« Greg Maddux | Main | Three Of A Rare Kind »

May 09, 2006

Win Probability Results

For those who are not aware, Win Probability is system in which a player is given credit toward helping his team win, based on play-by-play data and the impact each specific play has on the team's probability of winning. It's a neat stat.

And, you can find Win Probability results at FanGraphs.com. Here's the Yankees Win Probability totals to date:


It's a very interesting look at who has been helping and hurting the team to date.

Note Sturtze & Proctor. I never would have figured that one. Also, if you see Jaret Wright listed twice - it's not a mistake. The higher ranking here is Wright as a RP and the lower one is Wright as a SP. Either way, he's not helping. This stat also says that, depsite his season stats which look good, Sheffield is not helping the team so far this year. Fun stuff.

Posted by Steve Lombardi at May 9, 2006 11:18 AM


Which is why, with all due respect, the old saying goes "lies, damn lies, and statistics." I can't be convinced that Proctor's only marginally less hurtful to our team's chances to win than Tanyon Sturtze. Having watched Proctor have only two bad innings all year (Game 2 vs. Oakland and the 3-run homer vs. the Blue Jays 2 Fridays ago), there's no way he's "as bad as" Sturtze. Stats can tell you anything you want them to, no? I mean, that's why two politicians can get the same report and each spin it their own way.

Posted by: MJ [TypeKey Profile Page] at May 9, 2006 01:54 PM

I can't help but be skeptical of a stat that suggests that Gary Sheffield hurts the Yankees more than Tanyon Sturtze. How is that possible? And Sheffield's numbers are not that different from Cano's - in fact, better in almost every measure. Yet Sheffield comes out 60 points lower. It seems like something is screwy with this stat - Sheffield can't possibly be that much of a liability compared with his teammates.

Posted by: carla [TypeKey Profile Page] at May 9, 2006 01:57 PM

I will guess that the flaw in the WPA stat is that it allocates actual wins and losses. E.g., Aaron Small has pitched dreadfully, but the Yanks won the games in which he appeared, so he gets a WPA greater than zero. Sheffield may have done much of his great hitting in games the Yanks lost, leading to his negative WPA.

I'd like to draw a conclusion about the Bernie/Bubba comparison, but unfortunately, I cannot trust this statistic.

Posted by: David [TypeKey Profile Page] at May 9, 2006 02:14 PM

I don't think that it works that way David - but, I could be wrong. I think it results in "plays" that lead towards a good chance to win a game - and how a player does in those plays/chances. I don't think it goes by appearances in winnging games.

Posted by: Steve Lombardi [TypeKey Profile Page] at May 9, 2006 02:18 PM

I don't think it works at all. Take a look at Sunday's stat package. Proctor had 7 RSAA, which was in the top 5 in the AL. Are you telling me saving 7 runs over the league average is making the team lose ballgames? Dump it in the barrel with VORP, et al.

Posted by: hopbitters [TypeKey Profile Page] at May 9, 2006 03:32 PM

And I know that's how often Torre would like to use him, but there are not two Jaret Wrights on this team.

Posted by: hopbitters [TypeKey Profile Page] at May 9, 2006 03:35 PM

My goodness, such indignation toward a very simple stat. I wonder if you folks clicked on the link that Steve provided? It describes Win Probability, what it might be good for, what it isn't good for, and how it works.

For instance, Aaron Small doesn't get a WPA greater than zero just because the Yankees win a game, but he won't be hurt by WPA if he gives up runs in a blowout instead of a close game. Context does matter, but bad performances still get negative WPA -- just not as bad.

If you took some time to look at Proctor's actual appearances, you might see why he ranks so low in WPA.

4/4: gives up key hit in 4-3 loss. -.35
4/9: doesn't give up a run in a meaningless blowout. 0
4/11: pitches shutout inning in blowout loss, but the Yankees score five in the bottom of the eighth and he gets the win. The batters get the WPA credit, not Proctor. 1.5
4/13: pitches a shutout inning in a blowout win. 5 WPA
4/15: pitches three solid innings, but the Yankees are already losing 4-0. 6.5 WPA
4/18: enters with yanks already losing 7-4. 2.2.
4/21: relieves Wang in 5-4 game and walks in a run with the bases loaded, increasing O's lead to 6-4. Actually doesn't get credited with a run because he didn't put the runners on base, but -23 WPA even though he pitches two more innings of scoreless ball. He gave up a critical run and ERA doesn't capture it. WPA does.
4/25: pitches one inning in blowout win. 0 WPA
4/28: enters game with Yanks losing by only 2-0 and two men on. Gets two out but then yields HR to give Jays 5-0 lead. -6 WPA
5/1: gives up no runs after Sox have already taken 7-3 lead in ninth. 0 WPA
5/4: 2/3 inning after Yanks have taken 10-5 lead.
5/6: 2 innings after Yanks have 6-1 lead.

Summary: he's given up runs in close games and pitched well in blowouts. His WPA reflects that. No one is saying he's a bad pitcher. WPA clearly doesn't say he's "better" or "worse" than Sturtze (which you'd understand if you had read the article). WPA reflects contribution to wins, based on the context of the player's performance.

WPA isn't a perfect stat at all. It is what it is, which is dang insightful if you take a little time to think about it.

Posted by: studes [TypeKey Profile Page] at May 9, 2006 04:53 PM

I suspected that was the way it worked. Thanks for clearing that up

Posted by: Raf [TypeKey Profile Page] at May 9, 2006 05:55 PM

Nice explanation, Studes. I really enjoy your work at the Hardball Times.

It is an interesting stat. Am I wrong to say that this is manager-dependent? In order to do well, a relief pitcher would have to pitch well enough to move from low leverage situations to high leverage situations. I see how this could work for Proctor and, with him, it makes sense. The jury is still out on him until Torre moves him into high leverage situations. Unfortunately, I'm not confident at all in Torre's bullpen management.

Posted by: jonm [TypeKey Profile Page] at May 9, 2006 07:54 PM

jonm, yes, for relief pitchers WPA is very manager-dependent. A pitcher can't score high points unless the manager brings him in when it counts.

Posted by: studes [TypeKey Profile Page] at May 10, 2006 09:54 AM