drew Stubbs CF 498 -2.6 -1
Jonny Gomes LF 389 -4.9 -8

Adam Dunn 2009 540 -14 -18
2010 495 -1 -4


This notion that UZR and Dewan's consistently come up with a different answer and therefore defensive metrics should be rejected is way overblown.... The two systems don't calculate run values the same way so really what would be inappropriate would be to take Dewans and force it into a metric that calculates offensive run values like UZR calculates defensive run values. Just sayin'... Despite their differences regarding how they treat the data and the route they take to get an answer, the two systems generally arrive at the same place. Poking at the outliers on a given year ignores this reality.

bucksfan2
06-15-2010, 08:23 AM
A look at Cincinnati's OF is a case in point. Stubbs' speed and defensive ability on balls hit out of zone (31 so far this year) likely means Gomes has allowed Stubbs (his CF and the man who's "responsible" for any ball he can get to) to catch balls he could have gotten, thereby helping his own UZR. Just three or four of those lazy fly balls Gomes could have caught but allowed Stubbs to grab would put Gomes in the average category.

By comparison, St. Louis' CF, Colby Rasmus, has a poor UZR largely because of his poor OOZ numbers (21 so far this season). Meanwhile, Matt Holliday (who has 24 OOZ plays) is positive largely because he takes those 'tweener balls.

Makes little sense to penalize a defensive player when an out is made, regardless of who makes it.

Im glad you brought this up because I thought about the same thing. Basically what you are saying, and what I was thinking, is that Gomes is unfairly punished for balls that Stubbs gets to that are considered in Gomes zone or in Gomes "out of zone". Any ball that is caught should not go against a guy's UZR.

jojo
06-15-2010, 08:38 AM
Im glad you brought this up because I thought about the same thing. Basically what you are saying, and what I was thinking, is that Gomes is unfairly punished for balls that Stubbs gets to that are considered in Gomes zone or in Gomes "out of zone". Any ball that is caught should not go against a guy's UZR.

Again, it needs to be demonstrated that this is an endemic problem before its just accepted as a valid criticism.

nate
06-15-2010, 08:39 AM
But if all they reliably tell us is that Dunn was a trerrible defender, I don't need UZR to tell me. I know that much from your dreaded WoTV system.

That's only because of familiarity. For example, Jonny Gomes is nearly as bad of a defender as Dunn was when he played LF (with not nearly as good of a bat) yet because he makes "exciting" plays and has had recent success at the plate, the WoTV system ranks his defensive shortcomings as "insignificant" which is obviously wrong. As verification, see his paycheck. If he could play decent defense and exhibit sustained offensive success, he'd be getting paid. He's not. He's a bad to terrible defender.

Period.


Seems like your saying UZR is really no improvement over that.

UZR is an order of magnitude better than WoTV to a degree which can't be adequately described.


If so, I'd agree completely.

I see.

Do you have any real discussion points about defense?

nate
06-15-2010, 08:40 AM
WoTV...is there one person on here who thinks that is an adequate way to truly overall judge a player's defense? If so, I stand corrected. If not, it is your typical strawman.

See one post above yours.

About a week ago, it was offered that because one couldn't remember any terrible plays by Gomes, his defense wasn't bad.

So the answer to your question is, no...there's more than one.

nate
06-15-2010, 08:47 AM
From the primer (http://www.fangraphs.com/blogs/index.php/the-fangraphs-uzr-primer/#2):


In UZR, when a ball is caught and turned into an out by one fielder, no other fielder gets docked any runs.

Seems like there's a lot of good information there!

edabbs44
06-15-2010, 08:51 AM
Again, it needs to be demonstrated that this is an endemic problem before its just accepted as a valid criticism.

There is a problem with that...the average fan doesn't have the resources to actually chart each and every play and compare to what is being used here. Since the burden of proof should lie with the originator of the metric, have they fully demonstrated that this statistic is an endemic solution so that we can treat it as such?

With such a small sample size, it would be easy to isolate which plays have been categorized as outliers. Does this stat actually allow one to dig deeper into what they are saying or is it kind of a black box statistic? If we were given Gomes' outlier plays, for example, we may be able to then get a better measure of what they are saying.

jojo
06-15-2010, 09:32 AM
Since the burden of proof should lie with the originator of the metric, have they fully demonstrated that this statistic is an endemic solution so that we can treat it as such?

Seriously you need to read some of the many primers available.

When formulating an argument (i.e. UZR can't be trusted because of the issue of "poaching"), the burden of proof is actually on the one making the argument.

Concerning the burden the originator of the metric has, with UZR, MGL has written multiple primers (the latest is at fangraphs), he and Tango have written literally hundreds of pages on the Book blog addressing methodology and questions, multiple saber blogs have also written primers, the summarized UZR data is actually available for free and the originator is pretty accessible via email.

The guy has surpassed any reasonable burden that might be expected of him or the metric.

At some point, the person continually profiting supposition about the fatal flaws of the metric should be expected to do a little heavy lifting too.

MGL specifically addresses how fielder credit is assigned in the fangraph primer.

lollipopcurve
06-15-2010, 10:09 AM
From the creator of UZR, just last year:


Again, he forgets about measurement error which is important in defensive evaluation, because we don’t have those neat convenient little buckets, like hits and outs, that we do for offense. We only have, for example, “a hard ground ball hit to the 130 degree vector.” And if anyone thinks that the recording of that piece of data, and thus the “measurement” we need to compute UZR or plus/minus, is perfect, or even close to perfect, well, I have news for you…

The more I read about UZR, the more I see it is still very much a work in progress. They're making adjustments all the time -- for example, it still does not account for "scoops" at first, an amazing omission for that position. And they tweaked the park adjustment at Fenway so that Jason Bay was transformed from a "horrible" defender to an "above average" one.

Factor in the small sample size issue, which is massive, and the fact that folks are trying to use 50 games of defensive "evidence" in making an assessment about how productive Gomes has been this year, and I suspect that any of the UZR talk around here is just a drum circle of people banging their heads against walls.

I suggest other skeptics get off this board and do some reading about UZR in a more objective environment. Your suspicions will be confirmed.

westofyou
06-15-2010, 10:17 AM
Any system that needs a "primer" and can't be explained in two complete sentences is going to be looked at askew, the KISS approach is the everyman approach to stats, that's why OPS works so well to the masses.

Pretending that dogma is the detractor to the stat turns the burden of understanding on the people that the stat is trying to convert, when in fact the burden should be on the stat to be easily examined, compared, and quantified.

Which IMO it currently isn't.

jojo
06-15-2010, 10:22 AM
From the creator of UZR, just last year:



The more I read about UZR, the more I see it is still very much a work in progress. They're making adjustments all the time -- for example, it still does not account for "scoops" at first, an amazing omission for that position. And they tweaked the park adjustment at Fenway so that Jason Bay was transformed from a "horrible" defender to an "above average" one.

Factor in the small sample size issue, which is massive, and the fact that folks are trying to use 50 games of defensive "evidence" in making an assessment about how productive Gomes has been this year, and I suspect that any of the UZR talk around here is just a drum circle of people banging their heads against walls.

I suggest other skeptics get off this board and do some reading about UZR in a more objective environment. Your suspicions will be confirmed.

Again, UZR etc are massive advances so this notion of "work in progress" means "fatal flaw" isn't an appropriate extension of healthy skepticism.

Lets keep in mind that this latest discussion was precipitated by the assertion that Gomes' defense is bad and its bad enough to be a significant drag on his marginal value.

Even the UZR skeptics have to have a hard time arguing with that assertion.

But yes, everyone interested in defensive metrics should read and do their homework.

jojo
06-15-2010, 10:32 AM
Any system that needs a "primer" and can't be explained in two complete sentences is going to be looked at askew, the KISS approach is the everyman approach to stats, that's why OPS works so well to the masses.

Pretending that dogma is the detractor to the stat turns the burden of understanding on the people that the stat is trying to convert, when in fact the burden should be on the stat to be easily examined, compared, and quantified.

Which IMO it currently isn't.

Honest discussion demands that all participants be equally diligent in forming their conclusions.

The point of inventing UZR was to better understand and quantify defensive value and it clearly represents a dramatic advance in this regard. I'm not sure why that objective should be superseded by the arbitrary (and puzzling) edict that for any innovation to be worthwhile it can't be any more complicated than OPS. People seem to have no problem paying for and quoting VORP. The latest permutation of runs created isn't exactly OPS easy either (I bet most people couldn't go from the counting stats to THT's RC). Some of the skeptics of UZR have posted arguments based upon Winshares in the past. James had to write a 4 inch thick book to explain that one....

If the criticism is that UZR is unnecessarily complicated, then show where in fact that is the case. Given the (necessarily) complicated nature of UZR, the inventors have went to great lengths to explain in explicit detail how the numbers are derived and they share all of the data that they can freely (the original play by play data is proprietary). While the average fan can't recreate the UZR data, any fan with the money to purchase the play by play data could actually recreate the data given the extension commentary the inventors of UZR have freely posted on the internet. In other words, even if the average fan doesn't have the resources to recapitulate the stat, all that prevents them from understanding the metric and to accurately appreciate its strengths and weaknesses in order to appropriately interpret it is a simple google search away. I'm not sure what else the inventors could further do to meet their "burden".

nate
06-15-2010, 10:37 AM
Any system that needs a "primer" and can't be explained in two complete sentences is going to be looked at askew, the KISS approach is the everyman approach to stats, that's why OPS works so well to the masses.

Pretending that dogma is the detractor to the stat turns the burden of understanding on the people that the stat is trying to convert, when in fact the burden should be on the stat to be easily examined, compared, and quantified.

Which IMO it currently isn't.

No, it isn't.

Hence, this thread to discuss it.

I understand it's not an everyman issue and I'm not saying Thom should discuss the finer points of UZR in between chapters of Cowboy's latest trencherman endeavors. But that's kind of what we're here for, to discuss the Reds and baseball in detail. And if the detractors flippantly dismiss issues they're ignorant of, they should be pointed to the primer and asked to at least RTFM (read the fine manual.) Otherwise their contribution is simply noise.

lollipopcurve
06-15-2010, 10:39 AM
Again, UZR etc are massive advances so this notion of "work in progress" means "fatal flaw" isn't an appropriate extension of healthy skepticism.

OK, now you're the healthy skepticism authority, too!

Sorry, jojo, my skepticism is as healthy as ever when it comes to UZR. UZR is a cave drawing compared to the elegance that defensive stats will someday achieve with the help of advanced instruments that track ball and player speed and movement, and other factors such as wind currents, with remarkable precision.

Yes, UZR is an advance, but that does not make it authoritative. The fact that folks on here are trying to foist 50-game samples on the skeptics, knowing full well that amounts to a pathetically small sample, is an example of an argument being made on disingenuous grounds.

You guys need to be more forthcoming when it comes to the shortcomings of the systems you're upholding, IMO.

lollipopcurve
06-15-2010, 10:42 AM
the inventors have went to great lengths to explain in explicit detail how the numbers are derived

Ouch.

jojo
06-15-2010, 10:54 AM
OK, now you're the healthy skepticism authority, too!

Sorry, jojo, my skepticism is as healthy as ever when it comes to UZR. UZR is a cave drawing compared to the elegance that defensive stats will someday achieve with the help of advanced instruments that track ball and player speed and movement, and other factors such as wind currents, with remarkable precision.

Yes, UZR is an advance, but that does not make it authoritative. The fact that folks on here are trying to foist 50-game samples on the skeptics, knowing full well that amounts to a pathetically small sample, is an example of an argument being made on disingenuous grounds.

You guys need to be more forthcoming when it comes to the shortcomings of the systems you're upholding, IMO.

It's really this simple. Its the best available right now and it's good enough to not be thrown out with the bath water. In other words, it meaningfully informs discussions about player value. When there is a better way you can be sure that there won't be a dogmatic clinging to the current permutation.

I'm not sure why this notion is so controversial.

Frankly the notion that Gomes' defense is below average seems a surreal starting point for the rehashing of this issue.

jojo
06-15-2010, 10:56 AM
Ouch.

Seriously, this point is inarguable.

jojo
06-15-2010, 10:58 AM
You guys need to be more forthcoming when it comes to the shortcomings of the systems you're upholding, IMO.

Please, just search my posts on this issue and I think you'll find that statement to be inaccurate.

edabbs44
06-15-2010, 11:01 AM
Again, UZR etc are massive advances so this notion of "work in progress" means "fatal flaw" isn't an appropriate extension of healthy skepticism.

The part that bothers me about this is that a stat accepted by most as a work in progress isn't treated as such. Some posters on here use the stat like it is a finished product, which is pretty misleading.

The view that most of Gomes' offensive production is currently wiped out by his defensive shortcomings could easily be refuted 2 years from now by tweaks (or "progress") made to the same statistic used to prove that same assertion. If that is a possibility (and it seems to be a real possibility, just look at Jason Bay), then how can the stat be used as a legitimate measurement right now? Just because it is better than what was there before?

jojo
06-15-2010, 11:10 AM
The view that most of Gomes' offensive production is currently wiped out by his defensive shortcomings could easily be refuted 2 years from now by tweaks (or "progress") made to the same statistic used to prove that same assertion.

WAR isn't stating that conclusion. Right now WAR suggests Gomes' negative defensive value offsets his total marginal offensive value by about a third. Also keep in mind, due to the replacement level baseline being used, there is quite a bit of offensive value tied into Gomes that is assumed but not stated.

This just isn't a valid criticism.

bucksfan2
06-15-2010, 11:18 AM
OK, now you're the healthy skepticism authority, too!

Sorry, jojo, my skepticism is as healthy as ever when it comes to UZR. UZR is a cave drawing compared to the elegance that defensive stats will someday achieve with the help of advanced instruments that track ball and player speed and movement, and other factors such as wind currents, with remarkable precision.

Yes, UZR is an advance, but that does not make it authoritative. The fact that folks on here are trying to foist 50-game samples on the skeptics, knowing full well that amounts to a pathetically small sample, is an example of an argument being made on disingenuous grounds.

You guys need to be more forthcoming when it comes to the shortcomings of the systems you're upholding, IMO.

I am not saying that UZR is a bad measure, just one that has its inaccuracies. Even the creator of the stat has talked about its shortcomings. For me its difficult to get a true read on the value of UZR because of its inaccuracies. To further that just because its the best current available defensive stat doesn't mean that overall it is accurate.

