Win Probability

You may have noticed (or not) that I haven’t been writing much recently. That’s because I’ve been working away at putting together a Win Probability section which can be found under the “teams” tab. We’re going to try and plot each and every game during the season and include the Win Probability Added for each player on each team. You can read all about Win Probability here:

The Hardball Times: The One About Win Probability (Dave Studeman)
Major League Baseball Graphs (Dave Studeman)

Since we’re automating these graphs, our calculations might vary slightly from other WPA charts you’ve seen, but they should be “good enough” until I can tweak them to death. All credit/blame for each play is assigned to the pitcher or batter. We do not take good/bad fielding into account. Hopefully there aren’t any bugs, but if you find any, please let us know.

Click Here For Win Probability Graphs…

Print This Post

David Appelman is the creator of FanGraphs.

Sort by:   newest | oldest | most voted

Tango said this about assigning credit/blame for the pitcher and fielder:
“Btw, I did WPA for 1999-2002 a few years ago (for every game in the league), and I calculated the pitching/fielding split in two ways: gave 100% credit to pitcher, and gave 100% credit to fielders. The result? Virtually NO difference. That is, regardless of whether we gave the credit to the pitcher or not on his own BIP, it had no impact in the end. This is of course DIPS talk. So, in the grand scheme of 4 years of data, it really doesn’t matter how you split up the pitching/fielding split. I gave the two most extreme splits I could give without a noticeable change. In a given game though, it certainly makes a difference.”

He also said:
“Make an educated assumption as to what the average fielder would have done (98% out, 70% out, 20% out, etc), and compute the win% based on those expectations. That is the “ending win%” for the pitcher is the “starting win%” for the fielder. The “ending win%” for the fielder is the actual ending win%.”

You’d think it would make a difference, but I guess it doesn’t. Since your process is automated I guess it would be hard for you to include fielding since you would have to subjectively look at EVERY game right? I don’t know how that would be even possible when automating WPA graphs (which seems like a good idea).

Oh, and did you get my emails? I know you must be busy. :)