Spring Cleaning – WPA - Looking at Wins and RBIs

by Jeff Zimmerman on Mar 16, 2010 3:00 PM EDT

I have several article/projects I began previously, but never published/finished. This week I will publish several as I don't see myself going back to them to finish them to original level I wished. I was thinking of just deleting them, instead I am hoping someone may find the partial work useful.

A while back in a thread I mentioned that I didn't like WPA (win probality added) as a stat because all it measured was RBI's for hitters and wins for pitchers and I really preferred using WAR added. I looked at the data then and now finally publishing it. First I would like you to answer the poll question:

Of a pitchers' win total or a batters' RBI, which one correlates more to WPA (data of qualified Pitchers and Hitters from some point near the end of last season)?

Answer is RBI's with an r-squared of 0.37 compared Wins r-squared of 0.25. Neither correlated well, but RBIs seemed to be the king over wins

While I was at it, went ahead and mapped WPA vs RAR for this same pitchers and hitters

WPA given to pitchers almost directly collerates with WAR (r-squared of 0.96). Hitting WAR and WPA do somewhat correlate (r-squared = 0.65), but not as much as pitching for this sample size.

Poll

Of a pitchers' win total or a batters' RBI, which one correlates more to WPA (data of qualified Pitchers and Hitters from some point near the end of last season)?

Wins

31 votes

RBIs

75 votes

106 votes | Poll has closed

Tweet Comment 3 comments | 1 recs |

Do you like this story?

Comments

I've been thinking for a little while that maybe a RPA stat would be useful.

Run Probability Added. On the one hand, it wouldn’t get around the problem of correlating with RBI for hitters. On the other hand, it would provide equal weight to late game, lopsided score situations as to early game, scoreless situations. The concept is basically the same: look at the probability that that a certain number of runs are scored in the inning before and after the plate appearance.

by RoxnSox09 on Mar 17, 2010 12:07 PM EDT reply actions

FanGraphs has that

in their Win Probability section, under the name RE24 (Run Expectancy, based on the 24 base-out states). WPA/LI also does something similar in that it gives equal weight to each event whether it is in a close game or a blowout or whether it is early or late in the game, but it considers things like inning and score to determine which events (i.e. walk, home run, out, etc.) are worth more or less in that situation beyond just how they effect run expectancy, whereas RE24 does not.

by Kincaid on Mar 17, 2010 1:12 PM EDT up reply actions