Players with the most top-100 Win Probability Added games in 2011
Posted by Andy on August 4, 2011
I love Win Probability Added. Although it has its flaws (notably no good way to assign credit or blame for significant base running events or errors) I think it's very useful for getting a sense of just how helpful a player has been to his team's cause.
I have isolated the top 100 games in 2011 for individual player WPA.
Here are the players with multiple entries in the top 100:
| Rk | Player | Year | #Matching | PA | AB | H | 2B | 3B | HR | RBI | BB | SO | SH | SF | IBB | HBP | GDP | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Prince Fielder | 2011 | 3 | Ind. Games | 16 | 15 | 7 | 1 | 0 | 3 | 8 | 1 | 3 | .467 | .500 | 1.133 | 1.633 | 0 | 0 | 0 | 0 | 0 |
| 2 | Daniel Descalso | 2011 | 3 | Ind. Games | 13 | 11 | 8 | 2 | 1 | 1 | 7 | 2 | 1 | .727 | .769 | 1.364 | 2.133 | 0 | 0 | 1 | 0 | 0 |
| 3 | David Wright | 2011 | 2 | Ind. Games | 8 | 7 | 4 | 1 | 0 | 2 | 4 | 1 | 2 | .571 | .625 | 1.571 | 2.196 | 0 | 0 | 0 | 0 | 1 |
| 4 | Justin Smoak | 2011 | 2 | Ind. Games | 8 | 8 | 4 | 1 | 0 | 1 | 5 | 0 | 0 | .500 | .500 | 1.000 | 1.500 | 0 | 0 | 0 | 0 | 0 |
| 5 | Nate Schierholtz | 2011 | 2 | Ind. Games | 12 | 12 | 6 | 1 | 0 | 2 | 5 | 0 | 1 | .500 | .500 | 1.083 | 1.583 | 0 | 0 | 0 | 0 | 0 |
| 6 | Alexei Ramirez | 2011 | 2 | Ind. Games | 10 | 8 | 5 | 0 | 0 | 3 | 6 | 1 | 1 | .625 | .667 | 1.750 | 2.417 | 1 | 0 | 0 | 0 | 0 |
| 7 | Albert Pujols | 2011 | 2 | Ind. Games | 11 | 8 | 6 | 1 | 0 | 3 | 7 | 2 | 1 | .750 | .818 | 2.000 | 2.818 | 0 | 0 | 1 | 1 | 0 |
| 8 | Eric Hosmer | 2011 | 2 | Ind. Games | 10 | 9 | 3 | 0 | 0 | 1 | 4 | 1 | 2 | .333 | .400 | .667 | 1.067 | 0 | 0 | 0 | 0 | 0 |
| 9 | Travis Hafner | 2011 | 2 | Ind. Games | 9 | 9 | 5 | 0 | 0 | 2 | 6 | 0 | 1 | .556 | .556 | 1.222 | 1.778 | 0 | 0 | 0 | 0 | 1 |
| 10 | Adrian Gonzalez | 2011 | 2 | Ind. Games | 12 | 11 | 6 | 1 | 0 | 2 | 6 | 1 | 3 | .545 | .583 | 1.182 | 1.765 | 0 | 0 | 0 | 0 | 0 |
| 11 | Johnny Damon | 2011 | 2 | Ind. Games | 9 | 9 | 3 | 1 | 0 | 2 | 5 | 0 | 1 | .333 | .333 | 1.111 | 1.444 | 0 | 0 | 0 | 0 | 0 |
| 12 | Jay Bruce | 2011 | 2 | Ind. Games | 15 | 12 | 6 | 0 | 0 | 2 | 4 | 3 | 1 | .500 | .600 | 1.000 | 1.600 | 0 | 0 | 1 | 0 | 0 |
Prince Fielder and, umm, Daniel Descalso?
August 4th, 2011 at 10:25 am
It's a running joke on Cardinals broadcasts this year that Descalso should only play from the 7th inning on, as he has done little in the first few innings when he starts.... but has had more than what should be his share of big time hits in "clutch" situations late in the game....
