PDA

View Full Version : Pythagorean Question



Sea Ray
06-01-2008, 10:12 AM
One of the problems I have with saber metrics is how subtly the numbers change. There is an example of this in John Erardi's column today.

http://news.enquirer.com/apps/pbcs.dll/article?AID=/AB/20080601/SPT04/806010438/


One of the essential principles of sabermetrics - and this is our guiding light right here, the Holy Grail - is that in a 162-game season, for every 10 runs your team scores over the number of runs it allows, you get a victory. If you score the same number of runs as you allow, you are going to finish within one or two games of .500, 81-81. It is true every year for all but one or two teams in the majors, and those one or two that defy it are lucky, nothing else.

Last year, your team scored 70 fewer runs than it allowed. We know that your Wharton education allows you to calculate quickly that with a minus-70 run differential, your team finished within one or two games of seven games under .500, 74-88. Sure enough, the Reds were

72-90.

OK but then he changed his method of determining games over/under .500 when he said this:



That's right: You've scored 23 fewer runs than you've allowed, which by run differential should place your team within a game or two of being 2-3 games under .500.

Well, guess what? Through Friday's game, your team is dead on it, 3 games under, 26-29.

I suspect this is shotty journalism on his part, but which is the correct method to determine games under/over .500? In other words is 74-88 7 games under or is it really 14 games under. Is 26-29 3 games under or 1?

PuffyPig
06-01-2008, 11:14 AM
One of the problems I have with saber metrics is how subtly the numbers change. There is an example of this in John Erardi's column today.

http://news.enquirer.com/apps/pbcs.dll/article?AID=/AB/20080601/SPT04/806010438/



OK but then he changed his method of determining games over/under .500 when he said this:




I suspect this is shotty journalism on his part, but which is the correct method to determine games under/over .500? In other words is 74-88 7 games under or is it really 14 games under. Is 26-29 3 games under or 1?

74-88 is 7 games under .500.

RedsManRick
06-01-2008, 12:22 PM
Sea Ray, this is a fair observation. Clearly John did the math two different ways. But rather than use this as evidence towards dismissing the sabermetric approach as trying to trick you through manipulation of the numbers, this is a perfect opportunity for you to get clarification and move the discussion forward. Understand that Erardi wasn't likely trying to mislead, he just messed up. Perhaps you can ask John which way of calculating games below .500 (W-L) or ((W+L)/2)-L) he meant to use and why. Perhaps even ask him to speak further towards the tension you highlight here and he did in his article.

One of the traditions of the sabermetric approach is openness to inquiry and criticism. It's not about any one person being right and lording it over everybody else. Rather it's a search for truth, as lame/pompous that might sound. The idea that subtly changing the numbers is part of the approach is simply misguided. Yes, some people do things differently than others with the numbers, and some people interpret them differently, but nobody is trying to obfuscate through manipulation of the data.

Sabermetrics is like open source programming. Everybody has the same goal, in this case, of trying to better understand the reality of how baseball games are won. It's much more enjoyable to participate in the process than simply to poke holes in it -- though finding the holes, the inconsistencies in the logic is certain a very valuable part.

blumj
06-01-2008, 12:37 PM
I'll admit it, I've always been confused by "X games over/under .500". It doesn't make any sense to me.

UKFlounder
06-01-2008, 12:44 PM
I'll admit it, I've always been confused by "X games over/under .500". It doesn't make any sense to me.

It's strange, but the way I see it is this: the way it was described in the first post above (i.e. 74-88 being 7 games below .500) means you would only need to change the results of 7 past losses to wins (subtract 7 losses & add 7 wins) and suddenly the record is 81-81, i.e. .500, thus only 7 games takes the team to .500 since it affects both columns and 7 times 2 = the 14 game difference in the record

The way I look at it (and I'm clearly no sabrematician) is more of a forward looking approach - i.e. you can't change the results of the past so in order for a 74-88 team to reach .500, they'd have to win 14 straight games to reach .500 at 88-88 which makes them 14 games below .500. Obviously that would be impossible since there are not 176 games in a season.

Maybe I just prefer the overly simplistic way, I don't know, but the second way is how I've always looked at it - what do you need to do now and in the future to get to .500, not what you need to change about the past.

Sea Ray
06-01-2008, 01:11 PM
Perhaps you can ask John which way of calculating games below .500 (W-L) or ((W+L)/2)-L) he meant to use and why. Perhaps even ask him to speak further towards the tension you highlight here and he did in his article.


Sabermetrics is like open source programming. Everybody has the same goal, in this case, of trying to better understand the reality of how baseball games are won. It's much more enjoyable to participate in the process than simply to poke holes in it -- though finding the holes, the inconsistencies in the logic is certain a very valuable part.


I e-mailed John for clarification before posting here on Redszone and I have yet to hear back from him.

I'm an equal opportunity critic. There are strengths to sabermetrics such as OPS stats and such but there are holes as well such as this one and the 24 different ways to figure the RC stat. But this thread is not meant to debate the strengths/weaknesses of sabermetrics. It was to ask a fair question. What is the proper way to figure games over .500. I honestly didn't know. Then there is also the commentary of how saber experts (such as the ones who contributed to this article) lost track of their own sabermetric. Ironic to say the least, at worst it weakens the points being made.

RedsManRick
06-01-2008, 01:29 PM
I guess it depends on how you look at .500, looking forward or looking back.

If you're looking back, you'd want to use the .500 record given the number of games played minus, suggesting that if X games had been wins instead of losses, the team would be .500.

If you're looking forward, all you'd care about is how many games you need to win in order to get to a .500 record.

I'm not sure that either is better per se, though you should certainly be consistent within the course of article. Personally I think the former is more useful, but the latter is easier to calculate and perhaps more intuitive.

Falls City Beer
06-01-2008, 01:39 PM
Sea Ray, this is a fair observation. Clearly John did the math two different ways. But rather than use this as evidence towards dismissing the sabermetric approach as trying to trick you through manipulation of the numbers, this is a perfect opportunity for you to get clarification and move the discussion forward. Understand that Erardi wasn't likely trying to mislead, he just messed up. Perhaps you can ask John which way of calculating games below .500 (W-L) or ((W+L)/2)-L) he meant to use and why. Perhaps even ask him to speak further towards the tension you highlight here and he did in his article.

One of the traditions of the sabermetric approach is openness to inquiry and criticism. It's not about any one person being right and lording it over everybody else. Rather it's a search for truth, as lame/pompous that might sound. The idea that subtly changing the numbers is part of the approach is simply misguided. Yes, some people do things differently than others with the numbers, and some people interpret them differently, but nobody is trying to obfuscate through manipulation of the data.

Sabermetrics is like open source programming. Everybody has the same goal, in this case, of trying to better understand the reality of how baseball games are won. It's much more enjoyable to participate in the process than simply to poke holes in it -- though finding the holes, the inconsistencies in the logic is certain a very valuable part.

This may be true, but no one--sabermetric or scouting--wants to be "wrong."

I'd like to believe every sabermetrician (as well as every scout) was truly as magnanimous as you are saying every sabermetrician is. But the fact is, people fall in love with their theories. It happens. No one can crush his or her ego so to the point where truth obliterates all; and honestly, I think there's a danger in eliminating ego (or whatever narrative-creating mechanism it is that makes some people better at their jobs than others) altogether. Some folks' biases (for lack of a better term) are better than others, IMO.

RedsManRick
06-01-2008, 01:54 PM
This may be true, but no one--sabermetric or scouting--wants to be "wrong."

I'd like to believe every sabermetrician (as well as every scout) was truly as magnanimous as you are saying every sabermetrician is. But the fact is, people fall in love with their theories. It happens. No one can crush his or her ego so to the point where truth obliterates all; and honestly, I think there's a danger in eliminating ego (or whatever narrative-creating mechanism it is that makes some people better at their jobs than others) altogether. Some folks' biases (for lack of a better term) are better than others, IMO.

I agree, nobody wants to be wrong. The differentiator is how one reacts to being challenged as such. Some are more willing to be wrong than others. On balance, I've found sabermetric types to be more open to criticism and critique (assuming said criticism is well founded-- being called a basement dweller who doesn't love the game is not criticism) than anyone else discussing the game. Though as you point out, I'm sure my personal bias plays in to that perception to some extent.

We shouldn't confuse pride in one's work with a unwillingness to accept deserved criticism. It will be interesting to see Erardi's response to Sea Ray's request for clarification, if he does respond.

SteelSD
06-01-2008, 04:46 PM
I e-mailed John for clarification before posting here on Redszone and I have yet to hear back from him.

I'm an equal opportunity critic. There are strengths to sabermetrics such as OPS stats and such but there are holes as well such as this one and the 24 different ways to figure the RC stat. But this thread is not meant to debate the strengths/weaknesses of sabermetrics. It was to ask a fair question. What is the proper way to figure games over .500. I honestly didn't know. Then there is also the commentary of how saber experts (such as the ones who contributed to this article) lost track of their own sabermetric. Ironic to say the least, at worst it weakens the points being made.

Eradi's error really doesn't have anything to do with sabermetrics or, for that matter, Pythag.

