Bullpen regression versus Nolasco regression


Yesterday, I brought up how the Marlins pen may regress due to BABIP and strand rate performance from last year, and how that may affect the 2010 Marlins. This was all in response to this BP blog post by Russell Carleton on the topic. I weighed in on the topic myself, but I don’t think I added a whole lot to the discussion. One thing I did find interesting was a comment by reader Kampfer about Ricky Nolasco and how many “extra” runs he gave up last year on his way to a 5.06 ERA. Could those runs make up for some of the lost value from the pen?


I took the CHONE-projected FIP of each of the four relievers of concern (Leo Nunez, Renyel Pinto, Dan Meyer, Brian Sanches) and used that as a decent estimate of defense-independent ERA. That gives us a projected value for runs allowed without defensive assistance. Because this is a linear measure of runs, I’m assuming that FIP assumes a random distribution. I’m also assuming that this will yield a BABIP of .300 (league average). Of course, since the Marlins are not a league average defensive team as a whole, I took a projected -25 runs per season defensively and prorated it to the number of balls in play expected to be allowed by the four relievers. In order to calculate total BIP and runs, I used my estimates for reliever innings pitched from my earlier projections and estimated that those four players should pitch around 300 of the 523 innings I projected for the pen in 2010.

PlayerEst IPFIP RADefProj. RA
Leo Nunez754.63-140
Dan Meyer654.42-133
Brian Sanches554.46-128
Renyel Pinto654.84-136

In total, we would expect the bullpen to allow 137 runs based on this estimate, yielding an expected RA of 4.74. Now, that doesn’t sound good in the least bit, but we need to be able to compare it to something. So I decided to look at the 2009 numbers as a comparison point. I added up the runs allowed by the same four relievers and included the inherited runners allowed to score. This gives a more realistic portrayal of runners allowed than simple use of ERA or RA would have. However, to be fair to the relievers, I knocked off two runs from each player to simulate their own potential runners left on base during their time on the mound. With these inclusions, I got an RA of 4.80 for last season, leading to approximately 139 runs allowed if last year’s performance were prorated to this season.

OK, so what’s the difference? How big are those two runs? Well, before I determine that, I decided to also include leverage in the discussion; after all, individual bullpen innings are more important than starter innings. Weighted by innings pitched, the average LI for those four pitchers was 1.29 last season. I gave some estimates for this season’s LI and it came out to be 1.33. Calculating WAR with full leverage impact for both those performances gives a value of 0.2 WAR for this season’s model and 0 WAR for last season, an improvement of 0.2 WAR for this year’s model. For reference, if you don’t include the inherited runners that scored, you get the 2009 version (including defense) being worth 5.6 WAR, a difference of 5.4 WAR from last season to this season.

To determine Nolasco’s impact, I did the same thing, without regarding the impact of leverage. This gave me a projected WAR with defense of 3.3 WAR and a 2008 WAR of 0.7. So including those inherited runners kills the value of the pen and makes it so that the Marlins’ relievers actually gain from the move to this season. Meanwhile, not including the inherited runners makes it so that Nolasco’s expected improvement from his defense-included performance would make up for only half of the added value from the pen’s luck of 2009. I’ll leave it up to the reader to decide which method is the better one. What do you guys think?