What I can't agree with like "Johnny Gomes should be a platoon player" or "Gomes gives all of his offensive value back because of his defense" or "Gomes and Heisey are equal value players". To me this is a prime example of over valuing defense based upon an inaccurate metric. It would be similar to saying Adam Dunn is a bad offensive player because he has a low BA.

lollipopcurve
06-15-2010, 11:24 AM
I am not saying that UZR is a bad measure, just one that has its inaccuracies. Even the creator of the stat has talked about its shortcomings. For me its difficult to get a true read on the value of UZR because of its inaccuracies. To further that just because its the best current available defensive stat doesn't mean that overall it is accurate.

What I can't agree with like "Johnny Gomes should be a platoon player" or "Gomes gives all of his offensive value back because of his defense" or "Gomes and Heisey are equal value players". To me this is a prime example of over valuing defense based upon an inaccurate metric.

Agree.

jojo
06-15-2010, 11:44 AM
What I can't agree with like "Johnny Gomes should be a platoon player"

This is an honest question concerning Gomes/platooning-where is this screwy?

http://www.redszone.com/forums/showpost.php?p=2119806&postcount=651


"Gomes gives all of his offensive value back because of his defense"

Again that's not a valid conclusion based upon Gomes' current WAR so it really shouldn't be used as an example of a flaw in WAR.

Homer Bailey
06-15-2010, 11:52 AM
See one post above yours.

About a week ago, it was offered that because one couldn't remember any terrible plays by Gomes, his defense wasn't bad.

So the answer to your question is, no...there's more than one.

Assuming this post is referring to my comments, then this post is factually incorrect.

My case was that because Gomes hadn't directly cost the Reds runs with any plays that I could recall (at the time, this was before the gaffe in STL), that WAR was unfairly stating Gomes's value to the Reds, because it was docking him defensive points based on runs that never scored.

I never ever ever said that Gomes wasn't a bad defender, and never said "because I can't remember a bad play, his defense isn't bad."

Falls City Beer
06-15-2010, 11:53 AM
This thread really was a bunch of people talking past each other. Who could have guessed? :)

VR
06-15-2010, 11:59 AM
This thread really was a bunch of people talking past each other. Who could have guessed? :)

You must be new to RZ pal. :)

nate
06-15-2010, 12:02 PM
Assuming this post is referring to my comments, then this post is factually incorrect.

My case was that because Gomes hadn't directly cost the Reds runs with any plays that I could recall (at the time, this was before the gaffe in STL), that WAR was unfairly stating Gomes's value to the Reds, because it was docking him defensive points based on runs that never scored.

If you can't remember it, how do you know? How do you know his defense didn't allow a baserunner that eventually scored? I've already shown that he hasn't made 11 plays in zone and how those plays could lead to 5 runs. The Reds have scored runs at a rate higher than that this year (fewer hits per run.)


I never ever ever said that Gomes wasn't a bad defender, and never said "because I can't remember a bad play, his defense isn't bad."

But you seem to be making that argument that defense is only bad when it's memorably bad. The subtleties of _just_ missing a play or a slight misplay that allows a baserunner to get on, advance or even score are often beyond the total recall capacities of the human brain.

UZR doesn't have this limitation.

dougdirt
06-15-2010, 12:07 PM
Sorry, jojo, my skepticism is as healthy as ever when it comes to UZR. UZR is a cave drawing compared to the elegance that defensive stats will someday achieve with the help of advanced instruments that track ball and player speed and movement, and other factors such as wind currents, with remarkable precision.


Why do we need to know how far and fast the player ran? They either made the out or they didn't. The fact that they had to run far or not doesn't change that. Positioning is a part of defense. If Chase Utley only has to run 8 feet to make an out on a play Brandon Phillips had to run 12 feet on, both guys made the out (assuming exact same hit speed and played location of course).

Scrap Irony
06-15-2010, 12:13 PM
From the primer (http://www.fangraphs.com/blogs/index.php/the-fangraphs-uzr-primer/#2):


In UZR, when a ball is caught and turned into an out by one fielder, no other fielder gets docked any runs.

Seems like there's a lot of good information there!

But he doesn't get "credit" for it either-- which was my point all along. If Gomes becomes a "ball hog" a total (at this point) of four or five times on lazy fly balls hit between he and Stubbs, his UZR would become average to just below.

Is UZR better than a fielder's reputation and/or WoTV? Sure.

But it's a flawed metric that is susceptible to wild swings over the course of a few ABs, that is still being worked on and fine-tuned, and is highly subjectable at its core.

In short, it (and any metric that also uses UZR, such as WAR) shouldn't be trusted as the end-all, be-all of player value. Too many questions.

Scrap Irony
06-15-2010, 12:15 PM
Why do we need to know how far and fast the player ran? They either made the out or they didn't. The fact that they had to run far or not doesn't change that. Positioning is a part of defense. If Chase Utley only has to run 8 feet to make an out on a play Brandon Phillips had to run 12 feet on, both guys made the out (assuming exact same hit speed and played location of course).

If Phillips is positioned by his bench and doesn't therefore make the play, why should he be docked points for a play "in his zone"?

It's not Phillips' fault that his bench positioned him whre they did. Yet he is given negative value.

Homer Bailey
06-15-2010, 12:36 PM
If you can't remember it, how do you know? How do you know his defense didn't allow a baserunner that eventually scored? I've already shown that he hasn't made 11 plays in zone and how those plays could lead to 5 runs. The Reds have scored runs at a rate higher than that this year (fewer hits per run.)

I believe that I can be quoted that as saying "I can't not remember any plays over the last month that Gomes has failed to make that have cost the Reds runs" (this was pre-gaffe in STL). I conceded that I had not seen ALL of the games, and opened it up for ANYONE to mention a play that cost the Reds runs, and no one mentioned a single play. If you took this as me saying he is a good defender, then you are misunderstanding my point. If he didn't directly cost the Reds runs because of his defense, I don't see why his overall value should take a hit. This does not mean I don't think he won't cost the Reds runs in the future, or that I think he will be a good defender.

And addressing the whole "zone" thing. How do we know that balls in his "zone" were even catchable? What if 3 if them are line drives that fall just in front of him that weren't worth the risk of trying to catch them? I just don't think the fact that he didn't get to 11 balls in his zone should count as 5 runs docked against his defensive value. It could be way more than that, but it could also be way less. The whole point of my comments were that, to my knowledge, his D had not cost the Reds runs, and I'm well aware that UZR does not factor in whether or not runs were actually scored. Going forward, the next 11 balls in the zone that he does not get to, may cost the Reds 50 runs! Or it may cost them 2. My feeling, so far, is that the value is towards the low end so far. To my knowledge, there are no defensive metrics that factor in actual runs scored.




But you seem to be making that argument that defense is only bad when it's memorably bad. The subtleties of _just_ missing a play or a slight misplay that allows a baserunner to get on, advance or even score are often beyond the total recall capacities of the human brain.

UZR doesn't have this limitation.

No, that's not the argument at all. If Gomes was making poor defensive plays on a constant basis, and it was constantly costing the Reds runs, I would be glad to admit that Gomes's defense was costing the Reds. I'm just not seeing that so far this year. It does not mean I think he's a good defender.

Without looking at numbers, we can determine that Gomes is a poor defender. Without looking at numbers, would anyone say that Gomes's defense is significantly hurting the Reds? Or are we assuming his defense is costing the Reds runs by looking at UZR?

dougdirt
06-15-2010, 12:37 PM
If Phillips is positioned by his bench and doesn't therefore make the play, why should he be docked points for a play "in his zone"?

It's not Phillips' fault that his bench positioned him whre they did. Yet he is given negative value.

For the same reason he is docked points when he has to 'hit it to the right side' or hits a crappy pitch because the bench called for a hit and run and he had to flail at a poor pitch to save the runner and weakly hits the ball and gets out.

Scrap Irony
06-15-2010, 12:42 PM
I've already shown that he hasn't made 11 plays in zone and how those plays could lead to 5 runs. The Reds have scored runs at a rate higher than that this year (fewer hits per run.)


You've shown that it's possible, nothing more. How many times have Gomes' gaffes cost the Reds runs? How many runs?

If it's five or more, I'l gladly shut up and accept UZR as the stat you think it is.

Raisor
06-15-2010, 12:47 PM
You've shown that it's possible, nothing more. How many times have Gomes' gaffes cost the Reds runs? How many runs?

If it's five or more, I'l gladly shut up and accept UZR as the stat you think it is.

It's not just five or more. It's those five PLUS the number (x) that an "average" left fielder has made so far, PLUS the positional adjustment that a left fielder is penalized just for being a LFer..

nate, you're fond of saying "show your math". But I don't recall ever seeing the math with UZR.

nate
06-15-2010, 01:02 PM
But he doesn't get "credit" for it either-- which was my point all along. If Gomes becomes a "ball hog" a total (at this point) of four or five times on lazy fly balls hit between he and Stubbs, his UZR would become average to just below.

A reliable sample size of UZR wouldn't be greatly affected by this.


Is UZR better than a fielder's reputation and/or WoTV? Sure.

But it's a flawed metric that is susceptible to wild swings over the course of a few ABs,

Which is why we use a lot of reps and regress the data.


that is still being worked on and fine-tuned, and is highly subjectable at its core.

And yet still infinitely better than WoTV.


In short, it (and any metric that also uses UZR, such as WAR) shouldn't be trusted as the end-all, be-all of player value.

I'm not arguing that it's perfect, I'm arguing that it's good and WoTV is terrible.


Too many questions.

So let's talk about them.

nate
06-15-2010, 01:05 PM
If Phillips is positioned by his bench and doesn't therefore make the play, why should he be docked points for a play "in his zone"?

Is this likely to happen very often? If it does, I think we'd see a lot of turnover in the bench coaches.


It's not Phillips' fault that his bench positioned him whre they did. Yet he is given negative value.

If an an average 2B makes some 700 plays in a season and UZR requires a 3 year sample, that's 2100-ish plays. How many times does defensive positioning "rob" him and keep the data from being "good?"

Serious question.

edabbs44
06-15-2010, 01:08 PM
This is an honest question concerning Gomes/platooning-where is this screwy?

http://www.redszone.com/forums/showpost.php?p=2119806&postcount=651



Jojo, humor me for a minute...if you were to remove 2006-2008, would you consider him to be a full time player?

edabbs44
06-15-2010, 01:14 PM
It's not just five or more. It's those five PLUS the number (x) that an "average" left fielder has made so far, PLUS the positional adjustment that a left fielder is penalized just for being a LFer..

nate, you're fond of saying "show your math". But I don't recall ever seeing the math with UZR.

That part bothers me as well, especially in a 50 game sample.

nate
06-15-2010, 01:25 PM
I believe that I can be quoted that as saying "I can't not remember any plays over the last month that Gomes has failed to make that have cost the Reds runs" (this was pre-gaffe in STL). I conceded that I had not seen ALL of the games, and opened it up for ANYONE to mention a play that cost the Reds runs, and no one mentioned a single play.

That doesn't prove anything though. All it says is no one could remember/be bothered to post about it.


If you took this as me saying he is a good defender, then you are misunderstanding my point. If he didn't directly cost the Reds runs because of his defense, I don't see why his overall value should take a hit.

He's missed 11 plays in his zone. That means someone got on base because of his defense. I'd say the chance of a run or runs scoring in 11 missed opportunities is high.


This does not mean I don't think he won't cost the Reds runs in the future, or that I think he will be a good defender.

And addressing the whole "zone" thing. How do we know that balls in his "zone" were even catchable? What if 3 if them are line drives that fall just in front of him that weren't worth the risk of trying to catch them?

That is explained in the primer.


I just don't think the fact that he didn't get to 11 balls in his zone should count as 5 runs docked against his defensive value.

Again, you can't remember what the plays were/didn't see them so how do you know?


It could be way more than that, but it could also be way less.


It could be a little more, a little less, exactly the same, totally off.

However, millions of baseball events have been measured and correlate at a high rate with the actual run events.


The whole point of my comments were that, to my knowledge, his D had not cost the Reds runs,

But your knowledge is faulty. It doesn't remember, it doesn't pick up on the subtleties, you said you don't even watch all the games.

How can your knowledge be relied on?


and I'm well aware that UZR does not factor in whether or not runs were actually scored. Going forward, the next 11 balls in the zone that he does not get to, may cost the Reds 50 runs!

That seems unlikely.


Or it may cost them 2.

Mock if you must but I already showed how 5 runs on 11 hits is actually a higher rate than the Reds are scoring so far this year.


My feeling, so far, is that the value is towards the low end so far. To my knowledge, there are no defensive metrics that factor in actual runs scored.

Again, how would this work?

*LF misplays an single into a double, the next batter hits a home run. How many runs did the LF allow to score?

*LF misplays a ball that allows a runner on third to score. How many runs did the LF allow?

In the first case, although he allowed the baserunner to get into "scoring position," it wasn't his fault he scored. How do you divvy up the run he allowed on but the pitcher allowed to score?

In the second case, the LF didn't allow the runner to get to third but he allowed him in. How do you divvy the run that someone else allowed on but he allowed to score?


No, that's not the argument at all. If Gomes was making poor defensive plays on a constant basis, and it was constantly costing the Reds runs, I would be glad to admit that Gomes's defense was costing the Reds. I'm just not seeing that so far this year.

You said you don't even watch all the games. You don't even have the benefit of the WoTV system. The LoR (listen on radio) system is really maybe not so good.


It does not mean I think he's a good defender.

Without looking at numbers, we can determine that Gomes is a poor defender.

But it doesn't seem that's true. You don't WoTV all the games. You haven't seen them all and you're not comparing him to anyone else. You say he hasn't cost the Reds any runs but you say this in the same breath that you say you LoR sometimes and WoTV sometimes.


Without looking at numbers, would anyone say that Gomes's defense is significantly hurting the Reds?

An excellent point and one I've made often. The human remembers select peaks and valleys much better than the numerous, undifferentiable, subtle (I made that word up) events that make up the vast majority of a player's defensive profile.


Or are we assuming his defense is costing the Reds runs by looking at UZR?

I'm not assuming, I'm saying. His glove causes his bat to be less valuable.

nate
06-15-2010, 01:27 PM
It's not just five or more. It's those five PLUS the number (x) that an "average" left fielder has made so far, PLUS the positional adjustment that a left fielder is penalized just for being a LFer..

nate, you're fond of saying "show your math". But I don't recall ever seeing the math with UZR.

I say, "show your work."