Sea Ray
06-01-2008, 04:48 PM
Eradi's error really doesn't have anything to do with sabermetrics or, for that matter, Pythag.

Where did he err?

bucksfan2
06-01-2008, 05:19 PM
I guess it depends on how you look at .500, looking forward or looking back.

If you're looking back, you'd want to use the .500 record given the number of games played minus, suggesting that if X games had been wins instead of losses, the team would be .500.

If you're looking forward, all you'd care about is how many games you need to win in order to get to a .500 record.

I'm not sure that either is better per se, though you should certainly be consistent within the course of article. Personally I think the former is more useful, but the latter is easier to calculate and perhaps more intuitive.

Isn't it kind of pointless to look back during the season. IMO it is kind of pointless to look at a 26-29 record and say well "if we won this game we would be here".

As for Erardi's column I didn't like it this week. I thought it was written in a condesceding nature towards the non-saber baseball fan. It doesn't take much analisys to realize that Jr has been struggling in the 3 hole and should be moved (althought he is heating up now). Jr in the 3 hole has been a problem over the past few seasons. I don't need Erardi saying we (sabermetric fans) realize this that other non saber fans can.

SteelSD
06-01-2008, 05:20 PM
Where did he err?

He chose to apply one definition of "under .500" to a scenario where a team has no mathematical possibility of reaching .500. This created a contextual issue when he then presented another definition of "under .500" for a season still in flux. But that error is really a molehill rather than a mountain as the projected W/L records are accurate, which is what really matters.

And keep in mind that the "10-Run" rule Eradi is using isn't actual Pythag. He's likely using it in order to dumb down Pythag for reader mass consumption.

RedsManRick
06-01-2008, 05:37 PM
Isn't it kind of pointless to look back during the season. IMO it is kind of pointless to look at a 26-29 record and say well "if we won this game we would be here".

As for Erardi's column I didn't like it this week. I thought it was written in a condesceding nature towards the non-saber baseball fan. It doesn't take much analisys to realize that Jr has been struggling in the 3 hole and should be moved (althought he is heating up now). Jr in the 3 hole has been a problem over the past few seasons. I don't need Erardi saying we (sabermetric fans) realize this that other non saber fans can.

For purposes of evaluation. If you want to say "the Reds are a .500 team" as a description of their ability (and thus likely record looking forward), then you only need to look back and translate 1 or 2 losses in to wins.

Obviously in order to actually have a .500 record, they would need to win 3 more games than they lose, but that would not be the point of the original statement.

And to emphasize Steel's point about Eradi's use of 10 run per win, consider that 10 runs are more valuable in a lower run scoring environment.

That is to say, given a team who scores 600 runs and allows 600 (3.7 R/G per team), a 10 run swing represents more wins/losses than for a team who scores 800 runs and allows 800. 10 runs/win makes sense in an environment in which the combined run total of the average game is ~10 runs -- or 810 runs scored and allowed -- a pretty close estimate for the Reds of recent vintage.

Sea Ray
06-01-2008, 08:04 PM
He chose to apply one definition of "under .500" to a scenario where a team has no mathematical possibility of reaching .500. This created a contextual issue when he then presented another definition of "under .500" for a season still in flux. But that error is really a molehill rather than a mountain as the projected W/L records are accurate, which is what really matters.

And keep in mind that the "10-Run" rule Eradi is using isn't actual Pythag. He's likely using it in order to dumb down Pythag for reader mass consumption.

He didn't project any W/L records, so there's nothing to be accurate or inaccurate. He merely stated that the Reds current record is consistent with their RS/RA.

SteelSD
06-01-2008, 10:06 PM
He didn't project any W/L records, so there's nothing to be accurate or inaccurate. He merely stated that the Reds current record is consistent with their RS/RA.

Ray, when someone states that a team's record should be something based on Run Diff, they're projecting something.

Eradi's only real error in the article was not realizing that how one determines games over/under .500 can change when a team is mathematically incapable of reaching a .500 winning percentage. But that was a minor faux pas, and it has nothing to do with sabermetrics or Pythag, especially since Eradi wasn't actually using Pythag.

Sea Ray
06-02-2008, 11:43 AM
Ray, when someone states that a team's record should be something based on Run Diff, they're projecting something.

Eradi's only real error in the article was not realizing that how one determines games over/under .500 can change when a team is mathematically incapable of reaching a .500 winning percentage. But that was a minor faux pas, and it has nothing to do with sabermetrics or Pythag, especially since Eradi wasn't actually using Pythag.

No need to get defensive about sabermetrics or pythag. All I ever wanted to know by starting this tread was which method was the proper way to figure games over/under .500 when using sabermetrics calculations. Instead of answering that you seem intent on defining sabermetrics and its virtues. The very direct and succinct answer from Puffy was much appreciated. Steel, take note. ;)

SteelSD
06-02-2008, 12:01 PM
No need to get defensive about sabermetrics or pythag. All I ever wanted to know by starting this tread was which method was the proper way to figure games over/under .500 when using sabermetrics calculations. Instead of answering that you seem intent on defining sabermetrics and its virtues. The very direct and succinct answer from Puffy was much appreciated. Steel, take note. ;)

Again, the method of determining games over/under .500 has absolutely nothing to do with sabermetrics. The fact that you didn't know this prior to complaining about sabermetrics in the very first sentence of a thread you titled "Pythagorean Question" speaks volumes.

And for the record, I can't possibly "defend" a single sabermetric principle in this thread because what you're complaining and/or asking about has nothing to do with sabermetrics.

Sea Ray
06-02-2008, 12:59 PM
Again, the method of determining games over/under .500 has absolutely nothing to do with sabermetrics.

OK Steel then your complaint lies with Erardi not me. Afterall he begins his article refuting exactly what you just stated:


One of the essential principles of sabermetrics - and this is our guiding light right here, the Holy Grail - is that in a 162-game season, for every 10 runs your team scores over the number of runs it allows, you get a victory. If you score the same number of runs as you allow, you are going to finish within one or two games of .500, 81-81.

He says games over/under .500, based upon a "method" of 10 runs, are one of the essential principles of sabermetrics yet you claim it has nothing to do with sabermetrics. He not only says it is sabermetrics he says it's the Holy Grail. If you're right then his article is worthless. You can't take away someone's Holy Grail and have much left over.

IslandRed
06-02-2008, 01:51 PM
What folks are saying is, whether 74-88 is 7 games under .500 or 14 games under .500 is not a point of sabermetric contention.

The "10-run method," which is a method of sabermetric contention, is based on the math that says 74-88 is 7 games under .500.

If Erardi used it against the other method of calculating .500, that was a mistake.

For what it's worth, the 10-run method of Pythagorean shorthand applies to a full season's worth of games. Erardi is probably making a mistake by using the same principle applied to a third of the season's worth of games, as the expected deviation increases with a smaller sample size.

bucksfan2
06-02-2008, 02:46 PM
What folks are saying is, whether 74-88 is 7 games under .500 or 14 games under .500 is not a point of sabermetric contention.

The "10-run method," which is a method of sabermetric contention, is based on the math that says 74-88 is 7 games under .500.

If Erardi used it against the other method of calculating .500, that was a mistake.

For what it's worth, the 10-run method of Pythagorean shorthand applies to a full season's worth of games. Erardi is probably making a mistake by using the same principle applied to a third of the season's worth of games, as the expected deviation increases with a smaller sample size.

I don't think Erardi made a mistake. I think he took a manipulation of the 10 run/.500 winning percentage in order to prove his point. IMO Erardi thought that the majority of his readers wouldn't realize the two different understandings of .500 that he used.

jojo
06-02-2008, 03:13 PM
One of the problems I have with saber metrics is how subtly the numbers change. There is an example of this in John Erardi's column today.

While I can see where someone might be confused, really the inconsistency is Erardi's construct and not one inherent in the sabermetric principle in question. I think that's basically Steel's point.

You aren't having a problem with sabermetrics in this instance but rather, based upon your post, you're having a problem with the way a sabermetric principle is being explained. In other words, this really shouldn't be a reason you discount sabermetrics.

It's not easy to do what Erardi is trying to do recently (bring sabermetrics to an audience that is probably largely uninitiated) especially when Erardi himself really is more a translator (relying heavily upon some local saber-leaning bloggers). I give him a lot of credit and a lot of charity (i.e. benefit of the doubt), when reading his column.

Sea Ray
06-02-2008, 03:26 PM
What folks are saying is, whether 74-88 is 7 games under .500 or 14 games under .500 is not a point of sabermetric contention.

The "10-run method," which is a method of sabermetric contention, is based on the math that says 74-88 is 7 games under .500.



You didn't get Steel's memo. The method has absolutely nothing to do with sabermetrics so by definition you can't have a sabermetric method.



For what it's worth, the 10-run method of Pythagorean shorthand applies to a full season's worth of games. Erardi is probably making a mistake by using the same principle applied to a third of the season's worth of games, as the expected deviation increases with a smaller sample size.

I think you make an excellent point (above). I think that's the crux of it. Sabermetrics falls apart if you apply the wrong method or formula if you prefer.

