Wednesday, March 31, 2010
The case against ERA & introducing RPA
By Derek Czenczelewski, ESPN Stats & Info
Picture this: Top of the 9th inning, two out, men on second and third and the home team is up three. With the setup man unable to finish the game, the manager calls on the closer who he was trying to rest. The crowd goes wild as the bullpen doors open, but the noise quickly turns to silence as the closer gives up a two-run double. With the lead down to one run, the crowd rallies to the aid of the closer who induces a harmless ground ball. Game over.
Final stat line for the closer: (Save) 1/3 IP, 0 R, 0 ER, H, 0 BB, 0 K.
Situations similar to this occur somewhat frequently in baseball. For years pitchers have been judged on ERA, wins, losses, saves and strikeouts. While this formula may work for starters, it leaves a lot to be desired for relievers. Relievers can’t be evaluated by ERA. ERA is based on 9 innings and most relievers rarely pitch more than an inning or two per game. Runs, and even more importantly, inherited runs scored are far more important in measuring a reliever’s success.
Now, you may be wondering how unearned runs can be counted against a reliever. It doesn’t seem fair does it? Well, the reality is relievers have one job to do: Come into the game, and get out of jams. Whether a reliever comes in with the bases loaded or no one on base, his job remains the same. That job, again, is to come into the game and leave the game with the score the same way as when they entered.
So, with this in mind I set out to create a stat that took into account the two most important stats for a reliever: runs allowed, and inherited runs scored. Rather than divide this number by innings, I chose to divide by appearances.
The result is the following formula that I dubbed “Runs Per Appearance” (RPA): Runs allowed plus inherited runs scored divided by pitching appearances.
The results are simple, like the formula. Good pitchers had solid RPA averages while pitchers who were helped by errors and inherited runs scoring and not their own, were exposed. Check out the results below.
Here's another way to illustrate it. While both Jamey Wright (4.33 ERA last season) and Manny Acosta (4.34 ERA) had similar ERAs, their RPAs (1.123 and .639 respectively) differed greatly. The league average among all relievers last season was .935. That takes into account all long relievers, closers, setup men, and specialists. So Wright had a below-average performance in terms of RPA while Acosta actually had an above-average season in terms of RPA. His .639 RPA is far from Mariano Rivera or Joe Nathan status, but it was most certainly above average for a reliever.
Is (R)uns (P)er (A)ppearance (RPA) a perfect stat? Absolutely not. However, RPA does tell a much more accurate story for relievers than ERA. If nothing else, this should be something interesting to track in 2010.