clock menu more-arrow no yes

Filed under:

An Exercise in Win Probability - The Grand Slam Single

New, 6 comments

I had the day off, so I spent the day learning the WPA spreadsheet, from If you don't have or use this spreadsheet, get it. Now.

It's a great way to look at any baseball game.

So I went back to the annals of Retrosheet and decided to play around with WPA in my most memorable game - Game 5 of the 1999 NLCS. This game is most often known for the "Grand Slam single," in which Robin Ventura couldn't complete his trip around the bases for his game-winning grand slam because he was mobbed before he reached second base. Before that, however, there had been 14.5 innings of extraordinary and engaging baseball. Let's take a look.

First off, I decided not to incorporate defense into the equation unless there were an obvious credit due that could be seen in play-by-play data. For example, an error was credited against the fielder, entirely. An outfield assist was credited to the fielder. The best way to do this would be to credit different outs at different levels - for example, a K would go 100% to the pitcher, and a fielding out would be split at some proportion. I didn't have the information / a tape of the game to make any judgements, and I didn't want to be subjective.

So here's the graph, or "What I did on my day off":

For those new to win probability, the pink line represents the "P-value" of a situation, or basically, the "importance" of the situation. The blue line represents the percent chance for the Mets to win the game.

The MVP of the game for the Mets was neither Robin Ventura nor Todd Pratt, according to WPA. Rather, it was Orel Hershiser, who came into an extraordinarily high-leverage situation in the 4th inning and slammed the door shut. This was a game for the pitchers, no doubt.

PLAYER          WPA

Hershiser       .329
Pratt           .297
K. Rogers       .284
Remlinger       .284
Mullholland     .252
J. Franco       .235
Rocker          .201
Maddux          .172
C. Jones        .149
Benitez         .142
Springer        .142
Hamilton        .135
Dunston         .110
Agbayani        .100
Lockhart        .095
M. Franco       .072
E. Perez        .068
Mahomes         .054
B. Boone        .047
Battle          .042
Nixon           .022
Olerud          .022
Wendell         .015
Myers          -.004
Weiss          -.010
Mora           -.014
Williams       -.022
Jordan         -.046
Fabregas       -.060
Bonilla        -.064
Ventura        -.069
Hernandez      -.073
Guillen        -.077
Piazza         -.089
Henderson      -.095
Alfonzo        -.125
Klesko         -.133
Hunter         -.146
Dotel          -.153
G. Williams    -.252
Yoshii         -.275
A. Jones       -.310
Ordonez        -.462
McGlinchy      -.790
It's amazing how important Todd Pratt's game-tying walk was. At that point, with the bases loaded and one out, the game's result was pretty certain.

The other thing of interest was that Alfonzo's sacrifice bunt with 1st and 2nd and nobody out actually increased the win probability, albeit slightly. That might be the only sure-thing sac bunt situation - two on, no one out, down by 1 run.

So try that spreadsheet, if you get a chance. It's a lot of fun, and it absolutely changes the way you watch a game.

I'd love to see hitting totals for any player at the end of the year to see which players are really worth the most wins. In essence, this is the ultimate MVP stat. The league leader SHOULD be the MVP. The only problem is breaking up the "defense" aspect into "pitching" and "fielding." But that'll come with time.

Acknowledgments: I got the spreadsheet from I used the HTML from Eric Simon's "Amazin' Avenue" for the picture-image. All of the play-by-play data is from Retrosheet.