A small update on “clutch relief”

A few months ago, I had looked at whether relievers could be “clutch” in their performance.  That is, do they perform better when the game is particularly on the line as opposed to when it is just garbage time.  There’s been a good amount of work on clutch ability as applied to hitting which has largely found that there is no particular skill associated with hitting in the clutch.  Good hitters are good hitters independent of the circumstances of the game.  I found in that study that the same finding was true for relief pitchers.  Relievers, as a whole, didn’t appear to have any ability to “turn it up” when the game was more on the line either.
The older study was made possible by a new definition of “clutchness” proposed by Tango Tiger and David Appleman of Fan Graphs.  The definition was based on how much win probability a player had added to his team and what the leverage of each situation was.  For the curious, the definition was the sum of WPA for a player, minus how much WPA he would have contributed had all of his at-bats occured in an average game situation (that is, a leverage value of 1.00).  sum WPA – sum (WPA/LI).  However, following some discussion on Baseball Fever between Tango and former StatSpeaker Matt Souders, Matt pointed out that there was a small flaw in that definition, one that could be a big flaw for those relievers who often pitch in high leverage situations.  Hitters, generally have an average leverage value for when they hit around 1.00.  But, relievers are often put into situations based on the leverage index at the time.  Even some hitters just by sheer luck end up with average leverage scores (pLI) a good distance away from 1.00.
Matt suggested that the definition of clutch be revamped to read (sum WPA / pLI) – sum (WPA/LI).  I wanted to see whether my original conclusion still stood.  I assumed that everything that happened to the defensive team was the pitcher’s fault or credit (a bad assumption, but again, we sacrifice precision for direction) and calculated every pitcher’s clutch performance per year (only in plate appearances where he pitched in relief).  I looked at the years 2003-2006, and included a pitcher only if he faced at least 100 hitters in relief in that season.  The intra-class correlation was -.059, which means that between 3 and 4/10ths of one percent of relief performance can be understood through “clutch ability.”
For what it’s worth, I checked on clutch hitting as well using much the same method and found an ICC around .015.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: