The Baseball Graphs Blog
Monday, May 29, 2006
WPA in the ninth
This article from the Sox Watch site has a great insight into scoring WPA for some important, complicated plays.
The final play was a tricky one, and an important one for WPA, because there was a lot on the line at that point, obviously. The Sox WP before the play was 85.4%. Crawford got a base hit to right, scoring Norton, and then Harris threw out Gathright trying to score. On plays like this, I split the action into two separate plays. First, Crawford’s hit scores Norton. If the play had stopped there, we would have had a 5-4 score, with the bases loaded, with a Sox WP of 76.2%. Tavarez therefore takes a -0.092 WPA debit for his role in the play.
From here on, however, Tavarez is not involved in the play, and so all the remaining WPA credit/debit is split among the fielders involved. We start with a WP of 76.2% and the ball in center field, and we finish with Gathright out at the plate, the game over, and a WP of 100%. There is therefore 0.238 of WPA to be distributed. I elected this to split this into three equal parts: One share went to the “Errors” category, which represents errors (and also good plays) by the opposing team. In this case, I consider Gathright’s decision to try for home to be a judgement error.
It’s not clear at this point whether Gathright was waved home or made the call on his own, but in any case, it was a poor decision that helped the Sox. The remaining two-thirds of the credit for this play I split evenly between Harris, who threw a perfectly-located strike to the plate, and Mirabelli, who hauled it in, got set, and made the tag on Gathright.
There is no “correct way” to calculate the WPA of plays like this. But in general, splitting a play into components and logging each one separately is the best way to handle it.
Tuesday, May 23, 2006
Cyril Morong has created a historical look at the greatest FIP years for individual pitchers. FIP, which stands for Fielding-Independent Pitching, is a relatively simple yet powerful pitching stat. It weights strikeouts, walks and home runs—those things that a pitcher is solely responsible for—to produce a fair ERA without all the surrouding “noise” of bloop hits and bad fielders.
Cyril did some more things to standardize FIP even further. His weights are slightly different than ones I’ve used in the past, he didn’t create specific ERA constants for each year and he “normalized” the range of strikeouts, walks and home runs in each year. I guess these are “ultra-normalized” FIP’s but I wonder if sometimes stats become so sanitized that they lose value instead of adding insight.
It would be good to see some more discussion of why he added his extra calculations and what the impact of them was.
Monday, May 22, 2006
The Blue Jay WPA Way
Thomas has a nice post up on Batter’s Box analyzing the Blue Jay’s bullpen with WPA.
Friday, May 19, 2006
Fielding Win Shares
J.P. has a nice review of fielding Win Shares in his blog, inspired by the 2006 Win Shares posted at the Hardball Times. Here are the links to his main posts:
I agree with J.P. that fielding Win Shares should have negative totals for certain players (something Bill James would almost certainly disagree with) but he underestimates how hard it would be in this post. Essentially, giving players negative fielding Win Shares would require totally different formulas than the current ones.
Thursday, May 04, 2006
The Latest Spreadsheet
I’ve uploaded a new version of the WPA spreadsheet. I’ve added a “poor man’s” Leverage Index to the spreadsheet, replacing Doug Drinen’s “P” as a way to measure the criticality of a situation. This Leverage Index is based on the difference between the Win Probability of the current situation and the Win Probability of a strikeout (in that situation). This approach was suggested by Tangotiger, and it seems to follow the distribution of his Leverage Index pretty closely.
If you’re using the WPA spreadsheet (particularly if you’d like to track the crticiality of each situation), I suggest you start using this version (1.7). As always, please let me know if you find any problems with it.
The spreadsheet (which requires Windows Excel 2002 or later) can be downloaded from ftp://ftp.baseballgraphs.com/wpa
Don’t rename the spreadsheet, or you’ll lose the macros!
Wednesday, May 03, 2006
I have a fascination with timelines, especially ones that can help us grasp the length of time it takes certain things to happen. Along those lines, you can’t beat this java timeline that covers everything from the Big Bang to the present day. Highly recommended.
Monday, May 01, 2006
There are two “historic” landmarks today in the world of Win Probability Added.
- Tangotiger has published the methodology behind his Leverage Index at The Hardball Times. Leverage Index, as a spinoff of WPA, may actually be the most useful application of WPA. It measures the criticality of a situation and serves as a guide for relief pitching, pinch hitting and other in-game strategies. As far as I know, Doug Drinen was the first to publish such a measure. His was called “P,” and he published it in the late 1990’s in the Big, Bad Baseball Annual. P was based on the number of outs required to finish out the inning without a run scoring. Tango developed the Leverage Index a few years ago and published a number of fine related studies on his website, exhibiting the usefulness of LI. Then Keith Woolner appropriated Tango’s label but created his own definition of LI in the 2005 Baseball Prospectus, based on the difference in WP when one incremental run is allowed. In other words, Drinen’s methodology was based on incremental outs, Woolner’s was based on an incremental run. Tango’s system (actually, he has four different approaches) is based on the variability in outcomes OR the impact of an excellent hitter or pitcher on WPA. His approach is definitely the most sound of the three, and I encourage you to read the article.
- Fangraphs has improved its game WPA graphs, adding a box score below the graph (hey, David, score by inning would be huge. In your spare time…) which vastly improves the usefulness of the graphs. More importantly, David has also added running individual WPA totals to the team stats, so you can see who is leading each team in WPA. He also totals WPA by batting, starting and relieving, which is a superb “diagnostic” way of looking at a team. Those folks tracking WPA for each game and posting it on your websites, don’t give up! David is applying the “gross form” of WPA, which means that he isn’t adjusting WPA for the run environment and he isn’t splitting WPA credit between pitching and fielding. There’s is still real value in following a specific game and logging its WPA. But I’m not complaining; this is a tremendous step forward for baseball fans (whether they know it or not) and David deserves many thanks and praise.