Math is icky like a hug from nate on a hot day! :cool:

There is some math in the primer.

lollipopcurve
06-15-2010, 01:31 PM
I'm not arguing that it's perfect, I'm arguing that it's good and WoTV is terrible.

Disagree. I'll take the opinion of a scout who watches a full season of a team's games on TV over a year of UZR data. Every day of the week.

dougdirt
06-15-2010, 01:38 PM
Disagree. I'll take the opinion of a scout who watches a full season of a team's games on TV over a year of UZR data. Every day of the week.

A scout would laugh at the idea of judging defense on tv.

edabbs44
06-15-2010, 01:39 PM
Mock if you must but I already showed how 5 runs on 11 hits is actually a higher rate than the Reds are scoring so far this year.

That was kind of unfair math, since Gomes' defense isn't letting up HRs while your hit/run ratio for the Reds this year included their HRs. If you removed the HRs and HR related runs from the Reds' overall hits and runs, I'd think that it would take more than 1.87 hits to "create" a run.

dougdirt
06-15-2010, 01:40 PM
That was kind of unfair math, since Gomes' defense isn't letting up HRs while your hit/run ratio for the Reds this year included their HRs. If you removed the HRs and HR related runs from the Reds' overall hits and runs, I'd think that it would take more than 1.87 hits to "create" a run.

Yet the runners he allows with his defense can score on HR's.

jojo
06-15-2010, 01:40 PM
A scout would laugh at the idea of judging defense on tv.

And that comparison (a scout vs UZR) isn't even what is being debated in this thread.

lollipopcurve
06-15-2010, 01:44 PM
A scout would laugh at the idea of judging defense on tv.


www.mlbtraderumors.com/2010/05/the-growing-role-of-video-for-mlb-scouts.html

nate
06-15-2010, 01:47 PM
You've shown that it's possible, nothing more. How many times have Gomes' gaffes cost the Reds runs? How many runs?

I don't have access to the play data but I've heard one can buy it. I mean, it seems the only way you'd be satisfied if we could watch the video of each play in question and see what happened afterward.

There's no way that's going to happen.


If it's five or more, I'l gladly shut up and accept UZR as the stat you think it is.

I'll ask you for the third time.

So far this year in the National League, teams have gotten 8863 hits and scored 4480 runs, a ratio of 1.98 hits for every run. If Gomes has allowed 11 hits more than an average fielder would, is it really so farfetched that 5 runs (a ratio of 2.2 hits for each run) would've scored?

In the end, the difficulty of reviewing each play unless you track it yourself as it happens seems impossible. I think the fractional run values associated with every baseball play correlate rather highly and are reliable enough to inspire confidence.

edabbs44
06-15-2010, 01:47 PM
Yet the runners he allows with his defense can score on HR's.

Sure they can, but that is only part of the equation. If Gomes has something to do with HRs being allowed (like he would with hits), I could buy it a little more. Plus, for all the "advanced" statistical analysis being thrown around here you'd expect a better equation than dividing runs by hits, especially when the big run producing plays are of zero fault to the fielder.

Bottom line, it proved nothing.

dougdirt
06-15-2010, 01:48 PM
www.mlbtraderumors.com/2010/05/the-growing-role-of-video-for-mlb-scouts.html

That is for scouting amateur players and its not the same stuff you see on TV either. On TV, you can't see where anyone is positioned, the break the get on the ball or how far they truly had to go on most plays.

dougdirt
06-15-2010, 01:49 PM
Sure they can, but that is only part of the equation. If Gomes has something to do with HRs being allowed (like he would with hits), I could buy it a little more. Plus, for all the "advanced" statistical analysis being thrown around here you'd expect a better equation than dividing runs by hits, especially when the big run producing plays are of zero fault to the fielder.

Bottom line, it proved nothing.

So when a fielder, specifically an outfielder doesn't make a play that goes for a double or triple that another fielder would have made, that is of zero fault to the fielder?

jojo
06-15-2010, 01:55 PM
www.mlbtraderumors.com/2010/05/the-growing-role-of-video-for-mlb-scouts.html

Is a highly trained scout following a player's every move with a video camera really the same as a fan watching the game on tv?

dougdirt
06-15-2010, 01:57 PM
Is a highly trained scout following a player's every move with a video camera really the same as a fan watching the game on tv?
Or to that, even a scout watching the game on tv? It absolutely isn't.

Homer Bailey
06-15-2010, 02:03 PM
That doesn't prove anything though. All it says is no one could remember/be bothered to post about it.

If we're going to play the burden of proof game, then I'll say that UZR doesn't prove that he allowed runs either.



He's missed 11 plays in his zone. That means someone got on base because of his defense. I'd say the chance of a run or runs scoring in 11 missed opportunities is high.

I agree. I just think it's hard/not possible to estimate how many runs he (or anyone) "allowed" to score based on how many balls in the zone he's missed. I also do not accept that all balls in the zone are created equal.


That is explained in the primer.

Fair enough. I've read the primer twice, but still doubt I fully comprehend everything, which is why I enjoy these discussions.


Again, you can't remember what the plays were/didn't see them so how do you know?

I should clarify that I, at minimum, at least see the condensed game for every game played. However, I do not see all games live (I probably see 75% of the plays live).


It could be a little more, a little less, exactly the same, totally off.

However, millions of baseball events have been measured and correlate at a high rate with the actual run events.

Correlate, yes. I accept that. That doesn't mean actual runs though.


But your knowledge is faulty. It doesn't remember, it doesn't pick up on the subtleties, you said you don't even watch all the games.

How can your knowledge be relied on?

I never said my knowledge could be relied on, which is why I opened it to others to "refresh" my memory on other plays that I may have forgotten. Maybe people were too lazy to post, and no that does not prove anything, but I never claimed that "because I never saw it, it never happened." I was open to being corrected, and wasn't. Again, not saying that proves anything, but I was attempting to cover the hole in my argument that you are trying to poke, which is "you can't remember everything," which I agree with.



That seems unlikely.

Of course I was exaggerating.


Mock if you must but I already showed how 5 runs on 11 hits is actually a higher rate than the Reds are scoring so far this year.

No mocking. It may very well be a good estimation. However, it doesn't mean that 5 runs scored because of those 11 hits. It could mean 11 runs, it could mean zero.


Again, how would this work?

*LF misplays an single into a double, the next batter hits a home run. How many runs did the LF allow to score?

*LF misplays a ball that allows a runner on third to score. How many runs did the LF allow?

In the first case, although he allowed the baserunner to get into "scoring position," it wasn't his fault he scored. How do you divvy up the run he allowed on but the pitcher allowed to score?

In the second case, the LF didn't allow the runner to get to third but he allowed him in. How do you divvy the run that someone else allowed on but he allowed to score?

And here is where I believe most of the problems with UZR stem from. It attempts to assign those runs to the defender, when in fact, it is very difficult to quantify. Even with you describing in detail those particular scenarios, it is impossible to assign run values to each player. Imagine how tough it is to do that with actual plays where certain factors can be very subjective!


You said you don't even watch all the games. You don't even have the benefit of the WoTV system. The LoR (listen on radio) system is really maybe not so good.

Addressed above. I may not see all plays in the correct context, but I at least see the condensed games.


But it doesn't seem that's true. You don't WoTV all the games. You haven't seen them all and you're not comparing him to anyone else. You say he hasn't cost the Reds any runs but you say this in the same breath that you say you LoR sometimes and WoTV sometimes.

Again, refer above.



An excellent point and one I've made often. The human remembers select peaks and valleys much better than the numerous, undifferentiable, subtle (I made that word up) events that make up the vast majority of a player's defensive profile.

I've lost track of what point you're referring to. This discussion is wearing me out :p:. But yes, I agree that there are many plays other than the ones I necessarily remember that make up a defensive players profile.



I'm not assuming, I'm saying. His glove causes his bat to be less valuable.

I don't think there is anyone here that will debate that. The extent of which it devalues his bat is what I believe people (like myself) have difficulty agreeing with.

edabbs44
06-15-2010, 02:03 PM
So when a fielder, specifically an outfielder doesn't make a play that goes for a double or triple that another fielder would have made, that is of zero fault to the fielder?

Nope, you are 100% correct. But it is the usage of runs/hits that is wrong.

Let's do it another way, using the runs/hits equation the top 3 run producers in the majors this season are Cano, Prado and Ichiro. Poor Pujols is right behind Andrew McCutchen.

If I started a thread like that, how ridiculous would that be?

lollipopcurve
06-15-2010, 02:05 PM
Is a highly trained scout following a player's every move with a video camera really the same as a fan watching the game on tv?

As I stated, I'd take a scout's opinion, based on observations of a full season of games rendered on TV, than I'd take a season's worth of UZR data.

A season's worth of UZR data is considered less than half of a satisfactory sample, is it not?

jojo
06-15-2010, 02:11 PM
As I stated, I'd take a scout's opinion, based on observations of a full season of games rendered on TV, than I'd take a season's worth of UZR data.

A season's worth of UZR data is considered less than half of a satisfactory sample, is it not?

Where can the average fan get access to such scouting reports?

lollipopcurve
06-15-2010, 02:22 PM
Where can the average fan get access to such scouting reports?

The average fan can see what the scout sees and arrive at his own conclusions.

With a year's worth of UZR data, the average fan is armed with information that is not only useless (due to sample size issues) but incomprehensible.

RedsManRick
06-15-2010, 02:29 PM
People are acting like it's UZR vs. nothing. But that's not the case. Everybody is judging defense somehow. They coming up with player valuations somehow. And particularly for us fans who don't have access to copious scouting reports and aren't trained in the observation of performance, it's UZR versus our best subjective guess based on the games we've seen.

UZR is like Democracy in the famous Churchill quote (http://www.quotationspage.com/quote/24926.html). UZR is the worst way for a fan to assess defensive performance, except for all of the others that have been tried from time to time.

It's not close to perfect. There's plenty of error involved. But it sure beats the subjective judgment of an amateur exposed to a small sample of opportunities and even then never quantified. Even the professional scouts deal with the issues of sample size and a limited ability to benchmark performance.

I'd love to see the UZR detractors sit down and put together a list of the best/worst defenders at a given position and the relative value of that defense. But no, it's much easier to simply claim that defense can be sufficiently assessed without the quantitative machinations and that it's impossible to show otherwise.

RedsManRick
06-15-2010, 02:32 PM
The average fan can see what the scout sees and arrive at his own conclusions.

With a year's worth of UZR data, the average fan is armed with information that is not only useless (due to sample size issues) but incomprehensible.

And what makes you believe the fans own conclusions are any more accurate or useful? At least UZR can identify how little it knows and, when properly handled, can account for that lack of knowledge through regression to the mean.

nate
06-15-2010, 02:33 PM
If we're going to play the burden of proof game, then I'll say that UZR doesn't prove that he allowed runs either.

I think it's less likely that allowing 11 hits above what an average fielder would allow _DOESN'T_ allow any runs.


I agree. I just think it's hard/not possible to estimate how many runs he (or anyone) "allowed" to score based on how many balls in the zone he's missed. I also do not accept that all balls in the zone are created equal.

Do you mean the speed of the ball? If so, it takes that into account:

"The speed of the each batted ball is also considered and is indicated in the data as “slow/soft, medium, or fast/hard”


Fair enough. I've read the primer twice, but still doubt I fully comprehend everything, which is why I enjoy these discussions.


Yes!


I should clarify that I, at minimum, at least see the condensed game for every game played. However, I do not see all games live (I probably see 75% of the plays live).

Correlate, yes. I accept that. That doesn't mean actual runs though.

But it's likely that runs did occur.


No mocking. It may very well be a good estimation. However, it doesn't mean that 5 runs scored because of those 11 hits. It could mean 11 runs, it could mean zero.

And remember, these are 11 hits above what an average fielder misses so it's actually > 11 hits allowing 5 runs. However, in 11 hits in play, it could be between 0 runs (which would mean that whoever got on base never came around to score) or 33 (all 11 hits were triples with the bases loaded.)

I'm not seeing it as being totally crazy that 5 runs scored.


I don't think there is anyone here that will debate that. The extent of which it devalues his bat is what I believe people (like myself) have difficulty agreeing with.

What I don't get is why it's so hard to see 11 hits allowing 5 runs.

Even if one takes away HR totals from the total number of NL hits and runs, the ratio is 2.2:1, the same as this hypothetical 11 hits scoring 5 runs.

lollipopcurve
06-15-2010, 02:34 PM
I'd love to see the UZR detractors sit down and put together a list of the best/worst defenders at a given position and the relative value of that defense. But no, it's much easier to simply claim that defense can be sufficiently assessed without the quantitative machinations and that it's impossible to show otherwise.

What's the purpose of doing that? Is listmaking some kind of skill I need as a fan?

UZR is rough, we all agree. We do not agree about whether it should be used as part of a tabulation that attempts to provide a single number of wins that a player is worth to his team. There is complete disagreement on that issue, and it will not be bridged.

Quantitatively, defensive metrics will get there when the physicists are in charge. We're not there yet.

jojo
06-15-2010, 02:36 PM
The average fan can see what the scout sees and arrive at his own conclusions.

Sure but lets not pretend the two opinions should be weighted similarly


With a year's worth of UZR data, the average fan is armed with information that is not only useless (due to sample size issues) but incomprehensible.

The average fan is way farther ahead with a year's worth of UZR data.

jojo
06-15-2010, 02:38 PM
We do not agree about whether it should be used as part of a tabulation that attempts to provide a single number of wins that a player is worth to his team. There is complete disagreement on that issue, and it will not be bridged.

Here's a simple question. If your position is the one that is on the strongest ground, why does team WAR correlate highly (we're talking r=.9 highly) with RS/RA?

lollipopcurve
06-15-2010, 02:41 PM
And what makes you believe the fans own conclusions are any more accurate or useful? At least UZR can identify how little it knows and, when properly handled, can account for that lack of knowledge through regression to the mean.

A fan can get a sense if a play is difficult or not -- not whether it's in or out of a zone. A fan can tell if an outfielder has a good arm or not. UZR ignores that. A fan can tell if a player is affected by injury. UZR cannot. Stuff like that.

And again, a year's worth of UZR data is less than half of a satisfactory sample size -- true or false?

