PDA

View Full Version : Baseball-Reference adds Win Probability



RedsManRick
03-20-2010, 04:08 PM
Win probability is probably one of the most interesting ways to the tell the story of a game. Fangraphs has been doing it for a while, but B-R just released their version of it and I must say, it's pretty awesome. The graph clearly shows every play including the leverage (i.e. importance or clutchiness of the PA).

Check it out for the Reds final game of last season, a 6-0 win over PIT.

http://www.baseball-reference.com/boxes/CIN/CIN200910040.shtml

TheNext44
03-20-2010, 04:18 PM
Thx!

That is awesome, especially the mousing over each play and getting a breakdown of it on the graph. Great stuff.

I ♥ baseball-reference.

Redsfan320
03-20-2010, 05:15 PM
Awesome. BR really is a great site.

Thanks for posting. :)

320

oneupper
03-20-2010, 06:48 PM
I ♥ Win Probability, especially as a tool for strategic in-game decision-making.

It's a stat that is taking off in a big way. I've even seen it sporadically on some TV games.

kaldaniels
03-20-2010, 09:59 PM
I ♥ Win Probability, especially as a tool for strategic in-game decision-making.

It's a stat that is taking off in a big way. I've even seen it sporadically on some TV games.

Absolutely...when all is said and done, the goal is to win games. Great stat.

Johnny Footstool
03-21-2010, 01:36 AM
I think this stat is fairly pointless, because it measures events and situations that are beyond the player's control.

dougdirt
03-21-2010, 02:55 AM
I think this stat is fairly pointless, because it measures events and situations that are beyond the player's control.

It does give a player points for things beyond their control, but it is always rewarding you for helping the team (hit, walk, HBP, sac fly) and taking away from you when you are hurting the team (making an out).

This is the best way to view 'clutch' that we have right now.

redsmetz
03-21-2010, 07:12 AM
I ♥ Win Probability, especially as a tool for strategic in-game decision-making.

It's a stat that is taking off in a big way. I've even seen it sporadically on some TV games.

How is it possible to use this for "in game decision making"? I've really tried to understand what's being presented here, but to me this strikes me as fairly lifeless means to give a mathematical narrative to a game that has already occurred. It's really what makes all the new statistical analysis tools seem so soulless. And I'm completely missing how this can possibly be an in-game tool and how on earth there would be time for that.

My beef isn't with you on this, Oneupper, it's with the whole concept. I'm leaning gradually some of the value of more sophisticated analytic tools, although much of it, with their complex multi-lettered designations, strike me as a jumbled alphabet soup.

Even this tool leaves me with what I continue to stay, it always comes down to the flesh & blood players facing one another, the one succeeding, the other failing. And sometimes the one who shouldn't succeed, based on their previous record, does win out.

I'm really not trying to be a luddite on this. But honestly I don't much get it.

oneupper
03-21-2010, 10:17 AM
How is it possible to use this for "in game decision making"? I've really tried to understand what's being presented here, but to me this strikes me as fairly lifeless means to give a mathematical narrative to a game that has already occurred. It's really what makes all the new statistical analysis tools seem so soulless. And I'm completely missing how this can possibly be an in-game tool and how on earth there would be time for that.

My beef isn't with you on this, Oneupper, it's with the whole concept. I'm leaning gradually some of the value of more sophisticated analytic tools, although much of it, with their complex multi-lettered designations, strike me as a jumbled alphabet soup.

Even this tool leaves me with what I continue to stay, it always comes down to the flesh & blood players facing one another, the one succeeding, the other failing. And sometimes the one who shouldn't succeed, based on their previous record, does win out.

I'm really not trying to be a luddite on this. But honestly I don't much get it.

No, I get what you're saying. What happens is that WP is based around game "states", that is, what is your win probability in a certain game situation (down 1, bottom of ninth, man on second, one out). While the stat for players revolves around how each player contributed in getting the game from one "state" to another, it can also help in the decision-making process by identifying how important a certain "state" is at a given point of the game and acting accordingly.
Does pinch-hitting make sense?Base-stealing? Bunting? When? How? With whom?
Is the game out of reach? Does it make sense to put my closer in here? and so on.

Is there time for this stuff? Well, the tables exist and they use them on TV. Tango Tiger et al, wrote "The Book" which is all about this stuff.

Of course, it all means nothing if players don't execute. But putting your team in a better (more probable) position to win, makes sense IMO.

RedsManRick
03-21-2010, 10:45 AM
Even this tool leaves me with what I continue to stay, it always comes down to the flesh & blood players facing one another, the one succeeding, the other failing. And sometimes the one who shouldn't succeed, based on their previous record, does win out.

I'm really not trying to be a luddite on this. But honestly I don't much get it.

The one example I've seen used many times is that of "should I bunt here?". Theoretically, you could compare the possible outcomes and how they affect your win % and then decide, based on your own assessment of how likely it is that the guy will get the bunt down, etc., what to do.

In practice, I think you're right. No manager is going to do win% calculation while sitting on the bench. For me, it's more of an instructional tool; you can look at scenarios such as that ahead of time and develop a more well-founded decision for the future.

It's win% calculations that have shown that (successful) bunting early in games always reduces your chances of winning the game. At that point, the out given up is worth more than the base. That's where this sort of thing is helpful. Of course, it still requires some adjustment based on the particulars, as this is an average. If you're facing Roy Halladay and the runs are going to be very hard to come by, maybe it makes a bit more sense -- but generally speaking, it's pretty darn useful to understand how events contribute.

The other use is to communicate the flow of the game afterwards. The moments that are most exciting for fans just happen to align perfectly with those that have the highest potential to swing the Win% (aka Leverage). So this really tells the story from the fans point of view quite well, even if it doesn't attribute credit to individual players perfectly.

redhawkfish
03-21-2010, 11:21 AM
Probably a stupid question, but I am a newbie to the more sophisticated stats. Does this take in to account an individual batter's success against specific pitchers?

RedsManRick
03-21-2010, 12:36 PM
Probably a stupid question, but I am a newbie to the more sophisticated stats. Does this take in to account an individual batter's success against specific pitchers?

No, it is not adjusted for the particular people involved. It simply uses a big database of game states (score, base runners, outs) and looks to see how often each team won the game. So it's not about how likely a player is to have success -- it's about how the outcome of each PA affects the team's chances to win the game, in the abstract.

Obviously the particular players involve matter in reality. But for the most part it comes out in the wash. Think of this more along the lines of a box score, displaying how the game played out -- rather than as a real accurate measure of individual player performance.

gonelong
03-21-2010, 01:32 PM
Obviously the particular players involve matter in reality. But for the most part it comes out in the wash.

How much do they matter? IMO, that is what the numbers fail to tell us and therefore I don't feel it comes out in the wash. I wish they did, because I think there is enough information to make a solid study of teh differentiation of these scenarios between a 2-1 game vs. a 9-8 game.

For example, the decision to bunt against Lincecum with Wilson available in the pen is a completely different decision than to bunt with Ian Snell on the mound (and Capps in the pen) even when given the exact same scenario with inning, base runners, etc.

We both agree the who/when matters, but how much?

IMO it's quite possible (though I just made this up as an example) that the aggregate may be wrong for both scenarios. It might tell you bunting is a 50/50 decision, when in reality its a 40/60 for Lincecum and a 60/40 for Snell. The problem is, the data today will take both those scenarios and combine them to a 50/50 decision.

The problem with the way the numbers are presented today, is that I have no idea if my 2-1 game in the 7th inning with Lincecum or Snell on the mound (one side of league average to the other) skews the numbers inconsequentially, considerably, or drastically.

The problem is, IMO, the part where I really want the info (at the extremes) is where it fails to provide the punch I want.

As the information is presented today, it's at best a reference/guideline and should be taken with a huge grain of salt. The farther from the average any single player gets from the average player's skill set, the foggier the decision gets. Multiple players (+ and -) from the average makes it even foggier.

The thing is, maybe it doesn't matter as much as I think, but the numbers have never been presented in a manner in which I can make that determination.

GL

RedsManRick
03-21-2010, 01:42 PM
GL, I think we're more or less on the same page. I certainly don't know the answer to the "how much does it change based on specific players" question. And I'm fully willing to concede the point that the answer is quite important.

