Clutch Hitting Statistics Are Poor Predictors

 

by Cyril Morong

 

Email


Click here to see my sabermetric blog called Cybermetrics

 

 

This is the kind of analysis that David Grabiner has done (go to my home page and look for “Clutch Hitting Links). Using data from ESPN’s site, I looked at the correlation between the differential between a how player hit in the clutch and the non-clutch for the years 2003-4. Email me if you want the data on these differentials

 

 

Using Runners in Scoring Position (RISP)

 

To qualify, a player had to have at least 100 ABs with RISP in both 2003 and 2004. I calculated each player’s non-RISP batting average and then his differential. For example, if a player batted .280 without RISP and batted .300 with RISP, then his differential would be .020. Some players had a negative differential. I did this for both 2003 and 2004.

 

For all players with 100 or more ABs with RISP in both years, the correlation between their differentials was was .011. There were 129 players.

 

This means practically no relationship at all between a player’s differential in 2003 and his differential in 2004. So if a player’s RISP average was much higher than his non-RISP average in 2003, there was little chance that it would be this way in 2004. No chance he would repeat his good RISP differential in 2004. For example, Adrian Beltre’s RISP average in 2003 was .049 higher than his non-RISP average. But in 2004, his RISP average was .056 below his non-RISP average.

 

The table below summarizes this for Beltre:

 

Year

Non-RISP AVG

RISP AVG

2003

0.229

0.278

2004

0.349

0.293

 

 

For players with 125 or ABS with RISP in both years, the correlation was -.027. There were 71 players. So again, no relationship. If you tried to predict a player’s differential for 2004 based on his 2003 differential, you would predict very badly

 

 

Using Close and Late (CL)

 

CL = Situations when the game is in the 7th inning or later and the batting team is leading by one run, tied, or has the potential tying run on base, at bat or on deck.

 

I looked at all players with 50 or more ABs in CL situations in both years. The correlation between their differentials in the two years was .112. So again, very little relationship. There were 84 such players.

 

For example, Carlos Pena went from having a positive differential of about .050 in 2003 to having a negative differential of about .048 in 2004.

 

The table below summarizes this for Pena:

 

Year

Non-CL AVG

CL AVG

2003

0.239

0.289

2004

0.248

0.200

 

For players with 75 or more ABs in CL situations in both years, the correlation is .125. There were 41 such players.

 

Back to Cyril Morong's Sabermetric Research