The most important batting statistic

[spinowner]

OK, this is a proposed statistic for the stat-heads on this site to consider.

The purpose of playing a baseball game is to win it. My proposal is to quantify how much the outcome of a batter's plate appearance affects his team's chance of winning the game. That is the most important measure of a batter's performance. Batting average, OPS, all else, is secondary.

I am aware that win probabilities have been researched although I don't have the time, ability or resources to investigate that myself. For those reading this who aren't aware of win probability, or WinProb as I think it's often called, it's the likelihood of a team winning a game starting from a particular point in the game. The parameters taken into account are the inning and number of outs, number and position of baserunners, run differential and whether the team is visiting or home. The win probability is calculated from the percentage of times a major league team that was at that particular point has gone on to win a game. (As you can imagine, collecting these data is an extremely tedious and time-intensive process.)

I'd call this proposed stat ∆WinProb. (Delta is the symbol used in statistics for "change in". It can be positive or negative). For example, in the bottom of the ninth inning with two outs, the bases empty and the score tied a base on balls has a much lower ∆WinProb than a home run. In the bottom of the ninth with two outs, the bases loaded and the score tied a base on balls is equal in value to a home run.

There are at least two variations of this to consider. One is the cumulative ∆WinProb, and another is ∆WinProb per plate appearance.

There are many other factors to consider. One example: how should a batter be credited when an error occurs? I think the batter should receive some sort of partial credit for putting the ball in play but probably not as much as for a base hit.

I'd enjoy very much reading comments about and criticisms of this idea.

[Craig Arko]

If you get a chance, post some examples of how this calculation works in practice. Do you consider it to be predictive?

[ashbury]

http://www.fangraphs.com/library/misc/wpa/

 

Win Probability Added sounds like something you would be interested in.

[spinowner]

 

If you get a chance, post some examples of how this calculation works in practice. Do you consider it to be predictive?

Go to ashburyjohn's link. I haven't investigated it at all so I'm not the one to whom these questions should be addressed. It was just an idea I had.

[spinowner]

 

http://www.fangraphs.com/library/misc/wpa/

 

Win Probability Added sounds like something you would be interested in.

This is essentially what I wrote. I had assumed someone else must have investigated it.

I only briefly scanned the article and I'm sure there must be more research and analysis that has been done, is being done and will be done. I hope to see the day when this stat is fine tuned and reliable enough to give us the best measure of the quality of a player's plate appearances, and that this would be the primary stat cited by fans, media and sabermetricians.

 

[ashbury]

It's an interesting stat, but IMO it still is swayed too greatly by what your teammates do before and after you.

 

If you hit a 3-run homer that raises your team's chance of winning a close game by a lot, it depends on those two guys getting on base in front of you - otherwise it's a solo homer that raises the chances by a much smaller amount.

 

Ditto for you hitting a walk-off RBI in the ninth to win 6-5 - without those other 5 runs that same RBI didn't affect the win probability that much.

 

Ditto for having a powerful lineup around you, also with solid pitching, that typically wins games 8-2 - or a weak one that is typically on the losing end of those scores - your contribution may not add much to the overall probability of winning, through no fault of your own.

 

Still, it has a kernel of merit to it that other opportunity based counting stats like RBI don't.

[Paul Pleiss]

WPA is a good way to see how certain events changed the course of a game, but they are not, IMO, good at predicting future performance. As Ashbury stated, they are highly skewed based on opponent/teammate performance.

 

It highlights certain plate appearances that were the most influencial in determining the outcome of the game, but doesn't shed light on a players performance going forward.

 

A player who hits a first inning homerun and then fans three times in a low scoring affair is likely going to have a nice WPA value, but a player who hits two home runs in the 2nd half of a blow out isn't going to get nearly as much WPA value, although certaintly provided more to the team, albeit in a less important situation.

[spinowner]

 

It's an interesting stat, but IMO it still is swayed too greatly by what your teammates do before and after you.

 

If you hit a 3-run homer that raises your team's chance of winning a close game by a lot, it depends on those two guys getting on base in front of you - otherwise it's a solo homer that raises the chances by a much smaller amount.

 

Ditto for you hitting a walk-off RBI in the ninth to win 6-5 - without those other 5 runs that same RBI didn't affect the win probability that much.

This is exactly the point of this statistic. It's to evaluate how a player performs in a given situation relative to the best possible performance in that situation. It's more important to hit a home run with two runners on base than it is to hit one with the bases empty.  It's more important to drive in a run with the game tied in the bottom of the ninth inning than it is to do so in the first inning.