I am not discounting sabermetrics. This Pythag thing is interesting to look at and it has its flaws. Blowouts and other things can throw it off. It was way off last year for Colorado and Arizona, but there's nothing wrong with using it to set goals for your team with the idea of reaching a certain amount of wins for a year.

jojo
06-02-2008, 03:37 PM
I think that's the crux of it. Sabermetrics falls apart if you apply the wrong method or formula if you prefer.

To be fair, that's not really on "sabermetrics". It's on the person who is misapplying a sabermetric principle when making their reasoned argument.....

You're right though to point out that it's important to understand the limitations of a formula because they will inform the strength of the argument that relies upon the formula.

Sea Ray
06-02-2008, 04:08 PM
To be fair, that's not really on "sabermetrics". It's on the person who is misapplying a sabermetric principle when making their reasoned argument.....

You're right though to point out that it's important to understand the limitations of a formula because they will inform the strength of the argument that relies upon the formula.

Sabermetrics are only as good as the person putting together the sabermetric argument. Sure, that goes w/o saying. Sabermetrics are the study of baseball through statistics so it's only as good as the one applying it.

RedsManRick
06-02-2008, 05:01 PM
Sabermetrics falls apart if you apply the wrong method or formula if you prefer.

So does physics. Does that doesn't make it wrong as an approach to explaining the natural world, It just means you have to take care to use it properly.



I am not discounting sabermetrics. This Pythag thing is interesting to look at and it has its flaws. Blowouts and other things can throw it off. It was way off last year for Colorado and Arizona, but there's nothing wrong with using it to set goals for your team with the idea of reaching a certain amount of wins for a year.

I think this really gets at the heart of the complaint. Sabermetrically minded people read Erardi's article and understood he was using short-hand. A lay person may think he's being fully descriptive and come to the same sort of perspective you've outlined here. Pythag isn't fate and it was never meant to be. If that's how the article has come across to you, then perhaps Eradi could have been more clear.

It's a general principal saying that while wins and losses are a result of team performance, wins and losses are not the best measure of team quality. If you want to know how a team is most likely to perform looking forward, their RS and RA are more predictive than their current W/L record. W/L can be deceptive too, and looking at runs scored and allowed gets you closer to a measurement of actual team quality.

This is one of the common problems with teaching statistical methods to non-stat people. It doesn't deal with certainties and many people aren't comfortable in a world of mere likelihoods. Erardi's failing in the article was trying to bridge that gap by changing a principal which clearly recognizes the variable nature of the game and treating it like an exact formula. It's a very tricky tightrope to walk and perhaps Erardi slipped a bit.

Sea Ray
06-02-2008, 06:03 PM
Rick, my complaint was always with the inconsistency of Erardi's article. Once again you don't have to defend sabermetrics as a whole. Statistical analysis falls apart on anything if you're not consistent with your numbers.

RedsManRick
06-02-2008, 06:30 PM
Rick, my complaint was always with the inconsistency of Erardi's article. Once again you don't have to defend sabermetrics as a whole. Statistical analysis falls apart on anything if you're not consistent with your numbers.

I guess we're all in agreement then. Erardi screwed up. The end. I take it you never heard back from him?

jojo
06-02-2008, 06:42 PM
Sabermetrics are only as good as the person putting together the sabermetric argument. Sure, that goes w/o saying. Sabermetrics are the study of baseball through statistics so it's only as good as the one applying it.

Once again, that's an indictment of the one trying to formulate a sabermetric argument rather than an indication of the utility or power of sabermetrics.

A logical extension of your approach on this issue would be to suggest photoshop isn't a very powerful program because a particular user doesn't understand the concept of layers. The onus really should be on the user.


Statistical analysis falls apart on anything if you're not consistent with your numbers.

There really wasn't anything wrong with Erardi's analysis. The Reds run differential predicts a certain w-l record (i.e. 74-88). The point he was making is unaffected by whether that record is described as being 14 games under .500 or 7 games under .500....

SteelSD
06-02-2008, 07:18 PM
While I can see where someone might be confused, really the inconsistency is Erardi's construct and not one inherent in the sabermetric principle in question. I think that's basically Steel's point.

Bingo.


There really wasn't anything wrong with Erardi's analysis. The Reds run differential predicts a certain w-l record (i.e. 74-88). The point he was making is unaffected by whether that record is described as being 14 games under .500 or 7 games under .500....

And...you're 2 for 2!

Raisor
06-02-2008, 08:04 PM
Bingo.



And...you're 2 for 2!

Still not as good as Bruce.

Sea Ray
06-02-2008, 08:38 PM
I guess we're all in agreement then. Erardi screwed up. The end. I take it you never heard back from him?

No, I haven't. I sent it to jerardi@enquirer.com. Anyone know of a better link?

Sea Ray
06-02-2008, 08:52 PM
Once again, that's an indictment of the one trying to formulate a sabermetric argument rather than an indication of the utility or power of sabermetrics.

A logical extension of your approach on this issue would be to suggest photoshop isn't a very powerful program because a particular user doesn't understand the concept of layers. The onus really should be on the user.



There really wasn't anything wrong with Erardi's analysis. The Reds run differential predicts a certain w-l record (i.e. 74-88). The point he was making is unaffected by whether that record is described as being 14 games under .500 or 7 games under .500....

The defensiveness of sabermetrics around here is astounding. I think talking about religion and race would be easier with some of you folks. I nor anyone else in this thread am saying that sabermetrics isn't a "powerful program" to use your analogy to photoshop.

Getting back to Erardi's analysis, which is what this thread is all about, we've established that his point is accurate to use 10-run differential for every game under the .500 mark when using the standard of 74-88 is 7 games under. Where he erred was when he continued onto this point:


That's right: You've scored 23 fewer runs than you've allowed, which by run differential should place your team within a game or two of being 2-3 games under .500.

Well, guess what? Through Friday's game, your team is dead on it, 3 games under, 26-29.


So I would argue there is something wrong with his analysis, however I agree with Steel that it's not a big deal. It is more a molehill than a mountain. I merely wanted the inconsistency cleared up.

The debate of the importance of sabermetrics as a whole will have to be left to those with a lot more free time than I have...

jojo
06-02-2008, 11:49 PM
The defensiveness of sabermetrics around here is astounding.

Defensiveness? I thought it was a discussion.


I think talking about religion and race would be easier with some of you folks.

Several people have engaged the topic.


I nor anyone else in this thread am saying that sabermetrics isn't a "powerful program" to use your analogy to photoshop.

The point wasn't about the power of sabermetrics but rather that the issue you were having actually had nothing to do with the "sabermetric" part of Erardi's article. I think it was an important point to make given your thesis:


One of the problems I have with saber metrics is how subtly the numbers change. There is an example of this in John Erardi's column today.

In fact, citing Erardi's column does not support that the conclusion.


Where he erred was when he continued onto this point:

Where Erardi apparently erred was assuming readers understood there are typically 162 games in a completed regular season while there will be a lesser number of games played depending upon the point of the season-in-progress that is being referenced.


So I would argue there is something wrong with his analysis,

There actually isn't though. He just didn't word a sentence carefully enough apparently.

Sea Ray
06-03-2008, 09:47 AM
Jojo, I'm trying to figure out if we're not communicating or if you just have a thing for getting in the last word. I'll have my answer soon enough.

M2
06-03-2008, 03:28 PM
On the whole below .500 thing, if you're 50-62, then you're 12 games below .500. Literally it would take 12 more wins than losses to get you back to .500.

Now if you could go back in time and change six losses to victories you'd be at .500, but until someone invents a way to travel through time and alter the results of past games, then there really isn't any practical way to get back to .500 in that example in just six games. Plus, you'd also have to be able to change your results faster than other teams could steal victories from you. It would be a general mess.

So that's why you don't divide by two when counting games below/above .500. In actual linear timescape that is the actual W-L differential of your club.

RedsManRick
06-03-2008, 04:49 PM
Where Erardi apparently erred was assuming readers understood there are typically 162 games in a completed regular season while there will be a lesser number of games played depending upon the point of the season-in-progress that is being referenced.

To be fair Jojo, that was not Erardi' error. He did in fact use two different methods for calculating "games from .500".

He first refers to a 74-88 record as 7 games under, the divide by two method.

...your team finished within one or two games of seven games under .500, 74-88...

He later refers to a a 26-29 record as 3 games under, the simple subtraction method.

You've scored 23 fewer runs than you've allowed, which by run differential should place your team within a game or two of being 2-3 games under .500.

Unfortunately, by the same math, the Reds should be looked at as 1.5 games under .500, not 3. That was his error.

Obviously it doesn't change his point at all. The only real difference is that the Reds were .8 ahead of their pythagorean record, not .7 behind it.

The raw subtraction method while perhaps useful in the context M2 described (the Reds need to win 3 games to get to .500) is not appropriate in the pythagorean record context.

SeaRay's point was that sabermetrics can be confusing due to the subtle nature of interpretation. Both of Erardi's statements seem logical on the surface, but the inconsistency cited above provides ammunition to those who would question it's efficacy. Some people see a presumably honest mistake such as Erardi's and read it as an attempt to bend the numbers to make them say whatever you want to them to. It's the nature of all of analysis, and a constant challenge for the analyst.

