Sports
Sports

Analysis: Best Clutch Players in Major League Baseball

| by Hardball Times

My first take on the clutch issue went nearly unnoticed more than a year ago.
Here I'm getting at it once more, having the benefit of both the wide audience provided by The Hardball Times and the data from FanGraphs.

Leaving aside questions such as whether clutch hitting does exist, let's try to define the traits of a clutch hitter.

First warning. My choice is to treat every game the same; i.e., a plate appearance with the game on the line is always important, no matter if it's a wild card race contest or a meaningless game between two teams out of contention. Other analysts will be able to replicate this study weighting for the importance of the game. (Yes, I'm looking at you and your Championship Leverage Index, Mr Sky Andrecheck.

Update: It appears that Sky is now an analyst for the Cleveland Indians; congratulations to him.

I think of a clutch hitter as someone who, whatever his underlying talent, does his best when the game is close and leaves his worst at-bats in mop-up situations. Thus, a player who has 10 home runs in his bat would be perfectly clutch if he hit them in the 10 most important plate appearances of his season; and at the same time he piled up his strikeouts in the most lopsided games.

Second warning. According to how I'm defining the clutch hitter, you don't necessarily want the clutchiest hitter at the plate in the deciding at-bat of the game. A perfectly mediocre hitter could be superclutch, but his production when the stakes are high might nonetheless be lower than that of a great batter with no clutch ability.

Two independent measures


The Run Value of an at-bat gives us a measure of how successful the batter was, without taking into account the base/out situation, the score, the inning. A home run is always considered worth around 1.4 runs and a strikeout close to -.3.

The Leverage Index, courtesy of Tom Tango, on the other hand, summarizes in a single number everything pertaining to the situation, while completely ignoring the outcome of the at-bat. And, most importantly, that number varies according to the potential impact the at-bat has on the final result of the game.

For the perfectly clutch hitter, sorting his plate appearances by Run Value would be equal to sorting his PAs by Leverage Index.
In more technical words, the correlation between the ranking of at-bats by Run Value and the ranking of at-bats by Leverage Index would be one.

In the real world we are going to find a lot of zero-correlations.

Careers


Here are the top 10 clutch careers (data starting from 1974, minimun 2,500 PAs), where the value under the column "clutch" is the correlation between the ranking of at-bats by Run Value and the ranking of at-bats by Leverage Index.

rank     First      Last     clutch
   1    Rodney     Scott      0.068
   2      Bill      Hall      0.057
   3    Willie   McCovey      0.054
   4     Garth      Iorg      0.054
   5     Danny  Bautista      0.052
   6      Jeff  Reboulet      0.051
   7   Thurman    Munson      0.050
   8     Brian Schneider      0.050
   9       Jim   Leyritz      0.048
  10      Mike   Matheny      0.047



And now the chokers (justifying the title of the article).

    First       Last      clutch
   Nelson    Liriano      -0.033
     Mike Lieberthal      -0.030
 Robinson       Cano      -0.030
    Willy    Taveras      -0.029
    Shane  Victorino      -0.028
    David   Dellucci      -0.027
     Eric     Byrnes      -0.025
    Denny    Hocking      -0.024
   Warren  Cromartie      -0.024
      Tom      Foley      -0.024



As you can see even the most extreme performances show very weak correlations (those at the top come out as statistically significant, thanks to the huge sample size, plus the fact that looking for five percent statistical significance many times leads to finding it about once every 20 runs, by definition—but this doesn't make them strong correlations).

Plotting all the career clutch numbers as a histogram gives us a typical bell-shaped figure, only slightly skewed to the clutch side.

image

Seasons


David Ortiz's famous 2005 season is not the clutchiest of the last quarter of century. Though pretty impressive (correlation .153) it's the 46th on my list, topped by Milton Bradley's 2000 season (minimum 150 PAs).

     First     Last Season     clutch
    Milton  Bradley   2000      0.232
      Mike  Matheny   2006      0.227
      Jeff Reboulet   2000      0.224
     Randy    Ready   1990      0.204
      Ivan Calderon   1985      0.203
      Rick   Miller   1977      0.200
       Dan    Meyer   1983      0.195
      Luis  Salazar   1985      0.192
       Ben  Oglivie   1981      0.191
    Rafael  Ramirez   1980      0.189



Speaking of Big Papi's career, here are his seasons with at least 150 PAs.

Season clutch
  1998 -0.048
  2000  0.071
  2001  0.006
  2002 -0.005
  2003 -0.006
  2004  0.007
  2005  0.153
  2006  0.033
  2007  0.008
  2008  0.015
  2009  0.022



In his career, he posted just three seasons with negative correlations, compared to eight with positive correlations.
Other players have constantly beaten zero*, as you can see below.

* From now on, I'm loosely referring to seasons with correlation above zero as clutch and to seasons with correlation below zero as choke. Keep in mind that most of them are actually not significantly different from zero.

    First      Last   clutch seas.   choke seas.
     Dave    Parker             16             2
    Eddie    Murray             15             6
     Omar   Vizquel             15             6
  Carlton      Fisk             14             5
    Ozzie     Smith             14             5
     Paul   Molitor             14             6
     Gary Sheffield             14             7
  Darrell    Porter             13             1
    Manny    Trillo             13             1
     Mark  McLemore             13             2




Anyway, confirming findings by other authors, in general there doesn't seem to be anything persistent in clutch hitting: Plotting each player's clutch values in odd years versus the same player in even years we see absolutely no correlation.

image

Fangraphs clutch stat


Fangraphs has its own clutch number, calculated differently; but if you read the introduction to the stat, you'll see the definition of clutch hitting is absolutely in line with the one I've outlined a few paragraphs above. Here is a scatter plot of the two different measures of clutch hitting; the correlation is 0.569 (Confidence intervals: 0.555, 0.583)

image

Eddie Murray


In his celebrated book Weaver on Strategy, the Earl of Baltimore talks about Eddie Murray as a player who constantly delivered in important situations, while taking something off when the game was in the books. Seeing his name high in the previous table supports (however lightly) Weaver's impression. Murray never posted a season like Big Papi's 2005, but often had correlations on the positive side (see below).

Season clutch
  1977  0.019
  1978  0.022
  1979  0.004
  1980  0.057
  1981 -0.077
  1982  0.025
  1983  0.028
  1984  0.029
  1985  0.071
  1986 -0.044
  1987  0.008
  1988 -0.067
  1989 -0.009
  1990  0.092
  1991  0.017
  1992  0.028
  1993 -0.046
  1994  0.081
  1995 -0.081
  1996  0.008
  1997  0.005



Perceived leverage, platoon, bullpen usage


Again, leaving aside the debate of whether clutch hitting exists, why don't we find greater clutch numbers?

  • Players care about the outcome of the game, but also about their paycheck: When the batting average is used in contract negotiations, a hit in a meaningless at bat has the same weight as a hit in a crucial situation. Thus, it might not be always a good idea to relax when the result is set in stone.
  • Say you are such a clutch hitter that you can play 20 percent better whenever you think it's showdown time; your perceived importance of the moment won't necessarily match with Tango's Leverage Index values.
  • If managers used their bullpen efficiently, hitters would often face tougher opponents in high leverage at-bats; they would also likely have the platoon disadvantage.


For the reasons outlined above (especially the last one), I would have expected the histogram in the "Careers" paragraph to be slightly skewed toward the choke side, but the opposite is true.
Maybe the cause of this is the selection of players with at least 2,500 PAs, but I did a similar histogram (not shown here) for players with shorter careers and it had a similar right skewed shape. Any idea on that?

Finally a couple of links with the full lists of clutch values: careers - seasons.