However, my argument is not that win probability should be used blindly. Rather, I thinks it's the best place to start. It certainly beats going with "the book", your gut, or what Hank Aaron did back when you played together (yes, that's a joke). Start with the average, the general expectation and adjust according to the specifics. Absent better understanding on the influence of specific players, I can't fault anybody for adjusting more than I would. But just because it's only a general guide doesn't discount it's utility. In a game in which you can't expect to score but a few runs (against a Lincecum for example), the base you'd acquire does become more valuable. But that's a pretty extreme case. Most pitchers aren't Tim Lincecum and most managers aren't just bunting early in the game against great pitchers. Those that do it tend to do it regardless of who is on the mound.

I think the burden of proof should be on decisions that disagree with average win probability. Again, that's not to suggest that the decision to bunt, for example, is necessarily wrong. I'm sure there are exceptions. But there should be a common understanding of standard expectations should be so that the claim of exception can be tested.

When it comes to bunting in particular, it seems that the assumption is that there's something special about scoring first and "getting that first run across". I think that for most coaches, it's about doing what they were taught 30 years or what they're expected to do because "the book" says to rather than as a result of a logical decision making process based on facts (all limitations admitted) and an argument that the current circumstance is substantially different than "average".

TheNext44
03-21-2010, 02:16 PM
First, I think Doug hit it on the head when he said that In Game Win Probability is best used in trying to understand the concept of "clutch." This is where it is most useful. It's going to take a lot of work to go through all the data, but I think eventually, this will help us answer that question that has been so difficult up to now.

It also can be useful in understanding in game decisions, if not directly telling us what is best to do. I think this stat best represents what Bill James just said in his recent interview with C. Trent, which is in this thread:

C. Trent - Talking with Bill James: Part 1 (http://www.redszone.com/forums/showthread.php?t=80862)


CTR: Is our understanding of the game - and I see where so many people are trying to understand the game in new ways - do we understand the game any more than we did in 1906?

BJ: No. We don't. Absolutely not. Our knowledge is much larger and our ignorance is much larger. Is one larger relative than the other?

In Game Win Probability at the same time increases our knowledge of the game and our ignorance. By itself, it probably reveals more that we don't know than gives us new information, especially when it comes to in game decisions.

However, it a huge step in the right direction for understanding the game better, overall. If we take what it does tell us, along with digging deeper into what it reveals that we don't know, in the end we should come up with a better understanding of what to do in each situation, as they arise. This also is going to take a lot of work, but it should be very helpful if we do the work.

gonelong
03-21-2010, 07:04 PM
GL, I think we're more or less on the same page. I certainly don't know the answer to the "how much does it change based on specific players" question. And I'm fully willing to concede the point that the answer is quite important.

However, my argument is not that win probability should be used blindly. Rather, I thinks it's the best place to start. It certainly beats going with "the book", your gut, or what Hank Aaron did back when you played together (yes, that's a joke).

We aren't far off, but I enjoy the discussion. :)

If we don't know how much it changes, then it's not fair to state with any certainty (or authority) that it beats going with "the book", etc. That's how I see it. :p:

Furthermore, who gets to do the adjusting? Likely the guy that played with Hank Aaron.

If we are going to test we need a baseline. Currently WPA doesn't provide a baseline for any given situation, just a rule-of-thumb.

The burden of proof is squarely on WPA IMO. The "gut", "the book", and how Hank Aaron did it back in the day all have been set as acceptable precedence. WPA is the party crasher and needs to overcome the burden of being better than "we've always done it this way".


In a game in which you can't expect to score but a few runs (against a Lincecum for example), the base you'd acquire does become more valuable. But that's a pretty extreme case. Most pitchers aren't Tim Lincecum and most managers aren't just bunting early in the game against great pitchers. Those that do it tend to do it regardless of who is on the mound.

While not many are Lincecum, all sorts of them are high5-6 ERA types, as high above the league average as Lincecum is below the league average. It's not a rare occurrence to face these kinds of outliers, the ones to the high side in the regular season, and the ones to the low side in the post-season.


When it comes to bunting in particular, it seems that the assumption is that there's something special about scoring first and "getting that first run across". I think that for most coaches, it's about doing what they were taught 30 years or what they're expected to do because "the book" says to rather than as a result of a logical decision making process based on facts (all limitations admitted) and an argument that the current circumstance is substantially different than "average".

Toss bunting out, that is too easy a target. Try replacing it with a SB attempt, a hit and run, a double-steal. It's a much more interesting argument at that point. :D

GL