Raisor
06-15-2010, 02:43 PM
I don't have a real problem with UZR rankings. If it says so and so is 1st at his position and that so and so is number 2, fine. I can agree with that. Where I have a hard time is with the runs that are assigned. With Runs Created, we know what the formula is, and we can test the re****s. Now, we've been told that UZR uses the same process of assigning runs per event, but that, in this case, has to be subjective. If Drew Stubbs makes a catch in CF, how, other then subjectivly, can the person using the zones decide if that catch wasn't made, how many bases would have been lost? I don't have the chart in front of me, but if a double is worth (just for example) .35 runs, then preventing a double should be worth -.35 runs, but who is to say if the ball caught would have been a double, or a single, or what have you?

Someone up thread showed Dunn, a couple years ago, at -30ish runs. In reality, that's -27 due to the 7 runs he's penalized for being a LFer. That's 27 runs on top of whatever the average LFer makes (a number we never get to see, by the way). Now let's say Carl Crawford or someone is plus 30, which would actually be 37 since he started off at -7 due to the LF penalty. it's a difference of 60 runs difference, and it's all subjective since there's no way to know how many actual bases Crawford (or whomever) saved.

jojo
06-15-2010, 02:44 PM
A fan can get a sense if a play is difficult or not -- not whether it's in or out of a zone. A fan can tell if an outfielder has a good arm or not. UZR ignores that. A fan can tell if a player is affected by injury. UZR cannot. Stuff like that.

And again, a year's worth of UZR data is less than half of a satisfactory sample size -- true or false?

The more data, the safer the conclusion.

But a year's worth of UZR data is orders of magnitude more information than watching games on tv.

Seriously, if we're alluding to some type of spectrum, watching games on tv is farther right than listening to them on the radio but it's definitely to the left of surveying advanced defensive metrics.

RedsManRick
06-15-2010, 02:44 PM
Nope, you are 100% correct. But it is the usage of runs/hits that is wrong.

Let's do it another way, using the runs/hits equation the top 3 run producers in the majors this season are Cano, Prado and Ichiro. Poor Pujols is right behind Andrew McCutchen.

If I started a thread like that, how ridiculous would that be?

Maybe this has changed since you posted or maybe you're using different data, but Fangraphs has the top three run producers to date in 2010 as Morneau (39.9 RAR), Cano (37.3), and Rios (33.1). It has Pujols at 12th, .1 of a run behind Joey Votto. How ridiculous! I find it funny when people freak out that leaderboards after 1/3 of a season don't reflect true talent levels.

If one isn't open to the possibility of being wrong, he/she you might as well not even engage in the conversation to begin with.

And since when is "what other posters might think" a good standard for the legitimacy of an argument? It's an overused example, but most the world, especially the presumed experts, thought that Galileo's belief in a heliocentric "solar system" was nuts.

Just because a conclusion doesn't match our initial hypothesis or is controversial doesn't mean that it's necessarily is wrong. It just means we should be open about our process and go to town on figuring out where the process went wrong, if it did at all.

jojo
06-15-2010, 02:46 PM
I don't have a real problem with UZR rankings. If it says so and so is 1st at his position and that so and so is number 2, fine. I can agree with that. Where I have a hard time is with the runs that are assigned. With Runs Created, we know what the formula is, and we can test the re****s. Now, we've been told that UZR uses the same process of assigning runs per event, but that, in this case, has to be subjective. If Drew Stubbs makes a catch in CF, how, other then subjectivly, can the person using the zones decide if that catch wasn't made, how many bases would have been lost? I don't have the chart in front of me, but if a double is worth (just for example) .35 runs, then preventing a double should be worth -.35 runs, but who is to say if the ball caught would have been a double, or a single, or what have you?

Someone up thread showed Dunn, a couple years ago, at -30ish runs. In reality, that's -27 due to the 7 runs he's penalized for being a LFer. That's 27 runs on top of whatever the average LFer makes (a number we never get to see, by the way). Now let's say Carl Crawford or someone is plus 30, which would actually be 37 since he started off at -7 due to the LF penalty. it's a difference of 60 runs difference, and it's all subjective since there's no way to know how many actual bases Crawford (or whomever) saved.

The position adjustment in WAR is independent of UZR. It adjusts the player's total marginal value according to scarcity so that all players are compared on an apples to apples basis.

lollipopcurve
06-15-2010, 02:51 PM
The more data, the safer the conclusion.

But a year's worth of UZR data is orders of magnitude more information than watching games on tv.

So, you'd rather have a year's worth of UZR data about a player -- with no attendant information about the player -- than a year's worth of a scout's observations based on watching all of that player's games on TV, including as much attendant information about that player as the average person would get about the player?

edabbs44
06-15-2010, 02:51 PM
Maybe this has changed since you posted or maybe you're using different data, but Fangraphs has the top three run producers to date in 2010 as Morneau (39.9 RAR), Cano (37.3), and Rios (33.1). It has Pujols at 12th, .1 of a run behind Joey Votto. How ridiculous! I find it funny when people freak out that leaderboards after 1/3 of a season don't reflect true talent levels.

Read my post again.

RedsManRick
06-15-2010, 02:57 PM
Raisor, I'm not sure you're using subjective properly. UZR looks at the zone in the field where Stubbs caught the ball and it looks at how often balls hit in to that zone go for each type of hit based on the actual results of 1,000s of batted balls.

UZR doesn't claim to be measuring actual runs saved/prevented. It uses average values based on the observed results of what happens when a ball is hit to that zone. By their nature, specific observations are going vary around that average.

There is all sorts of room for error there. The angle/velocity of the ball itself can vary (liner vs. pop-up). Maybe the wind was blowing. Maybe the fielder was set up right there to begin with. Maybe the bases were loaded and Juan Pierre was the batter, threatening for a triple if the caught isn't made. Or maybe it was Bengie Molina and he's lucky to wind up on second. All of these (and more) are ways for the specific play being made in that zone to be different from other plays in that zone.

But UZR isn't trying to guess at the specific value of that play in that context. Rather, it is based on the knowledge that, over time, the variation in context comes out in the wash and that most of the factors which affect the value of the play are beyond the control of the fielder (and thus he should not be credited or penalized for). In small samples, it's precisely that variation that makes UZR unreliable. Over time, with enough balls fielded, it starts to stabilize.

That is, players tend to repeat their performances, suggesting that UZR is capturing a real skill. Think of UZR as a best guess as to how many runs the guy prevented/allowed. With a small amount of data, the possible "real" values can vary from that guess by a lot. Over time, those errors tend to cancel each other out and the error bars shrink.

But save for the coding of the zone in which the ball was fielded, nothing here is subjective.

lollipopcurve
06-15-2010, 02:58 PM
If your position is the one that is on the strongest ground, why does team WAR correlate highly (we're talking r=.9 highly) with RS/RA?

I don't know.

What's the correlation of UZR and RA, minus whatever influence a pitcher draws away from UZR?

RedsManRick
06-15-2010, 03:00 PM
Read my post again.

I read it the first time, thanks. You were critiquing a very rough illustrative example which was initially used to merely demonstrate that UZR values were in the range of possibility by taking it to an illogical extreme. You've shown Nate's conversion was very rough and should not be used broadly; congratulations. That doesn't exactly undermine the credibility of UZR values.

RedsManRick
06-15-2010, 03:06 PM
Aside from all of this conversation about the inaccuracies of UZR and the statements based on them (such as Gomes giving back significant value and of he and Heisey possibly being of equal value), I've yet to see somebody explain how the opposing viewpoints are somehow more accurate, more likely to be true because they use purely subjective assessment and valuation methods instead.

How is the "default" position of making such claims about comparative value based on a defensive assessment from only the games we've personally observed necessarily on stronger footing?

lollipopcurve
06-15-2010, 03:06 PM
There is all sorts of room for error there. The angle/velocity of the ball itself can vary (liner vs. pop-up).

To say the least. Once variables like velocity, trajectory, wind and player positioning are built in, 2-dimensional zones will be replaced by 3-dimensional ones. In other words, the zones in which players are actually moving.

When we get there, we can safely rank fielders.

Scrap Irony
06-15-2010, 03:07 PM
I don't have access to the play data but I've heard one can buy it. I mean, it seems the only way you'd be satisfied if we could watch the video of each play in question and see what happened afterward.

There's no way that's going to happen.



I'll ask you for the third time.

So far this year in the National League, teams have gotten 8863 hits and scored 4480 runs, a ratio of 1.98 hits for every run. If Gomes has allowed 11 hits more than an average fielder would, is it really so farfetched that 5 runs (a ratio of 2.2 hits for each run) would've scored?

In the end, the difficulty of reviewing each play unless you track it yourself as it happens seems impossible. I think the fractional run values associated with every baseball play correlate rather highly and are reliable enough to inspire confidence.

I want play-by-play information rather than the generalization of what might happen in most circumstances and assess some sort of fractional run equivalent, especially when the data is so subjective.

Again, is UZR better than WoTV? Yes.

Is it good enough to assess player value reliably, especially over less than half a season?

Obviously not.

Since the sample size is so very small, it is therefore unreliable. Therefore, because UZR over the course of a season is unreliable, WAR is also.

If you could use UZR as a blunt intrument that tells you only who's below average, right around average, and above average, it'd be fine. But, really, a scout (or even a well-versed fan armed with a notepad and a pencil could come close to the same thing.

Of course, I fully expect you to poo-pooh the rest of this post in favor of some sort of semantic argument about scouts, well-versed fans, notepads, and pencils.

Again, discussion* engaged.





* Discussion being defined as I'm right and you're wrong, no matter what proof is given for or againt an argument; similar to discussions regarding politics and religion.

jojo
06-15-2010, 03:16 PM
So, you'd rather have a year's worth of UZR data about a player -- with no attendant information about the player -- than a year's worth of a scout's observations based on watching all of that player's games on TV, including as much attendant information about that player as the average person would get about the player?

I'm mystified as to why you keep going back to its either UZR or a professional scout.

I think my tag line ought to answer any questions about how I feel about the value of scouting and metrics. If the choice was a year of UZR or the scouting report derived from a year long intensive effort to compare Gomes to other left fielders. You know what I'd do? I'd read the scouting report and look at the UZR data. If the two differed dramatically, I'd weight the intensive effort by professional scouts heavier.

RedsManRick
06-15-2010, 03:18 PM
I want play-by-play information rather than the generalization of what might happen in most circumstances and assess some sort of fractional run equivalent, especially when the data is so subjective.

Again, is UZR better than WoTV? Yes.

Is it good enough to assess player value reliably, especially over less than half a season?

Obviously not.

Since the sample size is so very small, it is therefore unreliable. Therefore, because UZR over the course of a season is unreliable, WAR is also.

If you could use UZR as a blunt intrument that tells you only who's below average, right around average, and above average, it'd be fine. But, really, a scout (or even a well-versed fan armed with a notepad and a pencil could come close to the same thing.

I actually agree with this. However, in a case such as the one we're in, where you're trying to decide between two players and (presumably) would want the guy who provides the most overall production, how would you incorporate your assessment of his defensive abilities in to that decision?

My answer is simply that UZR provides us a best guess and that, after regressing to the mean, we should put errors bars around it based on the sample size. When we compare the guys, we don't look at just the average values of those estimates, but the ranges and corresponding likelihoods as well (like PECOTA's percentile system).

With Gomes, we actually have a pretty big sample size suggesting he's a really poor defender in LF. Is he really worth -20 runs per 150 games? Maybe he's just -10. Maybe he's -25. I don't know. I freely admit that I don't know how big those error bars should be at various sample sizes and would love somebody to share that information if they have it.

But I don't know of a better way to estimate the production of a player (including defense) relative to his peers. I would love to get in to a discussion about the best way to do so, if not using WAR or WAR like process.

edabbs44
06-15-2010, 03:20 PM
Aside from all of this conversation about the inaccuracies of UZR and the statements based on them (such as Gomes giving back significant value and of he and Heisey possibly being of equal value), I've yet to see somebody explain how the opposing viewpoints are somehow more accurate, more likely to be true because they use purely subjective assessment and valuation methods instead.

How is the "default" position of making such claims about comparative value based on a defensive assessment from only the games we've personally observed necessarily on stronger footing?

Rick, no one is saying that they have some secret formula that is more accurate than anything else. The problem is that this stat is being treated like it is pinpoint accurate when everyone is in agreement that it likely isn't. If we are 50% of the way there on defensive metrics, instead of acting like we are 100% there right now why don't we shelve it until we get a little closer?

Saying that 2+2=5, while close, is wrong. Even if we don't find out that 2+2=4 until a few years from now, I don't think that I want to try and teach the world that 2+2=5 while using that thought process to further mathematics in other areas.

edabbs44
06-15-2010, 03:28 PM
But I don't know of a better way to estimate the production of a player (including defense) relative to his peers. I would love to get in to a discussion about the best way to do so, if not using WAR or WAR like process.

Is there a standard deviation for UZR right now? Maybe you use that instead of trying to come up with an exact number, since that is where I think a lot of us have an issue.

One thing that struck me from the LF UZR stats is that there were a couple of outliers on each side (examples of CC, Pierre and Gomes, Braun) and everyone else was pretty lumped together. Curious if you can derive anything from that.

nate
06-15-2010, 03:28 PM
I want play-by-play information rather than the generalization of what might happen in most circumstances and assess some sort of fractional run equivalent, especially when the data is so subjective.

Feel free to buy the PBP data and I'll bring the beer.


Again, is UZR better than WoTV? Yes.

Is it good enough to assess player value reliably, especially over less than half a season?

Obviously not.

Alright! High five!


Since the sample size is so very small, it is therefore unreliable. Therefore, because UZR over the course of a season is unreliable, WAR is also.

One can and should regress the UZR data as suggested by it's creator.


If you could use UZR as a blunt intrument that tells you only who's below average, right around average, and above average, it'd be fine. But, really, a scout (or even a well-versed fan armed with a notepad and a pencil could come close to the same thing.

I think UZR is useful to a point that exists between what you describe and excellent.


Of course, I fully expect you to poo-pooh the rest of this post in favor of some sort of semantic argument about scouts, well-versed fans, notepads, and pencils.

Again, discussion* engaged.

Consider having a cream soda; I though we were having a decent exchange.


* Discussion being defined as I'm right and you're wrong, no matter what proof is given for or againt an argument; similar to discussions regarding politics and religion.

Odd non-sequitur but aren't they always?

edabbs44
06-15-2010, 03:29 PM
Anyone know how much the PBP data costs? A few of you have written that it is available for purchase.

jojo
06-15-2010, 03:29 PM
To say the least. Once variables like velocity, trajectory, wind and player positioning are built in, 2-dimensional zones will be replaced by 3-dimensional ones. In other words, the zones in which players are actually moving.

When we get there, we can safely rank fielders.

So you'd support a system that accounted for type of ball classified by trajectory, the speed of the ball, the handedness, speed, and power of the batter, the outs and base runners, the size and configuration of the OF, the speed of the infield, the speed of batted balls in general as influenced by temperature, altitude, and the ground ball percentage of the pitcher?

RedsManRick
06-15-2010, 03:30 PM
Rick, no one is saying that they have some secret formula that is more accurate than anything else. The problem is that this stat is being treated like it is pinpoint accurate when everyone is in agreement that it likely isn't. If we are 50% of the way there on defensive metrics, instead of acting like we are 100% there right now why don't we shelve it until we get a little closer?

Saying that 2+2=5, while close, is wrong. Even if we don't find out that 2+2=4 until a few years from now, I don't think that I want to try and teach the world that 2+2=5 while using that thought process to further mathematics in other areas.

Because 50% > 0%. Why let the perfect be the enemy of the good? The problem with your example is that we don't know "4". All we know is that the formula is X + Y = Z (offense + defense = total). What we really care about is Z. We're pretty sure we've got a handle on X. We're starting to get a handle on Y but we know we've got a ways to go.

Abandoning our best current method simply leaves us with a worse method to use. We're going to use some process, why not use the best one we have?

Personally I'd rather have option A than option B

Option A: 2 + (somewhere between 1 and 3) = somewhere between 3 and 5 (with 4 being our best guess)

Option B: 2 + good = Aww heck, let's just go with 2 and call it a day.

I admit that UZR and WAR are occasionally, if not often, positioned as having greater certainty than they do. However, in so far as they are our best guess, functionally it doesn't make much of a difference. We're still going to come to conclusions based on the best information we can come up with.

I would absolutely love to have more information about the level of error involved in run estimators and start presenting claims with confidence intervals. But until then, I'm going to keep using my best guess, rather than ignoring information or using a worse version of it.

Raisor
06-15-2010, 03:30 PM
Raisor, I'm not sure you're using subjective properly. .

I know what subjective is, thanks.

If I'm misunderstanding how URZ works, it's one thing, but calling me out for this gets you on an ignore list. Enjoy yourself.

westofyou
06-15-2010, 03:43 PM
Anyone know how much the PBP data costs? A few of you have written that it is available for purchase.

http://www.retrosheet.org/game.htm

edabbs44
06-15-2010, 03:47 PM
Because 50% > 0%. Why let the perfect be the enemy of the good? The problem with your example is that we don't know "4". All we know is that the formula is X + Y = Z (offense + defense = total). What we really care about is Z. We're pretty sure we've got a handle on X. We're starting to get a handle on Y but we know we've got a ways to go.

Abandoning our best current method simply leaves us with a worse method to use. We're going to use some process, why not use the best one we have?

Personally I'd rather have option A than option B

Option A: 2 + (somewhere between 1 and 3) = somewhere between 3 and 5 (with 4 being our best guess)

Option B: 2 + good = Aww heck, let's just go with 2 and call it a day.

I admit that UZR and WAR are occasionally, if not often, positioned as having greater certainty than they do. However, in so far as they are our best guess, functionally it doesn't make much of a difference. We're still going to come to conclusions based on the best information we can come up with.

I would absolutely love to have more information about the level of error involved in run estimators and start presenting claims with confidence intervals. But until then, I'm going to keep using my best guess, rather than ignoring information or using a worse version of it.

Yeah but the issue is that we have people running around saying that certain players have certain values and, next year, all of this talk could be completely wrong. Next year UZR could be tweaked and Gomes could be seen as an average left fielder. Or Braun. Or even Crawford. I know that these are the extremes but so be it. And then you could say that all the discussions being had regarding this stat over the past two years were garbage.

Maybe the best way to talk about this currently is that Player X is a (good, bad, indifferent) fielder with (good, bad, indifferent) offensive value. To be honest, I'd rather know that Gomes is a .900 OPSer while not being good in the field than having someone tell me that he is worth X overall, where X may or may not be correct. Because the former is correct. I don't think we really need to drill down to run level, especially when it is potentially erroneous.

lollipopcurve
06-15-2010, 03:50 PM
So you'd support a system that accounted for type of ball classified by trajectory, the speed of the ball, the handedness, speed, and power of the batter, the outs and base runners, the size and configuration of the OF, the speed of the infield, the speed of batted balls in general as influenced by temperature, altitude, and the ground ball percentage of the pitcher?

The precise trajectory, the precise velocity, the wind (if possible) and the exact positioning and route of the fielder are the main components. Get the instruments that can capture that information, and you'll see who's got what defensive skills.

Scrap Irony
06-15-2010, 04:15 PM
I actually agree with this. However, in a case such as the one we're in, where you're trying to decide between two players and (presumably) would want the guy who provides the most overall production, how would you incorporate your assessment of his defensive abilities in to that decision?

My answer is simply that UZR provides us a best guess and that, after regressing to the mean, we should put errors bars around it based on the sample size. When we compare the guys, we don't look at just the average values of those estimates, but the ranges and corresponding likelihoods as well (like PECOTA's percentile system).

With Gomes, we actually have a pretty big sample size suggesting he's a really poor defender in LF. Is he really worth -20 runs per 150 games? Maybe he's just -10. Maybe he's -25. I don't know. I freely admit that I don't know how big those error bars should be at various sample sizes and would love somebody to share that information if they have it.

But I don't know of a better way to estimate the production of a player (including defense) relative to his peers. I would love to get in to a discussion about the best way to do so, if not using WAR or WAR like process.

I'm assuming the Reds, like other teams, have their own defensive data that is not readily available to the general public.

As for fans, I'd rank players in one of five categories-- Gold Glove level, above average glove, average, below average, and poor. No need to give a run evaluation at all, IMO, if you can't give one that's accurate. (And UZR can't.)

If you're comparing Crawford and Gomes, for example, you'd probably say that Gomes has a below average glove and an 850 OPS bat, while Crawford has Gold Glove level defense with 810 OPS bat.

The discussion between fans would then center on the extra 40 or so OPS points and whether they'd equal out to more runs than the difference between Gomes' and Crawford's offense. (Obviously, I'd argue that Crawford was the better player in that instance.)

You might also attempt to ascertain the difference between, say, Juan Pierre, and Gomes. I'd insist Gomes is by far the better player overall. But WAR would argue that Pierre was more valuable, despite his less than 600 OPS so far in 2010.

Since you can't reliably convert UZR to straight run production (prevention, actually), why try? Just discuss them as two different things until either UZR becomes more reliable or another statistic is more sound.

jojo
06-15-2010, 04:21 PM
The precise trajectory, the precise velocity, the wind (if possible) and the exact positioning and route of the fielder are the main components. Get the instruments that can capture that information, and you'll see who's got what defensive skills.

Lets apply these standards to offensive evaluation too when discussing just what it is we do and don't know.

RedsManRick
06-15-2010, 04:34 PM
Yeah but the issue is that we have people running around saying that certain players have certain values and, next year, all of this talk could be completely wrong. Next year UZR could be tweaked and Gomes could be seen as an average left fielder. Or Braun. Or even Crawford. I know that these are the extremes but so be it. And then you could say that all the discussions being had regarding this stat over the past two years were garbage.

I agree. Next year hopefully we'll know a little bit more than we know now. And when we know more, our conclusions will change. I don't think that people need to explicitly preface any statement with "to the best of our currently ability to determine..." -- that's assumed, no?

And while our conclusions next year will be better and possibly different, there are decisions to be made today. So all we can do is use the best information we have.

If I understand you correctly, you're suggesting that the process of UZR itself is so unsettled as to make it less reliable than an intelligent fan's personal subjective assessment. Is that fair? If it is, I guess we'll have to disagree -- I'll trust UZR, as a starting point, over my own guess based on my limited experience and knowledge.



Maybe the best way to talk about this currently is that Player X is a (good, bad, indifferent) fielder with (good, bad, indifferent) offensive value. To be honest, I'd rather know that Gomes is a .900 OPSer while not being good in the field than having someone tell me that he is worth X overall, where X may or may not be correct. Because the former is correct. I don't think we really need to drill down to run level, especially when it is potentially erroneous.

You've hit on the big point which is that in order to make judgments about a player's overall value, we have to have the components in the same unit of measurement. In isolation, knowing Gomes is a .900 OPS guy with poor defense is fine. But if I'm trying to decide between Gomes and an .800 OPS guy with good defense, I need some way of figuring out if the defensive difference is worth the .100 points of OPS.

While it's not perfect, WAR gives us a way to do this. While there is certainly error associated with it, I don't know how I'd do this absent a WAR-like framework. Pre-WAR I just had to guess.

"How much money is Dunn worth?"
"Well, offensively he's a complete stud. Let's give him $15M!"
"But his defense is horrible. No way is he worth that much. I wouldn't give him more than $5M"

We had this argument in 100 different flavors for years. Without some way of putting a run value (and thus a dollar value) on defense, you're stuck. While UZR carries a lot of error in it, I feel more confident using WAR than just guessing.

That UZR maps fairly closely with scouting assessments and that it is a key piece of the WAR model which maps very closely to actual RS/RA in the aggregate gives me confidence that it's on the right track and not systematically biased. The spread might be too big (or too small), but the same could be said about a subjective assessment process.

I wonder how people would react if UZR values were regressed back towards 0 based on the sample size/ level of certainty. So chop all current 2010 UZR values by 70% or something like that. Not that I'm not sure if 0 is the right number to regress towards, probably some existing estimate, even if subjective would be more appropriate. Which is interesting... what if we just classified guys subjectively as very poor (-10), poor (-5), average (0), good (5), very good (10) using our own eyes first? We could use those values as our best guess and then use them straight up or regress observed WAR back to those numbers.

In any event, the basic WAR premise has been tested and seems to be the right one. We can and should continue to debate the best way to measure the defense component. But I don't think you help yourself but tossing it out completely and making a soft decision with general terms like good and bad and a gut valuation.

RedsManRick
06-15-2010, 04:50 PM
I'm assuming the Reds, like other teams, have their own defensive data that is not readily available to the general public.

From what I have read, the insiders who have some knowledge of what data teams have say it runs the gauntlet from much more advanced than UZR to having no proprietary data beyond what's in a scouting report. Those people tend to suggest the Reds are closer to the latter. Obviously we don't know for sure, but Jocketty's reputation would support claim.



As for fans, I'd rank players in one of five categories-- Gold Glove level, above average glove, average, below average, and poor. No need to give a run evaluation at all, IMO, if you can't give one that's accurate. (And UZR can't.)

If you're comparing Crawford and Gomes, for example, you'd probably say that Gomes has a below average glove and an 850 OPS bat, while Crawford has Gold Glove level defense with 810 OPS bat.

The discussion between fans would then center on the extra 40 or so OPS points and whether they'd equal out to more runs than the difference between Gomes' and Crawford's offense. (Obviously, I'd argue that Crawford was the better player in that instance.)

You might also attempt to ascertain the difference between, say, Juan Pierre, and Gomes. I'd insist Gomes is by far the better player overall. But WAR would argue that Pierre was more valuable, despite his less than 600 OPS so far in 2010.

Since you can't reliably convert UZR to straight run production (prevention, actually), why try? Just discuss them as two different things until either UZR becomes more reliable or another statistic is more sound.

Iin your examples, the only way for you to come to a conclusion that Crawford > Gomes > Pierre was to combine offense and defense. Now, you didn't do it mathematically, but you still did it. You couldn't tell me how you weighted those two things exactly and you couldn't put a name or number on the final value, but you did have some concept of it or else you couldn't have said who came out on top.

You say "why try?" and yet, in order to compare two players, we have to combine all of our information about them in some form. So the question becomes, what method of combining information do we want to use? It's not UZR vs. nothing. It's UZR vs. some other method of adding defensive value to the rest of what we know.

That some other method is bound to have error in it too. With UZR, we have some idea about how much error there is. With other methods? Not as much. Maybe that process is more accurate. Maybe it's less. But we really have no way of even figuring that out. At least with UZR & WAR we can show our work. We can see how we arrived at "good" or "bad", how much "good" means relative to an .850 OPS, and then consistently apply those standards. If only other methods of player valuation were so open to critique.

lollipopcurve
06-15-2010, 04:51 PM
I wonder how people would react if UZR values were regressed back towards 0 based on the sample size/ level of certainty. So chop all current 2010 UZR values by 70% or something like that. Not that I'm not sure if 0 is the right number to regress towards, probably some existing estimate, even if subjective would be more appropriate. Which is interesting... what if we just classified guys subjectively as very poor (-10), poor (-5), average (0), good (5), very good (10) using our own eyes first? We could use those values as our best guess and then use them straight up or regress observed WAR back to those numbers.

People would be pretty comfortable with that, probably.

But the thing is, in addition to relaxing the rigidity of the defensive measure (a good thing, IMO), it is also important to consider other factors that carry importance. Is the player healthy? Does the player have the intangibles you want? Is the player playing for a contract? Those kinds of things probably carry more weight than whether a guy is a below average (but not sinfully awful) -- if not decent -- defender (as most would say about O. Cabrera).

lollipopcurve
06-15-2010, 05:02 PM
Lets apply these standards to offensive evaluation too when discussing just what it is we do and don't know.

Sure. But there's already much more granularity in the offensive results. It's not just "recorded an out" and "did not record an out." Plus, the difficulty inherent in hitting a pitched ball in the major leagues is much more consistent across all events than in fielding one. And an important component in being a successful hitter has nothing to do with physics -- that is, the decision to swing or not to swing. A fielder does not have the choice not to field a ball.

RedsManRick
06-15-2010, 05:03 PM
People would be pretty comfortable with that, probably.

But the thing is, in addition to relaxing the rigidity of the defensive measure (a good thing, IMO), it is also important to consider other factors that carry importance. Is the player healthy? Does the player have the intangibles you want? Is the player playing for a contract? Those kinds of things probably carry more weight than whether a guy is a below average (but not sinfully awful) -- if not decent -- defender (as most would say about O. Cabrera).

I think those concerns, along with all the "intangibles" go in to the eventual calculation. With all inputs, we should aspire to measure and incorporate them as best we can.

RedsManRick
06-15-2010, 05:05 PM
Sure. But there's already much more granularity in the offensive results. It's not just "recorded an out" and "did not record an out." Plus, the difficulty inherent in hitting a pitched ball in the major leagues is much more consistent across all events than in fielding one. And an important component in being a successful hitter has nothing to do with physics -- that is, the decision to swing or not to swing. A fielder does not have the choice not to field a ball.