[spinowner]

 

Ditto for having a powerful lineup around you, also with solid pitching, that typically wins games 8-2 - or a weak one that is typically on the losing end of those scores - your contribution may not add much to the overall probability of winning, through no fault of your own.

 

This criticism is well taken and has a lot of merit. A player whose team is involved in a lot of close games will have more opportunities to greatly increase WPA. This makes me realize that I didn't finish my thought in my initial post. My idea is not just to measure a player's WPA but also to measure it relative to the highest possible WPA for the situations that the player is involved in. That means a player striking out with two outs and the bases loaded in the bottom of the ninth would be penalized more for that than for striking out to lead off a game.

 

[lightfoot789]

You need a variety of statistical performers to have a successful offense.  A team of great OBP guys who can't hit to save their life or find a way to get to 2nd base won't win you a lot of games in the long run.  You need those contact guys and double guys and home run guys to help make all these offensive stats work in a winning way.

 

Does batting order determine offensive performance?  Do some players perform the same regardless to where they hit?  Good players find a way to make their skill set work within a team setting.  Or should I say Winners find a way to make their skill set work on winning teams.  

[spinowner]

 

WPA is a good way to see how certain events changed the course of a game, but they are not, IMO, good at predicting future performance. As Ashbury stated, they are highly skewed based on opponent/teammate performance.

 

It highlights certain plate appearances that were the most influencial in determining the outcome of the game, but doesn't shed light on a players performance going forward.

 

A player who hits a first inning homerun and then fans three times in a low scoring affair is likely going to have a nice WPA value, but a player who hits two home runs in the 2nd half of a blow out isn't going to get nearly as much WPA value, although certaintly provided more to the team, albeit in a less important situation.

I didn't say it was predictive, and in fact I'm sure it's not. I'm just trying to find the best way to evaluate how much a player has done to improve his team's chance of winning a game. And my apologies for not including in my original post that I'd also want to compare the actual WPA to the maximum possible WPA. (By the way, I'd like to see an example of a baseball stat that is predictive.)

[Paul Pleiss]

 

I didn't say it was predictive, and in fact I'm sure it's not. I'm just trying to find the best way to evaluate how much a player has done to improve his team's chance of winning a game. And my apologies for not including in my original post that I'd also want to compare the actual WPA to the maximum possible WPA. (By the way, I'd like to see an example of a baseball stat that is predictive.)

I think we're in agreement here. I was not trying to imply that WPA is flawed because it is not a strong predictor of future performance. I suppose I was just trying to qualify the value gleamed from WPA.

 

There's a table at the bottom of the FanGraphs article about what good/bad WPA values look like over the course of a season. 6+ is excellent. Lots of good links for further reading at the bottom of the article too, if you really want to get more into the weeds about WPA. The FanGraphs stat library is an excellent way to stretch the baseball statistical part of your brain.

[ashbury]

This is exactly the point of this statistic. It's to evaluate how a player performs in a given situation relative to the best possible performance in that situation.

And if a guy isn't offered many of those situations, his point total by this metric is going to lag other players who might not actually be as good.

 

As you go on to say later on, the sum needs to be normalized to something else in the denominator. But you have to be careful, as a player on a really bad team (with a tiny denominator) might be especially subject to small sample size variation.

 

I found this stat defined at baseball-reference.com:

 

 

 

WPA/LI Situational Wins. Sum of each plays WPA divided by the play's leverage index. SUM(WPA/LI) for all plays. This is similarly scaled to WPA, but removes the context from the outcome, so for this stat a player with 30 home runs all in blowouts would look very similar to a batter with 30 home runs all in tie games. They would look much different in WPA. Generally used for a season or career.

[jimmer]

RE24 is a good stat as well.

 

'Unlike wRAA, wRC, or Batting Runs, RE24 is context-dependent and assigns more credit for hits with men on base than with the bases empty. With league average set to 0, hitters with positive RE24 are creating more runs than we would expect given the situations they have been placed in and pitchers with positive numbers are preventing more runs than average given the situations in which they have been placed.

On the hitting side, RE24 is a measure of how well hitters are capitalizing on their opportunities while also not assigning extra credit (like RBI) to hitters who happen to come to the plate with men on base very often.'

 

'RE24 is valuable because it provides an answer to a different question than context-neutral run values like wRAA, wRC, and Batting Runs for hitters. If you want to know how frequently a batters hits a single or a double and how valuable his performance would have been in a neutral context, one of the other run value statistics is your best bet. Those stats tell you what happened in the batter’s box. RE24 tells you how what happened in the batter’s box impacted the context of the inning.'

 

 

http://www.fangraphs.com/library/misc/re24/