M2, take an team with 800 runs score and 800 runs allowed. Given them 810 runs scored and the new pythag is 82-80.

1/(1+(800/800)^2 = .500 = 81.0 wins = 81-81

1/(1+(810/800)^2)) = .506 = 81.972 wins = 82-80

Adding 10 runs of differential gives you one more win AND one fewer loss by pythagorean method. So saying that 10 runs equals 1 win is the same as saying increasing your run differential by 10 runs takes one of your existing losses and turns it in to a win.

The beauty of the pythagorean method is that it automatically scales with the number of games played. A 10 run differential in 60 games, say 290-280, equates to a .518 W% (31-29) whereas a 10 run differential in 162 games, say 810-800, equates to just a .506 W% (82-80). The won differential is the same as the run differential. So whether you've played 60 games or 162, that 10 run differential equates to the same effect on your record.

bucksfan2
06-04-2008, 09:12 AM
SeaRay have you heard back from Erardi yet?


SeaRay's point was that sabermetrics can be confusing due to the subtle nature of interpretation. Both of Erardi's statements seem logical on the surface, but the inconsistency cited above provides ammunition to those who would question it's efficacy. Some people see a presumably honest mistake such as Erardi's and read it as an attempt to bend the numbers to make them say whatever you want to them to. It's the nature of all of analysis, and a constant challenge for the analyst

One thing that I have noticed about the sabermetric community is that they are very detail oriented. Another that I have noticed with statiticians is that there is a tendency to manipulate the numbers in order for a desired result. I think Erardi did just that. There are two different intrepretations of being .500. I would find it hard to believe that a guy like Erardi didn't know the suttle differences between both and didn't used both in the same comparison in order prove his point.

Sea Ray
06-04-2008, 09:17 AM
SeaRay have you heard back from Erardi yet?



Nope. I doubt I'll hear anything. This is the letter I wrote him on Sunday. I tried to keep it brief and to the point.


John,

One of the problems I have with sabermetrics is how subtly the numbers change. There is an example of this in your column today. You mentioned that at 74-88 the Reds were considered 7 games under .500 yet later on you said that at 26-29 they are 3 games under. It can't be both. Either 74-88 is 14 games under or 26-29 is one game under. You can't change your method of determining over/under .500 in the midst of your argument. Please explain.

Thanks,

MWM
06-04-2008, 09:17 AM
Another that I have noticed with statiticians is that there is a tendency to manipulate the numbers in order for a desired result.

That's just not true at all. No true statistician would ever do this. And I don't think you've "noticed" this. If you have, then you should be able to cite several examples. Honestly, bucksfan, have you ever made an honest study of the statistical side of baseball? I don't think you have, which is why I don't think all the comments you make opposing statistical evaluation are all that credible. Until someone has really taken the time and effort to understand it, I don't think they can really comment as to their efficacy.

M2
06-04-2008, 09:19 AM
The raw subtraction method while perhaps useful in the context M2 described (the Reds need to win 3 games to get to .500) is not appropriate in the pythagorean record context.

I understand how pythag works without the tutorial (WOY and I even published an article using it once upon a time).

So I fully get how an ex post facto win swing would operate. However, since neither you nor anybody else on the planet has the power to go back and change results a one-win swing in either direction would net you two games above or below .500, which is a pure counting exercise (always has been, always will be). If you say you're six games below .500, then you'd better have six more losses than wins.

bucksfan2
06-04-2008, 09:47 AM
That's just not true at all. No true statistician would ever do this. And I don't think you've "noticed" this. If you have, then you should be able to cite several examples. Honestly, bucksfan, have you ever made an honest study of the statistical side of baseball? I don't think you have, which is why I don't think all the comments you make opposing statistical evaluation are all that credible. Until someone has really taken the time and effort to understand it, I don't think they can really comment as to their efficacy.

Erardi just did this. As for manipulation of numbers I don't mean changing the actual numbers involved. I mean taking different variations of the same ideas. Erardi took two different intrepretations of the same idea. Look at pollsters. They will ask the same question a several different ways in order to produce the statistic they want.

Have I ever done an honest study of the statistical side of baseball? NO. I have briefly looked into some statistics but I am very skeptical whenver an outside influence is introduced into a stat.

jojo
06-04-2008, 09:54 AM
Erardi just did this. As for manipulation of numbers I don't mean changing the actual numbers involved. I mean taking different variations of the same ideas. Erardi took two different intrepretations of the same idea. Look at pollsters. They will ask the same question a several different ways in order to produce the statistic they want.

Have I ever done an honest study of the statistical side of baseball? NO. I have briefly looked into some statistics but I am very skeptical whenver an outside influence is introduced into a stat.

Erardi's point was that run differential predicts record. The Reds run differential did just that despite whether he describes the Reds record as 7 under or 14 under .500 (BTW, he also included the records for everyone to see).

He didn't manipulate any stats to prove his point. He just got in his own way when explaining the principle.

M2
06-04-2008, 09:54 AM
Erardi just did this.

He ineloquently explained a particular concept (and not a central one to his larger point). If you look at his actual math (rather than his verbal explanation of the math), it's solid, but I'm guessing you haven't done that.

pahster
06-04-2008, 09:59 AM
Look at pollsters. They will ask the same question a several different ways in order to produce the statistic they want.


Incorrect. In survey research, questions are repeated with slightly different wordings in order to ensure that respondents give valid (i.e. consistent) responses.

RedsManRick
06-04-2008, 10:48 AM
I understand how pythag works without the tutorial (WOY and I even published an article using it once upon a time).

So I fully get how an ex post facto win swing would operate. However, since neither you nor anybody else on the planet has the power to go back and change results a one-win swing in either direction would net you two games above or below .500, which is a pure counting exercise (always has been, always will be). If you say you're six games below .500, then you'd better have six more losses than wins.

I wasn't trying to imply you didn't get -- writing for the broader audience. It sounds like your complaint is one of semantics.

It's interesting to me though because the "Games Behind" lingo that's used so much counts every win or loss by a given team as a 1/2 game, an appreciation that there is both a winner and loser to every game.

I'm not sure it's so easily resolved as you suggest. If the Reds are 27-30 and the Cubs 30-27, and the Reds win 1 game, they pick up just .5 games in the standings. If they beat the Cubs or the Cubs lose, the Cubs lose .5 games in the standings.

I'm not thinking too clearly this morning, but I don't think you can escape the need to use both methods of calculation. I just wish I had a clearer understanding of when you use each one.

Sea Ray
06-04-2008, 11:00 AM
I agree with M2. To me the Reds are 3 games under .500 right now

If it were me I'd standardize this formula and instead of using 10 run differential for games under .500/2 I'd use 5 stick with the literal under .500. For example if the Reds finish 74-88 they're 14 games under. A pythag of 14 X 5 runs = 70 differential between RS and RC. Then we're all on the same page.

Sea Ray
06-04-2008, 11:02 AM
II'm not sure it's so easily resolved as you suggest. If the Reds are 27-30 and the Cubs 30-27, and the Reds win 1 game, they pick up just .5 games in the standings. If they beat the Cubs or the Cubs lose, the Cubs lose .5 games in the standings.




Why complicate the matter by bringing the Cubs into it? This issue is getting back to .500 and that has nothing to do with another club.

M2
06-04-2008, 11:02 AM
I wasn't trying to imply you didn't get -- writing for the broader audience. It sounds like your complaint is one of semantics.

It's interesting to me though because the "Games Behind" lingo that's used so much counts every win or loss by a given team as a 1/2 game, an appreciation that there is both a winner and loser to every game.

I'm not sure it's so easily resolved as you suggest. If the Reds are 27-30 and the Cubs 30-27, and the Reds win 1 game, they pick up just .5 games in the standings. If they beat the Cubs or the Cubs lose, the Cubs lose .5 games in the standings.

I'm not thinking too clearly this morning, but I don't think you can escape the need to use both methods of calculation. I just wish I had a clearer understanding of when you use each one.

Games behind and games below/above .500 are two different things. For instance, in the example you provided above, the Reds are 3 games below .500 and the Cubs are 3 games above .500. Now the Reds could sweep a 3-game set from the Cubs and the teams would be even at .500, but that only reinforces their relation to .500. No one got .5 wins along the way. The Reds had to win three whole games to reach .500 and the Cubs had to lose three whole games to drop to .500.

So it's important to understand that games above/below .500 describes a team's relation to .500, not a team's relation to another team or to where it theoretically might be under different circumstances. And yes that is semantics, but semantics count when you're trying to impart information. In this case Erardi's semantic misstep on games below .500 has given bucksfan2 license to ignore the larger point Erardi was making.

M2
06-04-2008, 11:28 AM
I agree with M2. To me the Reds are 3 games under .500 right now

If it were me I'd standardize this formula and instead of using 10 run differential for games under .500/2 I'd use 5 stick with the literal under .500. For example if the Reds finish 74-88 they're 14 games under. A pythag of 14 X 5 runs = 70 differential between RS and RC. Then we're all on the same page.

Pythag doesn't work that way. Erardi's mistake was using the wrong terminology to describe the larger concept. Yes, 74-88 is 14 games below .500. No disputing that.