The absence of the swing/don't swing choice dynamic would seem to make defense simpler to me?

While offensive players have to worry about balls/strikes, situational hitting, etc., with defense it's almost always the same -- regardless of circumstance, the goal is the same: catch the ball. That would seem to make defensive measurement easier because we don't have to account for as much.

dougdirt
06-15-2010, 05:08 PM
Nope, you are 100% correct. But it is the usage of runs/hits that is wrong.

Let's do it another way, using the runs/hits equation the top 3 run producers in the majors this season are Cano, Prado and Ichiro. Poor Pujols is right behind Andrew McCutchen.

If I started a thread like that, how ridiculous would that be?

Which runs/hits equation are you using to come up with this?

Scrap Irony
06-15-2010, 05:08 PM
From what I have read, the insiders who have some knowledge of what data teams have say it runs the gauntlet from much more advanced than UZR to having no proprietary data beyond what's in a scouting report. Those people tend to suggest the Reds are closer to the latter. Obviously we don't know for sure, but Jocketty's reputation would support claim.

IIRC, someone (either here or on another sports site) posted that the Reds use their own defensive numbers that are more comprehensive than UZR. The Jocketty comment we'll save for another thread. (And, really, if you were sabre-oriented, wouldn't you want Jocketty to be pro-sabre as well, as he's been successful pretty much wherever he's been?)



Iin your examples, the only way for you to come to a conclusion that Crawford > Gomes > Pierre was to combine offense and defense. Now, you didn't do it mathematically, but you still did it. You couldn't tell me how you weighted those two things exactly and you couldn't put a name or number on the final value, but you did have some concept of it or else you couldn't have said who came out on top.

You say "why try?" and yet, in order to compare two players, we have to combine all of our information about them in some form. So the question becomes, what method of combining information do we want to use? It's not UZR vs. nothing. It's UZR vs. some other method of adding defensive value to the rest of what we know.

That some other method is bound to have error in it too. With UZR, we have some idea about how much error there is. With other methods? Not as much. Maybe that process is more accurate. Maybe it's less. But we really have no way of even figuring that out. At least with UZR & WAR we can show our work. We can see how we arrived at "good" or "bad", how much "good" means relative to an .850 OPS, and then consistently apply those standards. If only other methods of player valuation were so open to critique.

I'm willing to use numbers for offensive production (OPS being a particular favorite), but defense is, at this point, too nebulous to categorize with numbers. Why not just discuss defense individually?

edabbs44
06-15-2010, 05:11 PM
Which runs/hits equation are you using to come up with this?

Ask Nate, he's the one who used it. It was runs divided by hits, I believe.

lollipopcurve
06-15-2010, 05:16 PM
The absence of the swing/don't swing choice dynamic would seem to make defense simpler to me?

While offensive players have to worry about balls/strikes, situational hitting, etc., with defense it's almost always the same -- regardless of circumstance, the goal is the same: catch the ball. That would seem to make defensive measurement easier because we don't have to account for as much.

Because a better can do more (swing or not swing, walk, single, double, etc.), we know more. Like jojo said, the more information, the better the conclusion. Because a fielder can only do 1 of 2 things, the knowledge is more generalized and less reflective of his skill level, IMO.

Bring the physics in, and the granularity will rise to the surface. How good is a guy's jump? His routes? His acceleration? How much ground does he actually cover? How is his timing when leaving his feet? These are all defensive skills we can't measure now, but will probably be able to down the road.

dougdirt
06-15-2010, 05:26 PM
To say the least. Once variables like velocity, trajectory, wind and player positioning are built in, 2-dimensional zones will be replaced by 3-dimensional ones. In other words, the zones in which players are actually moving.

When we get there, we can safely rank fielders.

Why does the player positioning matter? That doesn't tell us if one guy is better than the other because positioning matters. If player B had to run 20 more feet to get to the flyball in Zone 187C than player A to get to the flyball in Zone 187C, the results is the same. Both players made the out. One guy just positioned himself much better. Simply because one guy ran further doesn't make him better if the balls were the same. It just means he positioned himself worse. The wind is the same way.... as a fielder you need to know the wind is blowing and adjust yourself accordingly.

And really, do certain fielders just see a ton more line drives than other fielders over a full season? If you think so, what makes you come to this conclusion?

dougdirt
06-15-2010, 05:28 PM
Bring the physics in, and the granularity will rise to the surface. How good is a guy's jump? His routes? His acceleration? How much ground does he actually cover? How is his timing when leaving his feet? These are all defensive skills we can't measure now, but will probably be able to down the road.

None of those things matter. Did the player catch the ball is what matters. How far he had to go means nothing in terms of the result.

Now if you want to argue that it means his skills are better, sure. But that doesn't mean his results are. It means they could be one day if the guy learns to position himself better, but for the 'right now how good is he', they mean absolutely nothing.

kaldaniels
06-15-2010, 05:33 PM
I think measured skills such as how far a player can go to get the ball would be helpful in predicting future results.

dougdirt
06-15-2010, 05:35 PM
I think measured skills such as how far a player can go to get the ball would be helpful in predicting future results.

They could be or they couldn't be. If a guy never learns to position himself well, all of those skills will be wasted. Some guys never figure it out.

nate
06-15-2010, 05:41 PM
I know I've already said this several times but how many defensive plays are affected by positioning? If the player is positioning himself poorly, that's part of his defense. If not, I would think positioning that results in missed plays gets addressed by getting better scouting, changing whoever sets the defense or a coaching change far before it has a huge impact on the defense.

nate
06-15-2010, 05:43 PM
They could be or they couldn't be. If a guy never learns to position himself well, all of those skills will be wasted. Some guys never figure it out.

Yeah. Willy T comes to mind. Fast but was like Vasco de Gamma out there.

RedsManRick
06-15-2010, 06:03 PM
IIRC, someone (either here or on another sports site) posted that the Reds use their own defensive numbers that are more comprehensive than UZR. The Jocketty comment we'll save for another thread. (And, really, if you were sabre-oriented, wouldn't you want Jocketty to be pro-sabre as well, as he's been successful pretty much wherever he's been?)

I missed that comment. I'd be very interested in hearing more about what the Reds use. As for wanting Jocketty to be pro-saber, I'd love for him to be. But I don't believe he is and that type of comment completely misses the point. It's not like individual stats resulting from the sabermetric approach are immutable religious tenets to be defended to death. They're tools.

I'm for using all good information at our disposal to uncover as much as we can about how the game works and how to play and manage it well. I'm not defending UZR because I have some personal attachment to it. Rather, it's the best publicly available method of defensive valuation that I know of. If there is a better way of measuring defensive performance and incorporating it in to player valuation, I'd love to hear/read about it.

Sabermetrics is precisely based on a recognition that we don't know as much as we once thought we did and that we should be open to all avenues of information from which we can learn. The only condition is that the approach be objective so that it cane be critiqued and improved.


I'm willing to use numbers for offensive production (OPS being a particular favorite), but defense is, at this point, too nebulous to categorize with numbers. Why not just discuss defense individually?

When you decide to sign a player and put him in your lineup, you get both his offense and his defense. You do not have the option to get just his offensive or defensive production. At the point where we have to make decisions about who to put on the roster and who to put in the lineup, we must consider the whole player.

As for the offensive metrics, I would recommend you take a look at wOBA. The slash stats still do a really good job at painting a picture of a player, but for a single summary stat of a guy's offensive ability, wOBA is current best in class and is very accessible. It's like OPS in that it captures both the value of getting on base and the value of how many bases you acquired, except it puts things in proper proportion and uses the familiar OBP scale. It's available on Fangraphs. Check out this link for more info: http://saberlibrary.com/offense/woba/

RedsManRick
06-15-2010, 06:14 PM
I think measured skills such as how far a player can go to get the ball would be helpful in predicting future results.

I agree, this would be interesting. Of course, we'd need to know how well those things translate on the field. Pretty much every on field event requires multiple skills.

HokieRed
06-15-2010, 06:18 PM
Do we have any measures of what a player's defense makes possible or prevents being possible in the way of pitching to hitters? Do we have any ways of quantifying what bad third base play, for instance, adds to a team's overall Runs Against not simply because the 3bman blows chances, lacks range, or is slow to start DP's but because a team's right-handers are reluctant to pitch as far inside to RHers as they might otherwise or as they need to (if they are not, in a small ball park, for instance, to allow the whole plate to become a place where the ball can be hit out of the ballpark)? Or to put this slightly otherwise, do we have any measures by which to quantify what good defense makes possible for pitchers--beyond simply catching the ball, making plays etc.?

jojo
06-15-2010, 06:18 PM
Why not just discuss defense individually?

Certainly but when discussing a player's overall value, it's not possible.

dougdirt
06-15-2010, 06:21 PM
Do we have any measures of what a player's defense makes possible or prevents being possible in the way of pitching to hitters? Do we have any ways of quantifying what bad third base play, for instance, adds to a team's overall Runs Against not simply because the 3bman blows chances, lacks range, or is slow to start DP's but because a team's right-handers are reluctant to pitch as far inside to RHers as they might otherwise or as they need to (if they are not, in a small ball park, for instance, to allow the whole plate to become a place where the ball can be hit out of the ballpark)? Or to put this slightly otherwise, do we have any measures by which to quantify what good defense makes possible for pitchers--beyond simply catching the ball, making plays etc.?

To be perfectly honest, if a team isn't going with their scouting report on a hitter because they are scared that a fielder might have to actually field the ball, they should just quit now because they should never have let that guy put on a glove.

HokieRed
06-15-2010, 06:26 PM
To be perfectly honest, if a team isn't going with their scouting report on a hitter because they are scared that a fielder might have to actually field the ball, they should just quit now because they should never have let that guy put on a glove.

Granted, but I don't think this is that simple for pitchers.

dougdirt
06-15-2010, 06:28 PM
Granted, but I don't think this is that simple for pitchers.

Well then someone needs to talk to them about how stupid they are actually being. Even Edwin Encarnacion, in his worst days, made 9 out of 10 plays just fine. It is stupid to go against a scouting report of a hitter for something like that. And while I do think you are probably correct, someone needs to talk to that guy and show him how wrong he actually is.

Scrap Irony
06-15-2010, 06:36 PM
Sabermetrics is precisely based on a recognition that we don't know as much as we once thought we did and that we should be open to all avenues of information from which we can learn. The only condition is that the approach be objective so that it cane be critiqued and improved.


I agree completely with this. Unfortunately, I think UZR is far too subjective to be taken as gospel.

From jojo, as per discussing defense individually:
Certainly but when discussing a player's overall value, it's not possible

Separate defense from offense (as defensive metrics aren't able to give reliable run estimates). Discuss defensive scouting reports, range numbers, and every other defensive number in your discussion. If possible, chart the runs defensive gaffes, lack of range, and poor arms have cost you. That's what I'd do, anyway.

HokieRed
06-15-2010, 06:37 PM
Would be interesting to know how the overall pattern of an opposing team's hits change from year to year and how it varies from park to park.

RedsManRick
06-15-2010, 06:49 PM
I agree completely with this. Unfortunately, I think UZR is far too subjective to be taken as gospel.

From jojo, as per discussing defense individually:

Separate defense from offense (as defensive metrics aren't able to give reliable run estimates). Discuss defensive scouting reports, range numbers, and every other defensive number in your discussion. If possible, chart the runs defensive gaffes, lack of range, and poor arms have cost you. That's what I'd do, anyway.

But how do you factor that defensive analysis in to your final decision between one player and another? That's where the problem arises. The question of, does player A's defensive advantage over player B make up for offensive deficiency. At the end of the day, you have to combine the two somehow. Even if you don't do it with numbers, you're doing it somehow or else you'd have no basis for a decision.

RedsManRick
06-15-2010, 06:52 PM
Doug, to an extent I agree with you about positioning. However, how much of a player's positioning can be attributed to the player himself versus a coach? I'm not sure we can tease that out and it could be a significant influence.

A comparison might be a basketball player playing in an offense which gets him open 3 point shots. To what degree do we credit the player for his increased scoring and 3-pt %? It might not matter so long as he plays for you, but if you're another GM looking to trade for the guy, or even if you're the player's GM and you don't want to pay the guy like a stud when you think any decent shooter can fill his shoes, you'd want to know the answer.

OUReds
06-15-2010, 06:56 PM
Here's a partial list examining only the primary starters (i.e. guys who have logged the most innings at each position) and including Dunn as a first baseman:



Name Pos Inn UZR Dewans
Joey Votto 1B 511 3.1 5
B Phillips 2B 561 3 1
Jay Bruce RF 549 0.2 2
Scott Rolen 3B 472 -0.5 -2
Orlando Cabrera SS 521 -1.7 -2
Drew Stubbs CF 498 -2.6 -1
Jonny Gomes LF 389 -4.9 -8

Adam Dunn 2009 540 -14 -18
2010 495 -1 -4


This notion that UZR and Dewan's consistently come up with a different answer and therefore defensive metrics should be rejected is way overblown.... The two systems don't calculate run values the same way so really what would be inappropriate would be to take Dewans and force it into a metric that calculates offensive run values like UZR calculates defensive run values. Just sayin'... Despite their differences regarding how they treat the data and the route they take to get an answer, the two systems generally arrive at the same place. Poking at the outliers on a given year ignores this reality.

That's a pretty off-hand dismissal of what to me is a real problem. Not by ANY means do I think defensive metrics should be rejected because of any inconsistency. However, these are the two most widely used systems, using the same data, purporting to measure the same thing, coming up with answers that are far from uniform. That's a problem. Below is a link with some more discussion of the differences between the two systems.

Link (http://www.hardballtimes.com/main/blog_article/from-twitter-uzr-and-plus-minus/)

Good news Nate, there is math!

Edit: There is a link towards the bottom of the above article discussing biases and problems with the current ball-in-play data we have as well.

Here's the Link (http://www.baseballprospectus.com/article.php?articleid=10523)

Money quote

The long-term answer is to get better data. Let’s face it—no matter how much we massage the data, there simply is not a way to objectively define the difference between a fly ball and a line drive. It is inherently a subjective and somewhat arbitrary distinction.

There’s a lot of work being done right now in precision batted-ball tracking, both with cameras and radars. Someday maybe that will percolate down—but it won’t tell us anything about players from before the introduction of those technologies. Failing that, a simple stopwatch could provide more accurate, quantifiable data than what we’re getting right now. And it is possible, to some extent, to review video of past games and get those measurements for players and seasons already passed.

In the meantime, consider this my sabermetric crisis of faith. It’s not that I don’t believe in the objective study of baseball. I’m just not convinced at this point that something dealing with batted-ball data is, at least wholly, an objective study. And where does this leave us with existing metrics that utilize batted-ball data? Again, I’m not sure. I can tell you I’m a lot less comfortable accepting their conclusions—even over a large number of seasons—than I was in the past.

nate
06-15-2010, 07:00 PM
That's a pretty off-hand dismissal of what to me is a real problem. Below is a link with some more discussion of the differences between the two systems.

Link (http://www.hardballtimes.com/main/blog_article/from-twitter-uzr-and-plus-minus/)

Good news Nate, there is math!

Dude, that's not good news!

Unless it involves the Fibonacci series and Bach!

:cool:

dougdirt
06-15-2010, 07:32 PM
Doug, to an extent I agree with you about positioning. However, how much of a player's positioning can be attributed to the player himself versus a coach? I'm not sure we can tease that out and it could be a significant influence.
Sure Rick, but we are then judging ones ability, not what one did on the field. Now if you are trying to determine how one could field if better positioned, then yes, that could come into play. That is where I would lean on the scouting reports of my scouts for how good ones range is.

HokieRed
06-15-2010, 08:36 PM
Positioning is inseparable from pitching approach.

jojo
06-15-2010, 08:55 PM
That's a pretty off-hand dismissal of what to me is a real problem. Not by ANY means do I think defensive metrics should be rejected because of any inconsistency. However, these are the two most widely used systems, using the same data, purporting to measure the same thing, coming up with answers that are far from uniform. That's a problem. Below is a link with some more discussion of the differences between the two systems.

A correlation of .8 between systems seeking to make sense of a complicated set of noise and commotion is much more a victory than a cause for panic.

A correlation of .8 for a cancer therapy would be called a cure.....

OUReds
06-15-2010, 09:01 PM
A correlation of .8 between systems seeking to make sense of a complicated set of noise and commotion is much more a victory than a cause for panic.

A correlation of .8 for a cancer therapy would be called a cure.....

I just can't agree here. Maybe not a great analogy, but there is a better correlation between Batting Average and Runs Scored. Nobody here would use Batting Average to judge a player's overall offensive value. It indicates real problems IMHO.

jojo
06-15-2010, 09:13 PM
I just can't agree here. Maybe not a great analogy, but there is a better correlation between Batting Average and Runs Scored. Nobody here would use Batting Average to judge a player's overall offensive value. It indicates real problems IMHO.

That's because there's things like base runs etc....

UZR can't be too wiggsers because WAR isn't too wiggsers:

http://www.redszone.com/forums/showpost.php?p=1969882&postcount=369

Brutus
06-15-2010, 09:14 PM
I just can't agree here. Maybe not a great analogy, but there is a better correlation between Batting Average and Runs Scored. Nobody here would use Batting Average to judge a player's overall offensive value. It indicates real problems IMHO.

Actually, batting average correlates roughly around .70-.72.

The r2 difference between a .7 and .8 correlation is nearly 15% (49 to 64). That's pretty significantly statistically.

OUReds
06-15-2010, 09:17 PM
Actually, batting average correlates roughly around .70-.72.

The r2 difference between a .7 and .8 correlation is nearly 15% (49 to 64). That's pretty significantly statistically.

Source? Not doubting you, just curious :)

nate
06-15-2010, 09:22 PM
That's a pretty off-hand dismissal of what to me is a real problem. Not by ANY means do I think defensive metrics should be rejected because of any inconsistency. However, these are the two most widely used systems, using the same data, purporting to measure the same thing, coming up with answers that are far from uniform. That's a problem. Below is a link with some more discussion of the differences between the two systems.

Link (http://www.hardballtimes.com/main/blog_article/from-twitter-uzr-and-plus-minus/)

Good news Nate, there is math!

Just wanted to let you know I read the link(s). I wish I had the statistical knowledge to fully grasp everything they talk about but I don't.

Thanks for posting them, though!

OUReds
06-15-2010, 09:33 PM
UZR can't be too wiggsers because WAR isn't too wiggsers:

http://www.redszone.com/forums/showpost.php?p=1969882&postcount=369

Yeah I've read that, and it's interesting, though I think your statement above is farther then I would go. To be clear, I certainly think UZR/Dewan is better then what we've had before, but they often get thrown around with the certainty of offensive stats. They shouldn't be. There are real issues with them yet to be ironed out. Some of them perhaps cannot be ironed out.

Brutus
06-15-2010, 10:05 PM
Source? Not doubting you, just curious :)

A lot of it probably depends on the sample used, but for me, that was the result I got a few years ago when I ran a correlation of stats from 1990 to 2005 (94 excluded). Granted, I don't have the results anymore and I could be off a few points thanks to the time that has passed, but it was in that area.

I do get your point though. All I'm saying is that even a difference in .1 is pretty significant.

OUReds
06-15-2010, 10:08 PM
A lot of it probably depends on the sample used, but for me, that was the result I got a few years ago when I ran a correlation of stats from 1990 to 2005 (94 excluded). Granted, I don't have the results anymore and I could be off a few points thanks to the time that has passed, but it was in that area.

I do get your point though. All I'm saying is that even a difference in .1 is pretty significant.

Thanks Brutus. For reference I was using this (http://www.redszone.com/forums/showthread.php?t=45245) thread.

Agreed on your conclusion of course.

OUReds
06-15-2010, 10:59 PM
More fun with inconsistencies within Batted Ball Data. Using one set of batted ball data (BIS or Baseball Information Systems) to calculate Ichiro's UZR from 2003-2008, we come up with +33 runs above average. Using another set of batted ball data (from STATS) we get +16 runs above average in the same time span. A not insignificant difference of about 4 runs per season.

Which is more accurate? Who knows?

Source, now with even more math! (http://www.insidethebook.com/ee/index.php/site/comments/uzr_on_fangraphs_using_bis_on_ichiro/)

RedsManRick
06-15-2010, 11:07 PM
More fun with inconsistencies within Batted Ball Data. Using one set of batted ball data (BIS or Baseball Information Systems) to calculate Ichiro's UZR from 2003-2008, we come up with +33 runs above average. Using another set of batted ball data (from STATS) we get +16 runs above average in the same time span. A not insignificant difference of about 4 runs per season.

Which is more accurate? Who knows?

Individual examples of inconsistencies aren't terribly compelling. What's the distribution of average error look like? How many players are within 1 run on average? 2? 4? 10? I can only assume you looked for a notable outlier, which really tells us nothing about the systems other than that it's possible to be off by 4 runs per season.

We can tear down the metrics all we want. I think we've clearly established the non perfection of UZR and +/-. (I'm curious how variation there is between different offensive run estimators since people are so confident in them. It's only fair to have a standard of what reliability look like.)

But if we have any interest in having informed discussions about which guy deserves playing time, how much a guy is worth in salary, etc., we're going to have to use something. Which one is right? I don't know. But I don't think I'd trust a number I pulled out of thin air or some nebulous mental math that I can't describe when I add 20 runs of offense plus "good" defense.

OUReds
06-15-2010, 11:09 PM
Individual examples of inconsistencies aren't terribly compelling. What's the distribution of average error look like? How many players are within 1 run on average? 2? 4? 10? I can only assume you looked for a notable outlier, which really tells us nothing about the systems other than that it's possible to be off by 4 runs per season.

We can tear down the metrics all we want. I think we've clearly established the non perfection of UZR and +/-. (I'm curious how variation there is between different offensive run estimators since people are so confident in them. It's only fair to have a standard of what reliability look like.)

But if we have any interest in having informed discussions about which guy deserves playing time, how much a guy is worth in salary, etc., we're going to have to use something. Which one is right? I don't know. But I don't think I'd trust a number I pulled out of thin air or some nebulous mental math that I can't describe when I add 20 runs of offense plus "good" defense.

The answer to the stats questions are in the comments in the link above. I generally agree with your other points, but I don't think it unreasonable to spell out exactly what the imperfections are. After all, Nate said this thread was the ULTIMATE!

jojo
06-15-2010, 11:18 PM
More fun with inconsistencies within Batted Ball Data. Using one set of batted ball data (BIS or Baseball Information Systems) to calculate Ichiro's UZR from 2003-2008, we come up with +33 runs above average. Using another set of batted ball data (from STATS) we get +16 runs above average in the same time span. A not insignificant difference of about 4 runs per season.

Which is more accurate? Who knows?

Source, now with even more math! (http://www.insidethebook.com/ee/index.php/site/comments/uzr_on_fangraphs_using_bis_on_ichiro/)

Tango picked out a a guy known to be an outlier and the the smoking gun was .4 win/yr. So its possible that instead of being a 5.2WAR/yr player over that period he might have been a 4.8 WAR/yr player.

I'm not seeing the fatal issues here.... we'd call him a 5 WAR player either way.

OUReds
06-15-2010, 11:21 PM
"The correlation ("r") between STATS UZR and BIS UZR for 05-08 for all players with at least 50 defensive games was .712 (620 data pairs). For players with at least 100 games, it was .727 (517 data pairs)"

Surely you don't think that correlation is ok? I think it points out a fair amount of subjectivity.

jojo
06-15-2010, 11:26 PM
"The correlation ("r") between STATS UZR and BIS UZR for 05-08 for all players with at least 50 defensive games was .712 (620 data pairs). For players with at least 100 games, it was .727 (517 data pairs)"

Surely you don't think that correlation is ok?

I think Ichiro +/- 4 runs a year is definitely workable.

Again, as a cancer therapy, it's a cure.

Brutus
06-15-2010, 11:27 PM
Thanks Brutus. For reference I was using this (http://www.redszone.com/forums/showthread.php?t=45245) thread.

Agreed on your conclusion of course.

Thanks! Good find.

I'll go with that one. That seems familiar ;)

OUReds
06-15-2010, 11:28 PM
I think Ichiro +/- 4 runs a year is definitely workable.

Again, as a cancer therapy, it's a cure.

LOL you're a stubborn one Jojo. Would you agree that advanced fielding stats should be taken with a healthy dose of skepticism? That's all I'm sayin.

jojo
06-15-2010, 11:40 PM
LOL you're a stubborn one Jojo. Would you agree that advanced fielding stats should be taken with a healthy dose of skepticism? That's all I'm sayin.

You're talking to a guy who has argued for an approach that takes a survey of advanced metrics to find a consensus and then brackets the estimate +/- 5. I've also pointed out numerous times that a problem with WAR is that it relies only on one defensive estimate (actually using this argument to suggest WAR underestimated Dunn's value in 2008) and that in season UZR should be taken with a grain of salt.

Seriously, when it comes to defensive metrics, I'm much more flexible than most. A survey of the archives would demonstrate that while I clearly favor these metrics, I don't present them in a one-sided fashion (i.e. as if they are flawless or absolute) and try pretty hard to focus these kinds of discussions by attempting to eliminate misstatements and misinterpretations (and dare I say, supposition in some cases).

OUReds
06-15-2010, 11:42 PM
You're talking to a guy who has argued for an approach that takes a survey of advanced metrics to find a consensus and then brackets the estimate +/- 5. I've also pointed out numerous times that a problem with WAR is that it relies only on one defensive estimate (actually using this argument to suggest WAR underestimated Dunn's value in 2008) and that in season UZR should be taken with a grain of salt.

Seriously, when it comes to defensive metrics, I'm much more flexible than most. A survey of the archives would demonstrate that while I clearly favor these metrics, I don't present them in a one-sided fashion (i.e. as if they are flawless or absolute) and try pretty hard to focus these kinds of discussions by attempting to eliminate misstatements and misinterpretations (and dare I say, supposition in some cases).

So that's a yes then? :)

jojo
06-15-2010, 11:47 PM
So that's a yes then? :)

Defensive metrics aren't perfect but they're good enough to be pretty useful. I think you've conceded that position so really, im not sure there's a great divide between our actual positions. It's basically pecking at fairly small differences isn't it?

OUReds
06-15-2010, 11:49 PM
Defensive metrics aren't perfect but they're good enough to be pretty useful. I think you've conceded that position so really, im not sure there's a great divide between our actual positions. It's basically pecking at fairly small differences isn't it?

Hmm, I suspect we disagree with exactly how useful they are, but there's certainly no harm in that.

jojo
06-15-2010, 11:51 PM
Hmm, I suspect we disagree with exactly how useful they are, but there's certainly no harm in that.

I'll say this about defensive metrics. I didn't need them to tell me that Gomes is a lousy defender... :p:

OUReds
06-15-2010, 11:54 PM
I'll say this about defensive metrics. I didn't need them to tell me that Gomes is a lousy defender... :p:

Poor Johnny. It's just part of his charm.

kaldaniels
06-16-2010, 12:38 AM
It boils down to this. As we know, UZR basically lets us know how often a defender is able to turn a ball that is hit into certain zones into outs. Thats a very useful thing...no argument here.

It just seems as of right now, people simply want to point to a fielders UZR and make like that is the true value of a fielders defense. It is a piece of the puzzle.

Give a seasoned baseball scout lists of fielding stats such as UZR, then let him go to stock tape of every play a fielder makes during a time period. Then let him make the decision on how valuable a player is on defense. Right now, thats the best way to evaluate defense. We as fans just don't have the camera shots on call to do this the right way.

There are many things on defense that can't be turned into a number...look at Votto's controversial cutoff decision the other night...that play is as big as it gets...yet I don't see a metric that quantifies its value. If in fact it was the wrong decision (don't go there please) a seasoned scout could dock him accordingly.

dougdirt
06-16-2010, 12:40 AM
There are many things on defense that can't be turned into a number...look at Votto's controversial cutoff decision the other night...that play is as big as it gets...yet I don't see a metric that quantifies its value. If in fact it was the wrong decision (don't go there please) a seasoned scout could dock him accordingly.

What would a scout dock him? What would that dock be based upon?

kaldaniels
06-16-2010, 01:03 AM
What would a scout dock him? What would that dock be based upon?

So let's say Votto made a terrible choice on that play. Are you telling me that that should not be considered when evaluating his defense and projecting his future defensive worth? I don't know a quantitative value that could be assigned to it or if there could even be one, but that would be a play where a human being could evaluate better than metrics. How can one disagree with that? I'm not knocking down UZR Doug, it just seems that some fail to acknowlege it is simply a piece of the puzzle.

nate
06-16-2010, 08:39 AM
It boils down to this. As we know, UZR basically lets us know how often a defender is able to turn a ball that is hit into certain zones into outs. Thats a very useful thing...no argument here.

