Happy Thanksgiving! Regularly Scheduled Articles Will Resume Monday, December 1
March 2, 2010
Always a Bridesmaid, Never a Bride
For a certain type of completist, memorizing all the relievers that get roster spots by virtue of 13-man pitching staffs is the troublesome last mile. Any fantasy player worth his salt can rattle off nearly every team's closer, even ones whose job security is questionable. But closers aren't always selected purely on merit, and there is usually a flamethrower waiting in the wings to become the anointed. Those relievers who, for whatever reason, do not close ballgames but that nevertheless mow down hitters are fascinating to me. Ask most fans who the best non-closer reliever in baseball is, they might stumble a bit, but you'll get a pretty interesting discussion going.
Like Janus, Looking Forward and Backwards
Any statistical attempts to answer the question will be highly sensitive to the metrics chosen. We know ERA is unacceptable for relievers, but most rate stats suffer similar problems. We can look at component stats, but even then small sample size problems still persist. Besides, there seems to be a stronger edict for asking how a player actually did, rather than how he ought to have done, or could have done), when it comes to relievers. On the other hand, we cannot simply look backwards to evaluate talent, since we would inevitably single out guys who were merely lucky to be there.
To solve these twin problems, I came up with a toy to try to rank relievers. It's not a complicated toy. In fact, it only has two moving parts: SIERA and WXRL/LEV. I hope by now you've had a chance to peruse Eric Seidman & Matt Swartz's series on SIERA, but the gist is that it gives an estimation of a pitcher's controllable skills (fly ball rate, strikeout rate, ground ball rate, and walk rate) and considers how they interact with one another. Put simply, it's a way to evaluate the totality of a pitcher's skills while looking beyond contingent (or luck-based) factors. WXRL, on the other hand, is a metric based on win expectancy. It simply measures, compared to replacement and adjusted for quality of opposing lineup, how the likelihood of the reliever's team winning changed from when he entered the game to when he left. Because it is dependent on the situations in which a pitcher is deployed (and thus would favor closers especially), I have divided it by LEV, which is a measure of the leverage of the average situation the pitcher faced. (I realize there have been criticisms of LEV, but as a rough and ready way to de-leverage WXRL, it should do fine.) Remember, with SIERA, lower is better, but with WXRL/LEV, higher is better (because it is denominated in wins).
Damn the Torpedoes
I've taken these two measures of reliever performance, and ranked the best relievers (with more than 25 IP) in each measure for the 2009 season. I've assigned a number rank based on position in each list. I've then added the ranks on the two lists together, and re-sorted that list from lowest to highest. In other words, to rank high on my hybrid list, a reliever must have done well in both WXRL/LEV and SIERA in 2009. Here are the top three overall (SIERA and WXRL/LEV in parentheses):
Wuertz snuck in to tie Rivera, placing ahead of closers Huston Street (tied for fifth), Jonathan Papelbon (seventh), Joakim Soria (eight), Andrew Bailey (ninth), and Rafael Soriano (10th). Obviously, he's No. 1 on the non-closer list, presented below.
1. Michael Wuertz (2.21, 3.13)
2. Matt Thornton (2.56, 2.81)
3. Luke Gregerson (2.91, 1.96)
4. Pedro Feliciano (3.07, 2.04)
5. Mike Adams (2.09, 1.58)
6. Jeremy Affeldt (3.46, 2.68)
7. Neftali Feliz (1.99, 1.35)
8. Trever Miller (2.88, 1.60)
9. Sergio Romo (2.64, 1.41)
10. Nick Masset (3.32, 1.92)
I limited myself to the methodology of my toy. Which deserving relievers did such a method leave out? Which included relievers are undeserving?