Erardi however is talking about what a team theoretically needs to do to play winning baseball (e.g. outscore its opponents on a consistent basis). So better performance by the team could net you a 7-game swing which could get you to .500 (theoretically).

So let's apply that to the current situation. The Reds are 28-31, 3 games below .500. If they continue playing baseball like they have been (-20 run differential) they can expect to finish roughly 77-85. Let's say they need to win 90 games to make the playoffs. That would necessitate a 62-41 record from here on out. Now compare that to the 49-54 record you'd expect if they maintained their current level of play. That means the team would need a 13-game swing from its current play to where it needs to be.

So what kind of baseball will the Reds need to play to achieve that 13-game swing? They will roughly need to outscore their opponents by 120 runs rather than be outscored by 30 runs. That means they need their actual run differential to improve 150 runs over its current path (not easy, but it's not uncommon). Mind you, these are thumbnail numbers, just like Erardi was using +/- 10 as a thumbnail. If you play a low scoring brand of baseball, then you need a smaller run gap to achieve a winning record. If you play a high scoring brand of baseball, then you need a larger gap. For instance, here's the pythag formula used by Baseball-Reference.com:

Pythagorean winning percentage is an estimate of a team's winning percentage given their runs scored and runs allowed. Developed by Bill James, it can tell you when teams were a bit lucky or unlucky. It is calculated by

(Runs Scored)^1.83
---------------------------------------------------------
(Runs Scored)^1.83 + (Runs Allowed)^1.83

The traditional formula uses an exponent of two, but this has proven to be a little more accurate.

As you can, the relationship of runs scored to runs allowed is a little more complex than simple addition or subtraction.

SteelSD
06-04-2008, 11:37 AM
I agree with M2. To me the Reds are 3 games under .500 right now

If it were me I'd standardize this formula and instead of using 10 run differential for games under .500/2 I'd use 5 stick with the literal under .500. For example if the Reds finish 74-88 they're 14 games under. A pythag of 14 X 5 runs = 70 differential between RS and RC. Then we're all on the same page.

Or you could just use the actual Pythag formula to figure out a team's projected record. How you explain games over/under .500 from there is up to you.

Sea Ray
06-04-2008, 11:42 AM
So let's apply that to the current situation. The Reds are 28-31, 3 games below .500. If they continue playing baseball like they have been (-20 run differential) they can expect to finish roughly 77-85. Let's say they need to win 90 games to make the playoffs. That would necessitate a 62-41 record from here on out. Now compare that to the 49-54 record you'd expect if they maintained their current level of play. That means the team would need a 13-game swing from its current play to where it needs to be.



Educate me 'cause I truly don't get this. I'm all ears and I'm clearly missing something. Why would a -20 run differential equate to a 77-85? Erardi said a -70 run differential only resulted in a three game difference, 74-88. I understand he's using a 10 run shorthand for his article but why wouldn't a -20 result in a record of 79-83? Show me the formula that shows -70 run projects to 74-88 and -20 equates to 77-85.

RedsManRick
06-04-2008, 11:59 AM
Educate me 'cause I truly don't get this. I'm all ears and I'm clearly missing something. Why would a -20 run differential equate to a 77-85? Erardi said a -70 run differential only resulted in a three game difference, 74-88. I understand he's using a 10 run shorthand for his article but why wouldn't a -20 result in a record of 79-83? Show me the formula that shows -70 run projects to 74-88 and -20 equates to 77-85.

I think it's because a 162 game projection requires projection the current run differential over the course of the season. It's -20 today, but we're only 40% of the way through the season. The Reds are on pace to be at -55 at the end of the season.

The easy way to avoid this confusion is to simply look at the pythagorean winning percentage and calculate that against the number of games you're interested in. The formula (The 1.83 exponent version) says the Reds are playing .467 baseball.



268^1.83 27764.30 27764.30
________________________ = _____________________ = _______________ = .467
(268^1.83) + (288^1.83) 27764.30 + 31672.93 59437.23

Over 59 games, that's a 28-31 record.
Over 162 games, that's a 76-86 record.

SteelSD
06-04-2008, 12:02 PM
Educate me 'cause I truly don't get this. I'm all ears and I'm clearly missing something. Why would a -20 run differential equate to a 77-85? Erardi said a -70 run differential only resulted in a three game difference, 74-88. I understand he's using a 10 run shorthand for his article but why wouldn't a -20 result in a record of 79-83? Show me the formula that shows -70 run projects to 74-88 and -20 equates to 77-85.

What you haven't considered is "If they continue playing baseball like they have been...", that means the Run Differential gap projects to widen rather than staying at a static -20. At current pace, the Reds would project to score 736 Runs and allow 791; producing a Pythag of 76-86 (or 75-87 if one is rounding down rather than up).

M2
06-04-2008, 12:12 PM
Educate me 'cause I truly don't get this. I'm all ears and I'm clearly missing something. Why would a -20 run differential equate to a 77-85? Erardi said a -70 run differential only resulted in a three game difference, 74-88. I understand he's using a 10 run shorthand for his article but why wouldn't a -20 result in a record of 79-83? Show me the formula that shows -70 run projects to 74-88 and -20 equates to 77-85.

Well, you've got the formula above if you want to play with it. Like Erardi, I thumbnailed the eventual record on the current runs pace. And let me underline this, I was talking about the runs pace. -20 right now continued over the course of a full season gets you to -55. For the record I was guesstimating (before I just picked up my calculator) at -50, which got me to 77-85 via shorthand.

You can plug the current runs scored pace (736 for a 162-game season) and runs allowed pace (791) into it. When I do that (use the exact formula rather than a thumbnail), I get a 76-86 record (actually I get 75.7-86.3). The Reds will need to be more like 811 runs scored and 715 runs allowed to be in the dogfight for the playoffs. That would get you a 90-72 pythag with the 150-run swing (in differential) I was referring to above.

Spring~Fields
06-04-2008, 12:47 PM
RedsManRick

Alright as a common fan I jump on this thread and read it entirely with interest, then I quickly click over to ESPN to look at the standings and see that the Reds 28-31 RS 268 RA 288, proceed to do the minor math thinking I will come up with -20 runs/2 = 2 games under, instead I find the actual 3 games under .500, what has common fan missed in this case?



RS 268 RA 288 -20, -20/162 = - 0.1234 * 162 = -20 runs, 10=1 or 2 games under .500
Yet at the same time 475 win pct x 162 = 76.95 or 77 wins

85 losses and 8 games under .500 in contrast to -20 runs and 2 games under .500,
8 games under x 10 would project to a -80 run in RS, though the projection using the current RS differentiation of -20 if constant projects to just 2 games under .500 for the season.

What am I missing ? If I am trying to discover where the Reds will be according to RS differentiation? and how many games under .500 they will finish by runs scored ?

RedsManRick
06-04-2008, 01:00 PM
SF, when making a pythagorean comparison you cannot add games. Looking at the Reds as 3 games under .500 implies a forward looking position in which 3 more games are played.

What you really want to compare 28-31 against is not 31-31, but the .500 winning percentage of 59 games played, 29.5-29.5. By this, you can see that the Reds are 1.5 wins below .500, which would equate to a -15 run differential, close to what we'd expect given their actual -20.

I think the least confusing way is to simply use the formula to calculate winning percentage, and then multiply that against the number of games.

Another way to think about it is that because pythagorean comparison works in a fixed number of games, a 10 run increase in run differential correlates to 1 additional win AND 1 fewer loss (by definition, since you can't add new games). So a 30-30 expectation becomes a 31-29 expectation if you up the run differential by 10, just like an 81-81 expectation becomes 82-80 by upping the differential by 10 runs. So if you know their run differential is going to be 10X (X being positive or negative), take a .500 record for the given number of games, add X wins and subtract X losses.

So what should we expect if the Reds have a +50 run differential over the next 50 games, say 250-200?

Current Record: 28-31.
Next 50 Games: Assumed 25-25 -> 30-20
New Record: 58-51

FWIW, the new Pythagorean record would be: 57-52 (actually 57.5-51.5)

The 'relative to .500' language acts is a forward looking environment in which more games can (and will) be played. But in doing so, it only counts the 1/2 of the game outcome which is relevant to the team you're looking at (If your team wins, the other team loses -- but you only care about counting the win). Looking at games already played, which is what we do in the pythagorean comparison, a change in outcome not only gives your team 1 extra win, but it also gives your team 1 fewer loss. Thus, in the 'relative to .500' context, it takes twice as many games as in the 'pythagorean' context.

One thing that might help, language wise, is to say we're X wins below/above .500 instead of games, and leave games for when we're really considering the entirety of the game outcome. In the 'games behind' context, games implies both wins and losses. In the .500 context, it doesn't.

Sea Ray
06-04-2008, 02:52 PM
Well, you've got the formula above if you want to play with it. Like Erardi, I thumbnailed the eventual record on the current runs pace. And let me underline this, I was talking about the runs pace. -20 right now continued over the course of a full season gets you to -55. For the record I was guesstimating (before I just picked up my calculator) at -50, which got me to 77-85 via shorthand.