It just seems as of right now, people simply want to point to a fielders UZR and make like that is the true value of a fielders defense. It is a piece of the puzzle.

Kal, I appreciate you sticking in this and try to have a discussion about it.

However, I think we've seen all of the pro-UZR guys repeat over and over that they absolutely do not think what you just said. We know it's not perfect, that doesn't make it useless.


Give a seasoned baseball scout lists of fielding stats such as UZR, then let him go to stock tape of every play a fielder makes during a time period. Then let him make the decision on how valuable a player is on defense. Right now, thats the best way to evaluate defense.

That's not perfect either. Nor does it produce a result that fans can use or appreciate. It's probably why the major league teams use a blend of both.


We as fans just don't have the camera shots on call to do this the right way.

UZR does. And for every play, there are three people "voting" on what happened.


There are many things on defense that can't be turned into a number...look at Votto's controversial cutoff decision the other night...that play is as big as it gets...yet I don't see a metric that quantifies its value.

How often in the 500-600-odd plays a first baseman makes would something like that happen?


If in fact it was the wrong decision (don't go there please) a seasoned scout could dock him accordingly.

Like I said, it's probably a hand full of finger's worth a play like that occurs in a season. If it's happening more, Joey needs Tom Emanski.

nate
06-16-2010, 08:41 AM
So let's say Votto made a terrible choice on that play. Are you telling me that that should not be considered when evaluating his defense and projecting his future defensive worth? I don't know a quantitative value that could be assigned to it or if there could even be one, but that would be a play where a human being could evaluate better than metrics. How can one disagree with that? I'm not knocking down UZR Doug, it just seems that some fail to acknowlege it is simply a piece of the puzzle.

It's maybe some fringey piece of a 500 piece puzzle.

We'll still be able to see what the picture is even if it's missing a border.

kaldaniels
06-16-2010, 08:59 AM
Lets say we are comparing Brandon Phillips and Chase Utley at second base. How do you Nate (or how should a big league scout) decide who is the better defender. Maybe I've got you all wrong...tell me your thought process.

lollipopcurve
06-16-2010, 09:11 AM
None of those things matter. Did the player catch the ball is what matters. How far he had to go means nothing in terms of the result.

Now if you want to argue that it means his skills are better, sure. But that doesn't mean his results are. It means they could be one day if the guy learns to position himself better, but for the 'right now how good is he', they mean absolutely nothing.

Entirely wrongheaded.

Positioning is really the responsibility of the coaching/scouting staff. They have the hit data, and they know how a player will be pitched. Players should be positioned accordingly. If they aren't, it's the fault of the coaching staff/scouting effort.

Doug, with all your talk about how "skillsets" are what matter, it's telling that you appear to have no idea what this entails on defense.

Jumps, routes, speed/quickness, arm. Those are the big four. (Of course, errors -- problems with throwing accuracy/glovework -- can be easily tracked, as they have been forever.) Track the player as precisely as possible and track the ball as precisely as possible and you'll find out all you need to know about how likely a player is to make plays successfully. You'll be able to throw all those zones in the trash.

blumj
06-16-2010, 09:21 AM
Lets say we are comparing Brandon Phillips and Chase Utley at second base. How do you Nate (or how should a big league scout) decide who is the better defender. Maybe I've got you all wrong...tell me your thought process.
If a team really ever gets a chance to choose between Phillips and Utley, aren't there much more likely to be about a dozen other things to base their choice on than who the better defender is between 2 excellent defenders?

edabbs44
06-16-2010, 09:45 AM
What would a scout dock him? What would that dock be based upon?

Dock him in the CS (common sense) column.

Then again, someone may have told him to cut it so it then may go in the CS column of Hernandez. Or Phillips. Or OCab. But not Rolen, he's the man. :)

nate
06-16-2010, 09:53 AM
Lets say we are comparing Brandon Phillips and Chase Utley at second base. How do you Nate (or how should a big league scout) decide who is the better defender. Maybe I've got you all wrong...tell me your thought process.

I'll speak for Nate because I'm most familiar with him. I have no idea about big league scouts but if you want to talk about some cool ways to reharmonize "Blackbird" using subdominant minor function substitution, I'm your guy! :cool:

Since I'm a Reds fan, I (like everyone else) have seen BP play a ton of games so I have a baseline set by my WoTV method.

With Utley, I don't have anywhere close to the number of WoTV reps so I go on:

*any and all defensive stats available like:
**UZR, regressed...rarely ever UZR/150
**any of the UZR components available
**secondary advanced defensive metrics
**usually not errors unless there is an extreme amount on either end

*WoTV - although I didn't re-up for HD so I'm rockin' Trinitron style this year. Quite honestly, I find it almost impossible to tell anything about defense on regular plays because unless it's blatantly bad or spectacularly good, you rarely get a shot of batter and fielder in the same frame.

*a little bit of reputation through my own patented "nonsense filter" (to which I have to thank Thom Brennaman for keeping fully stocked)

For me, when I find something that doesn't that doesn't jibe, I just have to make a decision about what _I_ think. Usually, that comes from erring on the side of who or what has seen the most and has the best experience to judge. I find that's typically UZR along with the other defensive metrics.

jojo
06-16-2010, 10:16 AM
I'll speak for Nate because I'm most familiar with him. I have no idea about big league scouts but if you want to talk about some cool ways to reharmonize "Blackbird" using subdominant minor function substitution, I'm your guy! :cool:

Seriously, why would you want to do that? Why not just play it on the banjo while you're at it? :cool:

kaldaniels
06-16-2010, 10:16 AM
If a team really ever gets a chance to choose between Phillips and Utley, aren't there much more likely to be about a dozen other things to base their choice on than who the better defender is between 2 excellent defenders?

I love when people go out of their way to completely attempt to stain the spirit of a question. Rock on! :thumbup:

nate
06-16-2010, 10:19 AM
Seriously, why would you want to do that? Why not just play it on the banjo while you're at it? :cool:

Hello? This IS for banjo!

:cool:

OUReds
06-16-2010, 10:23 AM
Hello? This IS for banjo!

:cool:

youtube video, or it hasn't happened.

blumj
06-16-2010, 11:43 AM
I love when people go out of their way to completely attempt to stain the spirit of a question. Rock on! :thumbup:
If that's what I was doing, I apologize.

dougdirt
06-16-2010, 01:01 PM
So let's say Votto made a terrible choice on that play. Are you telling me that that should not be considered when evaluating his defense and projecting his future defensive worth? I don't know a quantitative value that could be assigned to it or if there could even be one, but that would be a play where a human being could evaluate better than metrics. How can one disagree with that? I'm not knocking down UZR Doug, it just seems that some fail to acknowlege it is simply a piece of the puzzle.
I didn't say that at all. I am just asking what one docks him and how much it accounts for? What about the next time he does cut a ball and gets a guy trying to go to third? Does he get handed points back? How many? I can't and won't say that a human can't find that play to be a negative more than a system. However I can say that the system can make a better judgement on the other 99% of the plays guys make because well, they can remember them all in comparison to the other 15,000 plays made by players at that same position and a human simply can't.


Entirely wrongheaded.

Positioning is really the responsibility of the coaching/scouting staff. They have the hit data, and they know how a player will be pitched. Players should be positioned accordingly. If they aren't, it's the fault of the coaching staff/scouting effort.
Even if that is so, do we or do we not 'punish' a players batting line when they have to 'hit it to the right side' and weakly ground out to 2B because the coach told him to? We do, because maybe they got a nice inside pitch that they could have hammered, but instead tried to inside out it to get it to go the other way. Its an out. It takes away his avg/obp/slg and thus his 'offensive value' gets dinged some.



Doug, with all your talk about how "skillsets" are what matter, it's telling that you appear to have no idea what this entails on defense.
No, what matters is whether the out was recorded or not. In terms of what happened and the value of what did happen, skillsets do not matter. Results do. If Player X's manager positions him better or if player X himself does it and it leads to him making more outs than Player T who may have more 'skills' but he winds up making fewer outs because he is positioned much more poorly, then Player X was more valuable that season. It doesn't tell us who has more skill. It tells us who had the better season. You seem to continue to miss that point.



Jumps, routes, speed/quickness, arm. Those are the big four. (Of course, errors -- problems with throwing accuracy/glovework -- can be easily tracked, as they have been forever.) Track the player as precisely as possible and track the ball as precisely as possible and you'll find out all you need to know about how likely a player is to make plays successfully. You'll be able to throw all those zones in the trash.
No. You won't. Let me try this again, this time with graphics. Shown below is two players positioning and where they caught the ball at.
http://redsminorleagues.com/images/fielding.gif

Now I took the routes and put them side by side in the lower right corner. The Blue player clearly ran more to make the play, but why is his play better? Both guys got the ball in the same exact spot, one guy just positioned himself better for the batter at hand. Don't worry about the zones at all. Worry about why you seem to think the blue player is the better defender because he had to run further when both guys made the play, because it honestly doesn't make any sense at all.

nate
06-16-2010, 01:12 PM
Again I ask, how often does poor positioning lead to a missed play?

I'm thinking it's not very often.

kaldaniels
06-16-2010, 01:13 PM
Lets say the above play happens 100 times. (Assuming similar trajectory etc of each ball) Player A goes the further route to get the ball 100 times and Player B goes the short route. That tells me I know Player A is capable of going a certain distance to get the ball...but it doesn't tell me if Player B is able to go that far. They did make the same number of outs.

Now flash forward to a game situation. Who would you rather have with 2 outs and the winning run on 2nd base. I'd take the guy that has shown that he can go further to get the ball.

Now thats to the extreme, but that would show that Player B is capable of covering X amount of ground, but it shows that Player A is capable of covering X+Y amount of ground. Isn't that useful in some way?

dougdirt
06-16-2010, 01:19 PM
Lets say the above play happens 100 times. (Assuming similar trajectory etc of each ball) Player A goes the further route to get the ball 100 times and Player B goes the short route. That tells me I know Player A is capable of going a certain distance to get the ball...but it doesn't tell me if Player B is able to go that far. They did make the same number of outs.
Agreed 100%.



Now flash forward to a game situation. Who would you rather have with 2 outs and the winning run on 2nd base. I'd take the guy that has shown that he can go further to get the ball.
I will assume you solely mean on a fly ball so we can ignore who has the better arm. I think the answer depends on who is positioning Player Blue. If it is himself, then I want the other guy. If it is his coach, and he is now on a new team, I may want player Blue. But what we are again discussing is what MIGHT happen, not what DID happen. We don't exactly do that with hitters evaluations. I mean what would Albert Pujols numbers look like if he got pitches to like he were Juan Pierre? We don't play that game. We play the game of 'what did Albert Pujols do at the plate'.



Now thats to the extreme, but that would show that Player B is capable of covering X amount of ground, but it shows that Player A is capable of covering X+Y amount of ground. Isn't that useful in some way?
It is very useful in telling us what may happen. It isn't useful in telling us what did happen and what the value of that players actions WERE. It isn't useful in telling us what the value of that players actions COULD BE. I think that is where the largest difference is on the two sides of this argument. In terms of value, I want to know what DID happen. Not what may be able to happen one day (in terms of applying value to a players season - if I am making a trade for a player, I absolutely want to know what he COULD be able to do, but in terms of his value on the field in a given season, that does not matter).

kaldaniels
06-16-2010, 01:22 PM
Agreed 100%.


I will assume you solely mean on a fly ball so we can ignore who has the better arm. I think the answer depends on who is positioning Player Blue. If it is himself, then I want the other guy. If it is his coach, and he is now on a new team, I may want player Blue. But what we are again discussing is what MIGHT happen, not what DID happen. We don't exactly do that with hitters evaluations. I mean what would Albert Pujols numbers look like if he got pitches to like he were Juan Pierre? We don't play that game. We play the game of 'what did Albert Pujols do at the plate'.


It is very useful in telling us what may happen. It isn't useful in telling us what did happen and what the value of that players actions WERE. It isn't useful in telling us what the value of that players actions COULD BE. I think that is where the largest difference is on the two sides of this argument. In terms of value, I want to know what DID happen. Not what may be able to happen one day (in terms of applying value to a players season - if I am making a trade for a player, I absolutely want to know what he COULD be able to do, but in terms of his value on the field in a given season, that does not matter).

And I say this in all seriousness...am I just way off here? Because I am more interested in what a player will do. I have no problem looking back and judging what has been done, no problem whatsoever. But I want metrics in place to help determine what is going to happen. Maybe my outlook is just different than yours. :beerme:

RedsManRick
06-16-2010, 01:23 PM
Clearly there is agreement that fielding contains multiple components, two of which are the player's positioning and the player's movement.

From the team perspective, or the play perspective, these two components do not need to be separated.

However, if we want to credit the player for his contribution, we need to know the degree to which he is responsible for his positioning. It seems that Doug is suggesting the player is fully responsible whereas lollipop is suggesting the coach is.

If we go on Doug's premise, there is no need to deal with positioning, just the result of the play. UZR and the like operate on this premise, due to the lack of positioning data if not out of principle.

However, if we go on lollipop's premise, we need to isolate positioning so as to not credit the player for something beyond his control. In this model, the difficulty of all plays are comprised of not just the batted ball characteristics but also the initial positioning of the player. If a manager tells the CF to shade in to LF and then the CF gets to a ball that he may not have reached otherwise, it would be silly to assert that the CF is better than one who was directed to play it straight and did not reach that ball.

Until and unless you guys agree on the issue of who is responsible for fielder positioning, you'll continue to argue past each other. I think both ways of thinking have their place -- it depends on what question you're trying to answer.

jojo
06-16-2010, 01:26 PM
Again I ask, how often does poor positioning lead to a missed play?

I'm thinking it's not very often.

To me this is the important point relative to the issue-i.e. how much does positioning actually effect the results?

Alluding to Doug's graph, i'd say that making the blue play would require more "skill" than making the red play based upon range considerations all other things being equal.

dougdirt
06-16-2010, 01:29 PM
And I say this in all seriousness...am I just way off here? Because I am more interested in what a player will do. I have no problem looking back and judging what has been done, no problem whatsoever. But I want metrics in place to help determine what is going to happen. Maybe my outlook is just different than yours. :beerme:

I would rely on scouts to tell me how much range a guy has in terms of what a guy could be able to do.