This paper was published in the November, 2000 issue of By the Numbers, the statistical newsletter of the Society for American Baseball Research (SABR).

 

ONE THING I DID NOT MENTION IN THE PAPER WAS THAT THE AVERAGE OPS FOR EXPERIENCED PLAYERS (2000 OR MORE PA) IN THE NONCLUTCH WAS .815 AND .808 IN THE CLUTCH, A DROP OF ONLY .007.  FOR THE INEXPERIENCED PLAYERS, THEIR NONCLUTCH OPS WAS .792 AND NONCLUTCH WAS .741.  A DROP OF .051, MUCH LARGER THAN FOR THE EXPERIENCED PLAYERS.  THE DIFFERENCE IN DECLINES IS .044. THAT IS HIGH IN BASEBALL TERMS.  THERE IS A RELATIVELY SMALL DIFFERENCE BETWEEN THE TWO GROUPS OF PLAYERS IN THE NONCLUTCH BUT A MUCH LARGER ONE IN THE CLUTCH SITUATIONS.

 

Clutch Hitting and Experience

 

Cyril Morong

Email


Click here to see my sabermetric blog called Cybermetrics

 

Experience may have a positive influence on performance in certain clutch situations, holding hitting ability in general constant.  Hitting ability and performance were measured using Production, the sum of slugging percentage and on-base percentage.

 

 

Introduction

 

Does clutch hitting exist?  That is, do some hitters hit better in the clutch than others or better in the clutch than they themselves normally hit?  If a hitter bats .300 from the seventh inning on we might think he’s a great clutch hitter but he might be a .300 hitter all of the time, so his clutch performance is not really so extraordinary. If a hitter bats .400 from the seventh inning on we might think he’s a great clutch hitter but wonder why he does not hit .400 all the time if he can do so in the clutch.  What would allow him to hit so much better in the clutch?  Why would he do better when the pressure is on?  Do some hitters do worse when the pressure is on, and if so, why?  If a hitter is a good clutch hitter does it mean he simply maintains his normal level of performance in the clutch and does not wilt under pressure?

 

David Grabiner has done an excellent study on clutch hitting and finds no evidence that clutch hitting exists.  The correlations he found of clutch hitting in one year to previous years were quite low (.01).[1]  This study takes a different angle.

 

In general hitters do worse in the clutch (defined below) than they do in non-clutch situations. (more on this in the Data section below) This study raises the question of whether or not more experienced hitters hit better in the clutch than less experienced hitters. 

 

The Data

 

Data from the 1995 major league season was used.  The clutch situation used was “Close & Late” from the STATSTM Player Profiles 1996 from STATS, Inc. This occurs when “a) the game is in the seventh inning or later and b) the batting team is either leading by one run, tied, or has the potential tying run on base, at bat or on deck.”  Players with 60 or more plate appearances (PA’s, walks + at-bats) only were included.  That was the qualification for being among the league leaders in this statistic in the book.

 

How do hitters perform in the clutch in general?  In major league baseball in 1995, for example, the slugging percentage (SLG) in clutch situations was .393 while in non-clutch situations it was .421.  For on-base percentage (OBP) these numbers were .3346 and .3353, respectively.[2]  So hitters generally did worse in the clutch than non-clutch.  This was true in 1994 as well.  What might explain this is that relief pitchers are often brought in late in the game and a hitter has not faced them yet.  Also, a left-handed pitcher might be brought in to face a left-handed hitter and a  right-handed pitcher might be brought in to face a right-handed hitter.  It is not clear that the pressure of the situation is what makes these numbers lower, because the pitchers are under pressure, too.

 

With it being more difficult to hit in the clutch, any hitter who simply maintain his normal performance level might be considered to be a good clutch hitter.

 

Hitters were divided into two groups:  the inexperienced, those with less than 2000 previous major league PA’s (prior to 1995)  and the experienced, those with 2000 or more previous major league PA’s.  If a player plays every game and gets 4 PA’s per game, he would have about 2000 after 3 years.

 

Performance in both clutch and non-clutch situations was measured using Production (PROD), which is the sum of SLG and OBP.[3]

 

The Model

 

An ordinary least squares regression was run using a hitter’s PROD in the clutch (CPROD) as the dependent variable and his PROD in non-clutch situations (NCPROD) as the independent variable.  Here are the results:

 

The Results

 

N = 175

Adjusted R2 = .187

F-Ratio = 41.11

 

Variable           Coefficient        T value             P value

Constant               0.248           2.990              0.003

NCPROD            0.657           6.412              <.001

 

The Adjusted R2 = .187 means that 18.7% of the variation in CPROD is explained by variation in NCPROD.  It seems that very little of how well a hitter does in the clutch is explained by is general ability or overall performance .  What, then, would explain the other 81.3% of the variation in clutch performance?  Some real difference in how players react to pressure?  Perhaps.  But the number of PA’s for each player is low, the minimum being 60.  So it does not necessarily mean that clutch performance is all that much different from non-clutch performance.[4]

 

The second regression was the same as the first accept that a dummy variable was added, with 1 for experienced players and 0 inexperienced.   Here are the results:

 

N = 175

Adjusted R2 = .208

F-Ratio = 23.869

 

 

Variable           Coefficient        T value             P value

Constant               0.242           2.944              0.004

NCPROD            0.631           6.208              <.001

DUMMY             0.052           2.355              0.020

 

The Adjusted R2 is a little higher with the addition of the experience variable, but not much.  So we can do a better, but only slightly better, job of predicting how well someone will do in the clutch when we include experience. There is little change in the results for NCPROD.  The experience variable, however, is statistically significant with a P value of  0.020.  Also, the coefficient estimate, in a baseball sense, is significant.  At 0.052, it means that a player’s CPROD is that much higher if he is experienced.  So, instead of maybe a hitter having a CPROD of, say, 0.800, he would have a CPROD of 0.852.[5]  That could mean, for example, 26 points more in both SLG and OBP in clutch situations for the experienced hitter.

 

Experience may teach a player to relax in the clutch and not press.  Perhaps they learn how to concentrate better or what pitches to expect.  But, whatever the reason, an experienced player, holding ability constant (by using NCPROD to measure this), will do much better in the clutch than an inexperienced player. 

 

Similar results were found using only SLG as a measure of performance and ability instead of PROD.

 

Conclusions

 

A variable, experience, was found to be positively and significantly related to clutch hitting. But the results are certainly not definitive or conclusive on the question of whether or not clutch hitting exists or if some hitters have a special ability to do better in the clutch. This study only looked at one season.  Player career totals could be studied to raise the number of clutch PA’s needed to qualify for the study.  This may reduce random noise that could be in the model.  Other clutch situations like “men on base” could be studied.  Another question is how much difference individual clutch hitting makes in winning games and in the standings.  These results indicate that clutch might exist and that it is worth further study.

 

 

 

Cyril Morong is an economics professor in San Antonio.  Cyril Morong, 723 W. French Place, San Antonio, TX 78212.  CyrilMorong@aol.com



[1] His paper can be found at http://www.baseball1.com/bb-data/grabiner/fullclutch.html  This is the website for The Baseball Archive.

[2]  Only walks, hits and at bats were used to calculate OBP.

[3]  PROD, at the team level, is more highly correlated with team runs than batting average or SLG or OBP.  So it is a better measure of performance.

[4] The minimum level of clutch PA’s could be raised, but then there are fewer players in the study.

[5] Similar results were obtained when the dummy cutoff was at 1500, 2500 and 3000.  At 1000, the dummy was not significant.  This means a certain amount of experience is necessary before it helps a player do well in the clutch.


Back to Cyril Morong's Sabermetric Research