You can plug the current runs scored pace (736 for a 162-game season) and runs allowed pace (791) into it. When I do that (use the exact formula rather than a thumbnail), I get a 76-86 record (actually I get 75.7-86.3). The Reds will need to be more like 811 runs scored and 715 runs allowed to be in the dogfight for the playoffs. That would get you a 90-72 pythag with the 150-run swing (in differential) I was referring to above.

OK. that explains it. I took your post (as follows) to mean -20 run differential at year's end


If they continue playing baseball like they have been (-20 run differential) they can expect to finish roughly 77-85.

fearofpopvol1
07-07-2009, 10:59 AM
For those who are major believers in this idea...the Reds' pythag is shot after the last 2 days. That's a massive deficit to make up.

Falls City Beer
07-07-2009, 11:04 AM
For those who are major believers in this idea...the Reds' pythag is shot after the last 2 days. That's a massive deficit to make up.

Yeah, they're beyond dead. They've moved from the middle of the pack in the NL to near the bottom. They're not the Nats, but when you're - 50, you're headed nowhere, especially when you consider they aren't anticipating a ton of reinforcements.

BCubb2003
07-07-2009, 11:06 AM
For those who are major believers in this idea...the Reds' pythag is shot after the last 2 days. That's a massive deficit to make up.

That's what's difficult for many people to grasp about the pythag. Two games ago, it showed the Reds to be a certain kind of team. Two games later, the Reds are much different, off the charts. Yet there's nothing in the numbers to say that in today's game, the Reds are a 22-1 losing team. I personally think there's a trend happening here, where the starting pitching is faltering throughout the rotation, and we're in for a rough spell. But on a day-to-day basis, the only thing we really know about the pythag is that it'll change by night's end.

nate
07-07-2009, 11:07 AM
For those who are major believers in this idea...the Reds' pythag is shot after the last 2 days. That's a massive deficit to make up.


And yet by the end of the season, run differential and winning percentage will correlate at a very high percentage for all teams in the history of baseball.

In other words, there aren't "major believers" that follow it like a religion. Rather, some posit that run differential is closely tied to record.

Falls City Beer
07-07-2009, 11:10 AM
Basically, what happened was luck (defense?) forestalled a drubbing like this for some weeks, but last night, against a competent offense, the dam burst.

jojo
07-07-2009, 11:30 AM
In the last 31 games the Reds are 13-18 with 107 RS and 158 RA.

PythagenPat would predict a record of 11-20 given the run differential.

After Saturday, the Reds were 13-16 and PythagenPat predicted the Reds should be 13-16 based upon their RS/RA over that period.

Over a 30 game stretch, the blasting of the last two nights might seem like an untenable error that underestimates the Reds talent. But over the course of a season, it becomes more like a hill than a mountain peak in the RS/RA landscape unless of course, they continue to get blasted fairly regularly.

But really, at that point, wouldn't it be more a sign that the Reds might really be as bad as a skewed RA/RS would suggest rather than an indictment of Pythag? After all good teams blast bad ones. So a team that got blasted routinely, might not be as good as we'd like to think.

Sea Ray
07-07-2009, 11:36 AM
The Reds got blasted at a historical level last night. They've never lost by that many runs in their long history. That has to throw off the Pythag, but so what? The Pythag is what it is, another statistical tool and sometimes it's predicative and sometimes it isn't.

Falls City Beer
07-07-2009, 11:42 AM
The Reds got blasted at a historical level last night. They've never lost by that many runs in their long history. That has to throw off the Pythag, but so what? The Pythag is what it is, another statistical tool and sometimes it's predicative and sometimes it isn't.

I think Pythag is a story, not a prophecy.

Raisor
07-07-2009, 11:42 AM
Reds are currently +4 over the ExW-L.

nate
07-07-2009, 11:47 AM
The Reds got blasted at a historical level last night. They've never lost by that many runs in their long history. That has to throw off the Pythag, but so what? The Pythag is what it is, another statistical tool and sometimes it's predicative and sometimes it isn't.

It's not predictive.

fearofpopvol1
07-07-2009, 12:01 PM
And yet by the end of the season, run differential and winning percentage will correlate at a very high percentage for all teams in the history of baseball.

In other words, there aren't "major believers" that follow it like a religion. Rather, some posit that run differential is closely tied to record.

I wasn't really indicting the system. Just some put more emphasis on it than others.

I think overall it's a good measuring tool.

mth123
07-07-2009, 10:03 PM
Phillis 25 Reds 5. Each team has one win.

Ron Madden
07-08-2009, 04:23 AM
Phillis 25 Reds 5. Each team has one win.

We must wait until seasons end to truly appreciate how accurate Pythag really is. It doesn't work so well in a small sample like two, three or Five games. Like FCB said Pythag is a story not a prophecy.

:)

mth123
07-08-2009, 04:51 AM
We must wait until seasons end to truly appreciate how accurate Pythag really is. It doesn't work so well in a small sample like two, three or Five games. Like FCB said Pythag is a story not a prophecy.

:)

I agree. Just pointing out that a 30 run hit to the pythag after a couple blow-outs doesn't mean much. Over time, there will be a couple blow outs that the Reds win to even that out a little. Looking at it now and predicting a record based on it seems like a dubious exercise and it seems to be what many in this thread are doing.

bucksfan2
07-08-2009, 08:38 AM
I agree. Just pointing out that a 30 run hit to the pythag after a couple blow-outs doesn't mean much. Over time, there will be a couple blow outs that the Reds win to even that out a little. Looking at it now and predicting a record based on it seems like a dubious exercise and it seems to be what many in this thread are doing.

I have mentioned this before but shouldn't there be a cap put on margin of victory in a more advanced Pythag?

I understand that there will be a balancing out factor throughout the season. Every team has a couple of games in which they blow the team out of the water. But the 22-1 loss is such an outlier that it could skew the data to a certain degree. Even more so the Reds lost two back to back games 32-2 which equates for a 3 game swing over the course of a 162 games season not a 2 day span. Wouldn't it be more accurate if they limited the margin of victory to twice the average margin of victory? What if an advanced Pythag limited blow out wins to an 8 run difference? Wouldn't that make the end results more accurate?

BCubb2003
07-08-2009, 08:54 AM
I have mentioned this before but shouldn't there be a cap put on margin of victory in a more advanced Pythag?

I understand that there will be a balancing out factor throughout the season. Every team has a couple of games in which they blow the team out of the water. But the 22-1 loss is such an outlier that it could skew the data to a certain degree. Even more so the Reds lost two back to back games 32-2 which equates for a 3 game swing over the course of a 162 games season not a 2 day span. Wouldn't it be more accurate if they limited the margin of victory to twice the average margin of victory? What if an advanced Pythag limited blow out wins to an 8 run difference? Wouldn't that make the end results more accurate?

That's an interesting question. After a certain point, the dynamics of that game change to such a degree that they don't truly reflect the team. The shortstop coming in to pitch, for instance.

Highlifeman21
07-08-2009, 09:02 AM
That's an interesting question. After a certain point, the dynamics of that game change to such a degree that they don't truly reflect the team. The shortstop coming in to pitch, for instance.

Seems to be the only way Janish gets any playing time.

nate
07-08-2009, 09:19 AM
I have mentioned this before but shouldn't there be a cap put on margin of victory in a more advanced Pythag?

I understand that there will be a balancing out factor throughout the season. Every team has a couple of games in which they blow the team out of the water. But the 22-1 loss is such an outlier that it could skew the data to a certain degree. Even more so the Reds lost two back to back games 32-2 which equates for a 3 game swing over the course of a 162 games season not a 2 day span. Wouldn't it be more accurate if they limited the margin of victory to twice the average margin of victory? What if an advanced Pythag limited blow out wins to an 8 run difference? Wouldn't that make the end results more accurate?

Maybe so, but why? It correlates very highly already. It's not used as a predictive measure that needs to be more accurate; it is what it is. If you want higher accuracy, why limit it to 8 runs? Make it 1 and the advanced pythag will ALWAYS correlate 100% to win/loss.

BCubb2003
07-08-2009, 09:33 AM
I understand that it's not predictive, especially on a game-to-game basis, but if it's not predictive at all, what's the point? Doesn't everybody use it to get a better idea of what to expect from a team? That is, the team's playing over their heads and expect a fall, or better than they look and expect a rise?

bucksfan2
07-08-2009, 09:44 AM
Maybe so, but why? It correlates very highly already. It's not used as a predictive measure that needs to be more accurate; it is what it is.

Yes it is what it is. I don't want to take anything away from the original pythag. However, whenever you have a large data pool, roughly 162 different data points, you don't want outliers to skew the data. That is basic stats. The Reds had a two game stretch that may skew that data to an extent that it is not accurate. Hence the reason to put a cap on the run difference. Thats why I asked if there was an advanced Pythag.

For example, IIRC two years ago the DBacks defied the Pythag and made the playoffs with a negative run differential. If you had a way to put an advanced Pythag analysis on the data would that have shown anything different?


If you want higher accuracy, why limit it to 8 runs? Make it 1 and the advanced pythag will ALWAYS correlate 100% to win/loss.