August 4th, 2011 at 10:26 am
Scott Cousins was 0-1 with a walk, but that was the 73rd highest score of the whole year?
August 4th, 2011 at 10:29 am
Jeff, that's a good example of the caveat in my original post. Cousins gets credit for the tying run scoring in the bottom of the 9th when he reached on an error. There's no really good way to handle WPA in such cases.
Here's the box score for that game:
http://www.baseball-reference.com/boxes/FLO/FLO201104250.shtml?utm_source=direct&utm_medium=Share&utm_campaign=ShareTool
August 4th, 2011 at 10:53 am
@1
With apologies to someone else, Evil, Daniel Descalso should be named "Mr.Late-Innings". Ah, darn ......... no one laughed!
I think the mixture of names on the list and the individual games they were in, brings out some of the nuances of WPA. I don't understand that statistic very well and would like to learn more about it and the plate-appearance circumstances that contribute to it.
August 4th, 2011 at 12:53 pm
SInce both teams start off with a 50% chance (.500 in decimal form) to win a game, the winning team has a combined WPA of .500 in a game (batting plus pitching), and the losing team -.500. Josh Hamilton, the player with the biggest WPA this season at 1.084, therefore did enough to win two games, if only his teammates could tread water and maintain a zero WPA! In that game his teammates actually had a negative WPA of -.391 batting and -.192 pitching (it doesn't quite add to .500, likely due to round-off error).
August 4th, 2011 at 1:15 pm
@5
Whiz, I grasp the big-picture of what WPA is, thank you.
And I understand that the entire-game WPA is the sum of the individual WPA's for each plate appearance.
What I find more challenging is relating change in WPA for the team to the specific plate appearance circumstances, inning and base-running events.
Am I correct in thinking that WPA for a plate-appearance extends to the end of your base running outcome, assuming that you got on base?
August 4th, 2011 at 3:29 pm
Neil, yes.
August 4th, 2011 at 3:34 pm
Neil,
WPA is the change in win probability (aka win expectancy) from before a PA to what it is after. The win probability is determined by the current base-out state (e.g., man on first with nobody out), inning, and game score.
Take the Josh Hamilton game on July 9 as an example:
http://www.baseball-reference.com/boxes/TEX/TEX201107090.shtml
When he came up in the bottom of the 9th there were two out, a man on first and the Rangers were trailing 6-5. Apparently the Rangers had about a 10% probability of winning the game at that point, and when he hit the walk-off HR, he increased it to 100%, for a WPA of 90%, or about .90, most of his WPA for the whole game.
WPA is calculated using a Markov chain analysis assuming league average batting profiles; only 10% of the time would an average team come back to win against an average pitcher in that situation. For more info on WPA, see http://www.baseball-reference.com/about/wpa.shtml
The bb-ref play-by-play gives the WPA for each play. Hamilton had a positive WPA for each PA in the game, even for the out he made since it drove in a run.
August 4th, 2011 at 3:48 pm
One more thing, Hamilton's is the 24th highest WPA for a game in the play-by-play era. The highest is Art Shamsky, on 1966-08-12, who had a WPA of 1.503 in a losing effort. This game has been discussed here before.
Shamsky entered the game in the top of the 8th in a double switch, hit a 2-run HR in the bottom of the 8th to give Reds an 8-7 lead, hit a solo HR in the bottom of the 10th to tie the game 9-9, and hit a 2-run HR in the bottom of the 11th to tie the game again at 11-11.
Unfortunately for the Reds, the Pirates scored 3 in the top of the 13th and Shamsky never got to bat again 🙂 Shamsky's teammates had a whopping -2.003 WPA, -.168 for batting and -1.337 for pitching (there's some round-off error again).
August 4th, 2011 at 7:41 pm
@7-@9
Whiz & Johnny T., so to put a specific set of circumstances on WPA, this afternoon Rajai Davis pinch runs for J.P. Arencibia in the top of the tenth with one out and gets picked off. Care to guess a WPA for that?
Escobar, Rasmus, and Bautista to follow, but I assume that WPA is based on an average hitter and takes no account of who is batting next.
Does WPA take into account how many runs have already been scored in the game, for example, if it's a slugfest or not, or does it just use league-average runs?
Not being critical of WPA, just trying to understand it. So will Davis get a
-0.100, -0.200 or some other WPA?
And is WPA pro-rated for events that happen after the plate appearance later in the game or is it etched in stone by the base-state and out-state at the time?
August 4th, 2011 at 9:21 pm
WPA is based on average expectation and does not account for the specific pitcher or batters. It does adjust for league- and park-adjusted scoring, but not on a game-by-game basis. So if you have one of those hurricane gale days at Wrigley, the WPA estimates will probably be low because they are not intended to account for the likelihood a team will need 15 runs to win. And the WPA does not change based on what happens later. It's just, before this PA the team had a 50% chance to win, afterwards it had a 55% chance to win, and that change is the WPA.
I'm not that sharp on knowing what win expectancies are at different stages, but I'll give a guess on Davis. You didn't say what base he was on. If he was on first, I'd guess it took Toronto from something like 60% win expectancy to something like 48%, so -0.12 WPA?
August 4th, 2011 at 9:34 pm
@11
Yes, JT, on first since he pinch-ran after Arencibia's walk. Thanks for the clarification. It seems like a small reduction in WPA, if you are correct, for the magnitude of the blunder. I think the count was 1-0 on Escobar, it may have been 2-0 when Davis got erased.
Rasmus subsequently doubled, scoring Escobar, after Escobar walked, but the butterfly effect says you can't project what woulda, coulda, shoulda ......
August 4th, 2011 at 9:42 pm
Don't forget to look it up tomorrow so we can see. I could be way off. With a runner on 1st and 1 out, there's probably only about a 25% chance of scoring, and even if they score, the home team still has a shot to tie or win, so...
For the batting team to have a big loss in run/win expectancy, they've gotta do something like GIDP with the bases loaded. -0.12 is pretty substantial though.
August 4th, 2011 at 10:23 pm
Schierholz is on their twice and that doesn't even count his walk-off catch against the Dodgers.
August 4th, 2011 at 10:33 pm
@13
Johnny Twisto, will do. And for your next act ...... lottery numbers?
@14
Nate, puhleez, what is a "walk-off catch"? (Afraid to ask.)
No more "walk-off" plays. ESPN may patent the phrase. 🙂
August 4th, 2011 at 10:53 pm
Neil @10,
Yes, it assumes league average batters, so WPA depends on the year or years you use to calculate league averages -- 1968 WPA would be different than 1930 or 2000 WPA. I don't remember if bb-ref adjusts WPA from year to year or not.
For the example you mentioned, according to my tables the win expectancy was about .50 before and about .39 after, for a WPA of about -.11. My tables are based on 2009 stats. Tango et al.'s tables from The Book, based on year 2000 stats, give about the same.
JT, you were really close on the WPA, although the absolute values were each off by the same amount. As you said, we can look at the bb-ref box score tomorrow to see if it agrees. BTW, with one out and a runner on first, there's about a 29% chance of scoring at least one run, so you were pretty close there, too!
August 4th, 2011 at 11:19 pm
@9:
I've heard of Art Shamsky before although I haven't heard the name in years. So I looked up his stats and it seems the guy had a lot of pop in his bat.
He didn't have a nickname though, and based on his record-setting performance listed here I'm going to call him Art "Blam" Shamsky.
August 4th, 2011 at 11:22 pm
Whiz, so your tables are showing that the home team has about a 65+% chance of winning the game if it clears the top of an extra inning? That seems high but I guess I'm not thinking about it right.
I'll blame it on my air conditioner which doesn't seem to be cooling efficiently, even though the temperatures dropped today.
August 4th, 2011 at 11:44 pm
JT, yes, you gain 15% by getting the visitors out in the top of the 9th or later inning. That's because there's a 30% chance of scoring a run -- in which case you win right away -- plus you win half of the other 70% later in the game.