They do have this. It is called the Win/Loss record. No need to change that.

westofyou
07-08-2009, 09:52 AM
Every year there is a blow out that brings this subject up, and every year somebody who eschews stats tries to say the phythag is a load.

I can set my sundial by it.. every year.

And at the end of the year every team comes pretty darn close to what the pythag said it should have done.

Instead of trying to prove the pythag incorrect while it's being compiled (and make no mistake it's the final result of the pythag that is the one that matters) the real test is finding all the outliers at the end of the season (or other seasons) and find out how wrong it is. Trying to address while it's in process is like picking up mercury.

nate
07-08-2009, 10:06 AM
I understand that it's not predictive, especially on a game-to-game basis, but if it's not predictive at all, what's the point? Doesn't everybody use it to get a better idea of what to expect from a team? That is, the team's playing over their heads and expect a fall, or better than they look and expect a rise?

To me, the biggest point to take away from pythag is that for a team to be successful, it has to score more runs than it allows. I know that seems a bit "duh," but that's always the thing I take away from it.

nate
07-08-2009, 10:10 AM
Yes it is what it is. I don't want to take anything away from the original pythag. However, whenever you have a large data pool, roughly 162 different data points, you don't want outliers to skew the data.

It's not 100% accurate and I don't expect it to be. I'm not sure why it's so important to you that it be more accurate.


That is basic stats. The Reds had a two game stretch that may skew that data to an extent that it is not accurate. Again, it's not 100% accurate. Nor is the season over yet.


Hence the reason to put a cap on the run difference. Thats why I asked if there was an advanced Pythag.There are variants one can find but they don't dismiss outliers.


For example, IIRC two years ago the DBacks defied the Pythag and made the playoffs with a negative run differential. If you had a way to put an advanced Pythag analysis on the data would that have shown anything different?Don't know and don't care. It's not predictive, it's not 100% accurate. It's an excellent indicator of whether a team winning via domination or smoke and mirrors.


They do have this. It is called the Win/Loss record. No need to change that.And by capping the margin of victory, you're trying to turn it into win/loss record.

I really don't understand the point you're trying to make.

bucksfan2
07-08-2009, 10:10 AM
Every year there is a blow out that brings this subject up, and every year somebody who eschews stats tries to say the phythag is a load.

I can set my sundial by it.. every year.

And at the end of the year every team comes pretty darn close to what the pythag said it should have done.

Instead of trying to prove the pythag incorrect while it's being compiled (and make no mistake it's the final result of the pythag that is the one that matters) the real test is finding all the outliers at the end of the season (or other seasons) and find out how wrong it is. Trying to address while it's in process is like picking up mercury.

Why stop at pretty darn close if you can find a way to be more accurate than pretty darn close?

nate
07-08-2009, 10:10 AM
Every year there is a blow out that brings this subject up, and every year somebody who eschews stats tries to say the phythag is a load.

I can set my sundial by it.. every year.

And at the end of the year every team comes pretty darn close to what the pythag said it should have done.

Instead of trying to prove the pythag incorrect while it's being compiled (and make no mistake it's the final result of the pythag that is the one that matters) the real test is finding all the outliers at the end of the season (or other seasons) and find out how wrong it is. Trying to address while it's in process is like picking up mercury.

Yep.

nate
07-08-2009, 10:11 AM
Why stop at pretty darn close if you can find a way to be more accurate than pretty darn close?

Because it's not a predictive measure.

BCubb2003
07-08-2009, 10:16 AM
Because it's not a predictive measure.

So why do we talk about it now?

Not being a pest, just trying to figure out what it's used for.

nate
07-08-2009, 10:21 AM
So why do we talk about it now?

Not being a pest, just trying to figure out what it's used for.

Because it's interesting to talk about. Both the times when there's a delta between pythag and actual record and when it matches. When there's a delta, why is it? Is it due to inaccuracy? Luck/something we can't measure? Both? Neither? Something else?

When it matches, well that makes it interesting to talk about how pythag manages to be accurate such a high percentage of the time.

Or have threads like this!

:cool:

jojo
07-08-2009, 10:22 AM
So why do we talk about it now?

Not being a pest, just trying to figure out what it's used for.

Mostly because those inclined to use stats evoke pythag after a blow out in jest and those disinclined to use stats fail to see the humor.......

Pythag provides a way to measure the gap between a teams actual record and what their record should normally be with such a run differential.

The meaning of a disparity between actual and pythag records can be debated and the cause is likely highly variable but probably most often just due to randomness (so most likely differences probably don't mean anything).

That said, run differential and winning percentage are related in an undeniable fashion.

westofyou
07-08-2009, 10:26 AM
Why stop at pretty darn close if you can find a way to be more accurate than pretty darn close?

Because life isn't a perfect circle? If you can get closer than the Pythag does have at it, but I've looked at it and I can't see anything that can get any closer.

And just because it doesn't wring the essence of truth from the subject in totality doesn't mean it it is without merits.

So go through the seasons, find the outliers (the very few) see how teams in the midst of seasons fool the pythag for awhile but then go to where their run differential thinks they should be. I've spent hours doing it to get my belief, while some seem to just look at the creators name in print and/or a couple of blow outs to get theirs.

edabbs44
07-08-2009, 10:30 AM
I've never gotten the love affair that some have with this theory, but here is what I have learned as a passive observer:

- It isn't designed to predict the future
- It isn't 100% accurate
- It shouldn't be looked at too seriously during the year, as it becomes closer to accurate as the season moves forward

So the only true worth that it could have is as a look back at the end of the season to see if a team played above or below the expectation. But I believe that I remember that there were some excpetions to this theory (i.e. stronger bullpens might skew the data) so basically this should be used as a novelty and nothing more.

BCubb2003
07-08-2009, 10:31 AM
Mostly because those inclined to use stats evoke pythag after a blow out in jest and those disinclined to use stats fail to see the humor.......

Pythag provides a way to measure the gap between a teams actual record and what their record should normally be with such a run differential.

The meaning of a disparity between actual and pythag records can be debated and the cause is likely highly variable but probably most often just due to randomness (so most likely differences probably don't mean anything).

That said, run differential and winning percentage are related in an undeniable fashion.

I agree it's often used in jest. And maybe it's nothing more than a "hmm, how bout that" otherwise. But if it's not really an indicator of something, then isn't it at the end of the season just a slightly less accurate way of showing who won the pennant? Don't people sometimes create a "standings" that show where the teams "should be" based on their pythags?

Raisor
07-08-2009, 10:38 AM
When building a team using the pythg from the pervious season can help give an idea of what you're shooting for for the next season.

If you're shooting for 95 wins, and you have a good idea what your offense is going to do, you can see where your pitching needs to be, etc.

In the middle of the season, and you're looking at the trade deadline, and you know you need to go 45-30 the rest of the way, and you know your pitching is good and has been consistent, you can see about how many runs you'll need for the rest of the season, and trade accordingly.

And if you're a fan on Redszone, you can use it to make fun of former poster Bad Fundamentals.

nate
07-08-2009, 10:38 AM
I've never gotten the love affair that some have with this theory, but here is what I have learned as a passive observer:

You don't understand it because there isn't a "love affair." That is hyperbole.


- It isn't designed to predict the future
- It isn't 100% accurate
- It shouldn't be looked at too seriously during the year, as it becomes closer to accurate as the season moves forward

So the only true worth that it could have is as a look back at the end of the season to see if a team played above or below the expectation. But I believe that I remember that there were some excpetions to this theory (i.e. stronger bullpens might skew the data)Mostly correct.


so basically this should be used as a novelty and nothing more.Incorrect.

Eric_the_Red
07-08-2009, 10:39 AM
If I'm reading correctly it seems the best "use" of Pythag would be to compare where a team's WL record sits compared to what Pythag says it likely should be. Then it could be determined if a team is likely to improve, decline or stay the same as the team has so far. Kind of like looking at a pitchers K/BB, K/IP, etc. to determine if the pitcher is performing over/under what is expected.

Or have I got it all wrong?

Raisor
07-08-2009, 10:42 AM
If I'm reading correctly it seems the best "use" of Pythag would be to compare where a team's WL record sits compared to what Pythag says it likely should be. Then it could be determined if a team is likely to improve, decline or stay the same as the team has so far. Kind of like looking at a pitchers K/BB, K/IP, etc. to determine if the pitcher is performing over/under what is expected.

Or have I got it all wrong?


You're about 75% correct. You forgot about using it to make fun of former RZer Bad Fundamentals.

pahster
07-08-2009, 10:46 AM
You're about 75% correct. You forgot about using it to make fun of former RZer Bad Fundamentals.

http://www.redszone.com/forums/showthread.php?t=38566&highlight=badfundamentals

He never did understand your "Pyth Thm of BBall," did he?

jojo
07-08-2009, 10:47 AM
Pythag is like any other stat..... it's useful as long as it's used properly.

Even "in season" pythag can be useful but caveats need to be appreciated. As it stands, the Reds look like overachievers based upon pythag. But we know they just got hammered two nights in a row so caveat......

On July 3rd, 2005 the Nationals were 50-31 despite a RS/RA of 335/333 and a pythag that suggested they should be 41-40. They had just finished a 32 game stretch where they posted a 26-6 record that included a 14-1 record in 1 run games. Safe bet.... expect a regression to the mean in the second half....

westofyou
07-08-2009, 10:48 AM
If I'm reading correctly it seems the best "use" of Pythag would be to compare where a team's WL record sits compared to what Pythag says it likely should be. Then it could be determined if a team is likely to improve, decline or stay the same as the team has so far. Kind of like looking at a pitchers K/BB, K/IP, etc. to determine if the pitcher is performing over/under what is expected.

Or have I got it all wrong?
Exactly, rate stats are nice indicators of a certain behavior, sure there are areas of the behavior that cause different results. But I like to think of the Pythag as data, all the items that make up the pythag are the meta data, they help define the data and are important too. But the Pythag just looks at the data point and reports it's findings. It's up to us to find out what bends it, just like when a guy has a 9 k per 9 rate the scouts define the type of pitches he uses to achieve that rate, that's the meta data.

M2
07-08-2009, 10:54 AM
Here's the thing with pythag - when your run differential starts to go in the wrong direction, you either have to correct that or your record will follow.

BCubb2003
07-08-2009, 10:57 AM
Exactly, rate stats are nice indicators of a certain behavior,

That sounds mildly predictive, with various caveats, and I'm OK with that.

bucksfan2
07-08-2009, 11:15 AM
If you were to do a mid season Pythag would 1 win be equal to a 5 run difference?

nate
07-08-2009, 11:49 AM
If you were to do a mid season Pythag would 1 win be equal to a 5 run difference?

I think it depends on the ratio of RA/RS rather than the difference.

For example:


RS RA pyth% W L
5 0 1.000 81 0
10 5 0.800 65 16
20 15 0.640 52 29
50 45 0.552 45 36
100 95 0.526 43 38
200 195 0.513 42 39
400 395 0.506 41 40

Obviously you're not going to win 81 games if you score 5 runs but you get the point.

Highlifeman21
07-08-2009, 11:53 AM
I've never gotten the love affair that some have with this theory, but here is what I have learned as a passive observer:

- It isn't designed to predict the future
- It isn't 100% accurate
- It shouldn't be looked at too seriously during the year, as it becomes closer to accurate as the season moves forward

So the only true worth that it could have is as a look back at the end of the season to see if a team played above or below the expectation. But I believe that I remember that there were some excpetions to this theory (i.e. stronger bullpens might skew the data) so basically this should be used as a novelty and nothing more.

Bad Fundamentals, is that you?

When did you get into tORG?

Falls City Beer
07-08-2009, 01:05 PM
That sounds mildly predictive, with various caveats, and I'm OK with that.

This is right, with what M2 said above appended to it.

Pythag to me is like a story with the last sentence ending in an ellipses....

it points the way to the ending of the story and locks certain determinants in place, but it's not "fate" or "prophecy."

Right now the Reds better be hoping for some deus ex machina.

kaldaniels
07-08-2009, 02:08 PM
Nothings perfect.

But how could the amount of runs you scored and the amount of runs you have allowed not be predictive in any way? And if they aren't, what is a predictive stat?

Sea Ray
07-09-2009, 12:07 PM
And at the end of the year every team comes pretty darn close to what the pythag said it should have done.



So it is predicative?

Raisor
07-09-2009, 12:18 PM
20 of the 30 major league teams are within +/- 2 wins of their ExW-L record.

28 of 30 are within 4 wins.

MWM
07-09-2009, 01:32 PM
So it is predicative?

Only if you know how many runs they're going to score and give up ahead of time.

WOY's comment was not a descriptor of predictive. If anything, it's prescriptive. As Raisor has pointed out, it's value is in understanding a framework for where a team is and how much improvement they need to make. There are methods that can give a team a decent estimate of how many runs they might score and give up, so it's only predictave in a roundabout way, and the level of accuracy is solely dependent on the accuracy of the other predictions.

Sea Ray
07-09-2009, 01:39 PM
Only if you know how many runs they're going to score and give up ahead of time.

WOY's comment was not a descriptor of predictive. If anything, it's prescriptive. As Raisor has pointed out, it's value is in understanding a framework for where a team is and how much improvement they need to make. There are methods that can give a team a decent estimate of how many runs they might score and give up, so it's only predictave in a roundabout way, and the level of accuracy is solely dependent on the accuracy of the other predictions.

So if our Reds are currently 5 wins better their Pythag and their record is pretty much .500, that doesn't predict that they'll finish with a sub .500 record?

nate
07-09-2009, 01:39 PM
Only if you know how many runs they're going to score and give up ahead of time.

WOY's comment was not a descriptor of predictive. If anything, it's prescriptive. As Raisor has pointed out, it's value is in understanding a framework for where a team is and how much improvement they need to make. There are methods that can give a team a decent estimate of how many runs they might score and give up, so it's only predictave in a roundabout way, and the level of accuracy is solely dependent on the accuracy of the other predictions.

Prescriptive is an excellent way to put it.

nate
07-09-2009, 01:40 PM
So if our Reds are currently 5 wins better their Pythag and their record is pretty much .500, that doesn't predict that they'll finish with a sub .500 record?

No.

MWM
07-09-2009, 02:36 PM
So if our Reds are currently 5 wins better their Pythag and their record is pretty much .500, that doesn't predict that they'll finish with a sub .500 record?

No, because it's not predicting how many runs they'll score and give up the rest of the way. Pythag is an output, you're describing an input. If you think a team's current run differential is going to continue out for the rest of the season, then yes, you can reasonbly predict what the record will be. Without runs scored and against, pythag does not exist. It all depends on the accuracy of the inputs.

Highlifeman21
07-09-2009, 03:43 PM
So if our Reds are currently 5 wins better their Pythag and their record is pretty much .500, that doesn't predict that they'll finish with a sub .500 record?

I would take it to mean that right now the Reds should be under .500, not that they'll finish with a sub .500 record.

The Reds could magically find some offense and cut into the RS vs RA gap, which would help the pythag, as well as hopefully yield more actually Ws.

M2
07-09-2009, 05:17 PM
Six years later (http://www.redszone.com/forums/showthread.php?t=11888)and we're still trying to explain pythag.

westofyou
07-09-2009, 05:23 PM
Six years later (http://www.redszone.com/forums/showthread.php?t=11888)and we're still trying to explain pythag.

Those fingers in my hair
That sly come hither stare
That strips my conscience bare
It's witchcraft

And I've got no defense for it
The heat is too intense for it
What good would common sense for it do

because it's witchcraft, wicked witchcraft
And although, I know, its strictly taboo

Ltlabner
07-09-2009, 05:32 PM
Six years later (http://www.redszone.com/forums/showthread.php?t=11888)and we're still trying to explain pythag.

You really should get out from behind your computer and watch some actual baseball games.

BCubb2003
07-09-2009, 06:01 PM
Six years later (http://www.redszone.com/forums/showthread.php?t=11888)and we're still trying to explain pythag.

Who could have predicted that? I guess pythag is kind of like the standings. It gives you a good idea, but you can change the future.

Sea Ray
07-09-2009, 07:06 PM
There's a lot of turnover here in 6 years. How many current Redszoners were around to read that stuff 6 yrs ago?

M2
07-09-2009, 07:14 PM
Who could have predicted that? I guess pythag is kind of like the standings. It gives you a good idea, but you can change the future.

I think the problem some folks have with pythag is that, at some inexact point, it begins to tell you what the team is. If the Reds keep getting outscored by 20+ runs a month (which happened in June), then we all know where the record is headed and we didn't need a formula to tell us that.

So pythag sometimes provides unwelcome sobriety to fans who want to suspend their disbelief.

BCubb2003
07-09-2009, 07:18 PM
There's a lot of turnover here in 6 years. How many current Redszoners were around to read that stuff 6 yrs ago?

How many current Reds players were on the team then? One. Can you name him?

MWM
07-09-2009, 08:09 PM
Ed E?

Maybe Weathers?

dougdirt
07-09-2009, 08:27 PM
How many current Reds players were on the team then? One. Can you name him?

Harang?

westofyou
07-09-2009, 08:32 PM
Harang is the longest tenured Red

Sea Ray
07-09-2009, 09:11 PM
Without cheating I'd be hard pressed to say which Red was around longer, Harang or Weathers. It's hard to believe it's been 6 yrs since the Harang trade

BCubb2003
07-09-2009, 10:08 PM
Yes, Harang's the one.

westofyou
07-09-2009, 11:24 PM
Prior it went Dunn--> Javy ---> Harang

Now it's Harang-->EE--->Arroyo/Phillips

Raisor
07-10-2009, 07:47 AM
Yes, Harang's the one.

Like Neo from the Matrix?

redsbuckeye
07-10-2009, 09:04 AM
Like Neo from the Matrix?

Does he know Kung Fu?

Big Klu
07-10-2009, 04:16 PM
The five players with the longest continuous service with the Reds are:

1. Aaron Harang -- 8/9/2003
2. David Weathers -- 4/4/2005
3. Edwin Encarnacion -- 6/24/2005
4. Bronson Arroyo -- 4/3/2006
5. Brandon Phillips -- 4/7/2006