August 4th, 2011 at 11:55 pm
I checked some games in 1968 and in 2000, and in the same situation the win expectancy was different -- for example, in 1968, a team that was tied going into the bottom of the 9th had a 61% chance of winning, instead of about 65% in 2000. That makes sense since it was a lower run environment -- those numbers suggest that there was only a 22% chance of scoring at least one run in any given inning in 1968, compared to 30% in 2000.
So the win expectancy on bb-ref does change from one year to the next. The numbers seem to vary within a season, so perhaps there are park effects, too. Looking at Colorado in 2000, they had a 68% chance of winning when starting the bottom of the 9th tied, so apparently at Coors Field in 2000 there was a 36% chance of scoring in any given inning.
August 5th, 2011 at 9:54 am
Whiz and JT, thanks for bring me up to speed on WPA.
Whiz, if I may, what are "my tables" that you refer to in #16? I'm going to try and look up Tango's win expectancy table and download a copy of it.
August 5th, 2011 at 11:10 am
Neil,
I generated my own win expectancy tables using the play-by-play data from Retrosheet, mainly because I wanted it updated for the current run environment. I was using them to determine the value of different save opportunities, and develop an alternative stat to saves (WPA in save situations).
Tango et al. describe it here: http://www.insidethebook.com/c01.shtml and give the table for the 6th inning -- you have to buy The Book to get the rest!
August 5th, 2011 at 11:39 am
@22
Whiz, first paragraph. Heavy duty!
I can only get a portion of Tango's win expectancy table, outside of the book, I think. On his site there a "crucial situation" win expectancy table, covering the seventh inning on and tied or up/down by a run, but that's all.
Philbirnbaum has a comma-delimited, text posting from which one could calculate win expectancy for every inning, number of outs and score. It would have to be pasted into a spreadsheet and cleaned up a bit. He says it was generated from Retrosheet data covering 1979-1990, but it would miss the higher-offence era.
http://www.philbirnbaum.com/probs2.txt
I guess I need The Book. A dynamic win expectancy table that had minor correction factors for the current run-scoring environment wuld be the best, I guess. But at that point, maybe we're talking super-computers.
August 5th, 2011 at 1:00 pm
Where are yesterday's games?!
Anyway, since B-R is coming up short, I checked at Fangraphs. Their numbers may be marginally different because of different park factors or something. They say the Davis pickoff improved TB's chances of winning from 49.8% to 59.6%, so -.098 WPA.
http://www.fangraphs.com/plays.aspx?date=2011-08-04&team=Rays&dh=0
August 5th, 2011 at 1:40 pm
@24
Johnny T., I, too, have been awaiting the BBRef WPA for yesterday's event with bated breath to see if you get the Karnak award.
A WPA 0f -0.098 doesn't seem punitive enough, looking at Rajai Davis's base-running blunder. But I guess that shows the difference between seeing something as a fan and seeing the same thing statistically with no emotional attachment.
Of course, one other thing concerning Davis and the assigned WPA. I 'm sure the win-expectancy tables assume an average baserunner, which Davis is not. He's at the top of the league in steals and has stolen both 2nd and 3rd in the same inning 8 separate times this year.
The point is, if he had been a little more patient, he probably could have stolen second and maybe third during Escobar's at bat anyway. But to get erased on a pick-off, when your whole role with the team right now is a designated pinch runner. Enough! I'll get over it.
If "If's and but's" were candy and nuts, we'd all have a Merry Christmas. I don't see Phils, Red Sox or Yanks fans having to type "If's and but's".
August 5th, 2011 at 4:32 pm
B-R is updated: http://www.baseball-reference.com/boxes/TBA/TBA201108040.shtml#play_by_play::none
They have the play as -.099, taking TB's chance of winning from 50 to 60% (they don't show the extra decimal place in the play-by-play). Only the 8th biggest play of the game. 4th biggest non-scoring play. 3rd biggest play at the time it happened.