The following article was part of Baseball Prospectus’ April Fool’s Day content for 2004.
One of the biggest controversies between sabermetrics and conventional wisdom has been the existence of “clutch hitters.” Those close to the game and how it’s played are convinced that some players have the ability to “rise to the occasion” and deliver key hits in the clutch. Even casual fans can point to dramatic game-turning events such as Kirk Gibson‘s pinch-hit home run in the 1988 World Series as evidence for clutch hitters.
Sabermetrics has grappled with this issue for the past two decades, trying to discover whether clutch hitting existed, who the clutch hitters were if they did exist, and how much effect they had on the game. Most studies focused on situations that could be defined by objective criteria that related to the subjective impression of being “clutch”–batting with runners in scoring position, and batting in the late innings of close games being the two most common examples.
However, in thinking about this recently, I realized that we had been approaching this in entirely the wrong way. Defining clutch in terms of a particular characteristic in a point in time fails to capture the common understanding of the term–delivering when it means the most to your team. Without having the larger context of the game in which to evaluate clutchness, any attempt to measure it is doomed to failure.
So, what larger context applies in this situation? Clearly, the outcome of the game is paramount. The goal of the team is to win the game (and secondarily, the pennant and the World Series, both of which are related to winning today’s game), and clutchness necessarily contributes to that outcome. It’s not how a player hits with runners on base that’s important, it’s how a player hits when his team wins that defines clutchness. Hitting well in losses is wasted effort, and contributes nothing to the team’s bottom-line results. True clutch hitters are the ones who save their best for when their team achieves victory.
We now turn to an examination of the data to search for evidence to support this theory. I examined the batting performance for all players in 2003 who had at least 50 PA in both winning and losing games. In order to compare the level of production for each batter in clutch (win) and non-clutch (loss) situations, I turned to the best single measure of offense known: On-Base Percentage Plus Slugging Average, or OPS.
There were 416 players who qualified with at least 50 PA in both clutch and non-clutch games. Comparing the OPS of these players in clutch and non-clutch situations, most had superior performance in clutch games, supporting the notion that players are selected from the right-hand tail of the normal distribution for clutch ability. We would expect such players to perform better in the clutch than not, and indeed only 28 players hit better in non-clutch losing efforts, meaning over 93% of players demonstrated some level of clutch ability.
In some of those other 28 cases, this may be an issue of small sample size, but can we really say that about, say, Chris Widger, whose OPS was 421 points higher in losses than in wins (.151/.213/.189 in clutch wins vs. .327/.353/.469 in non-clutch losses)? That 421 seems to be a pretty large sample, larger, in fact, than the number of qualifying players in 2003 (416), and thus is tough to ignore.
To take the study a step further, I looked at the clutch/non-clutch splits for all players with 100+ PA in two consecutive seasons since 1972. I also wanted to perform a more advanced analysis, so rather than relying on differences in OPS, I computed the ratio between clutch and non-clutch OPS, as division has a higher order of precedence than subtraction. Thus defining a new stat, ClutchRatio as:
ClutchRatio = ClutchOPS / non-ClutchOPS
ClutchRatio of greater than 1 indicates higher performance in clutch situations as I’ve defined them (we’ll call these hitters “Clutch”). ClutchRatio less than or equal to 1 indicates no improvement in performance in clutch situations (which we’ll label “Choke”).
I looked at the year-to-year behavior of ClutchRatio to see if the results were randomly distributed; that is, whether the totals tended towards a 50-50 split. Of course, an exact 50-50 split is not necessary, or even expected. For example, a 52-48 split would still yield logical results without causing problems. Over 7,300 data points (players with consecutive qualifying years of clutch/non-clutch data) were used in compiling this table:
Year 2 ClutchRatio <=1 (Choke) >1 (Clutch) Total Year 1 <=1 (Choke) 0.002447 0.034933 0.03738 ClutchRatio >1 (Clutch) 0.037243 0.925377 0.96262 Total 0.03969 0.960310
As you can see, the above table does not even closely resemble a 50-50 split between clutch and non-clutch, thus proving to six decimal places that this new theory view of clutch hitting correctly and conclusively demonstrates what it shows.
So, how do some of the top hitters in baseball fare under this new view of clutch:
Clutch NonClutch NAME G PA OPS G PA OPS DIFF RATIO --------------- ---- ---- ------ ---- ---- ------ ----- ----- Barry Bonds 86 373 1.363 44 177 1.100 .263 1.24 Albert Pujols 83 376 1.289 74 309 .895 .394 1.44 Todd Helton 73 334 1.194 87 369 .994 .200 1.20 Javy Lopez 80 321 1.198 49 174 .820 .378 1.46 Gary Sheffield 96 431 1.151 59 247 .813 .338 1.42 Carlos Delgado 85 378 1.201 76 327 .812 .389 1.48 Manny Ramirez 91 408 1.124 63 271 .858 .266 1.31 Vladimir Guerrero 61 263 1.053 50 204 .962 .091 1.09 Jim Edmonds 66 278 1.135 65 253 .856 .278 1.32 Alex Rodriguez 71 336 1.161 90 379 .855 .306 1.36 Trot Nixon 78 324 1.052 52 189 .842 .209 1.25 David Ortiz 73 309 1.151 55 200 .677 .473 1.70 Jim Thome 84 383 1.156 75 315 .725 .432 1.60 Richard Hidalgo 74 309 1.247 67 276 .629 .618 1.98 Frank Thomas 81 366 1.127 72 296 .737 .390 1.53 Matt Stairs 55 173 1.094 66 184 .811 .283 1.35 Brian Giles 62 291 .978 72 318 .909 .069 1.08 Jason Giambi 95 443 1.033 61 247 .774 .259 1.33 Bill Mueller 83 354 1.124 59 246 .667 .457 1.68 Luis Gonzalez 81 367 1.064 75 312 .786 .278 1.35 Ken Griffey Jr. 24 100 .912 29 101 .957 -.045 .95 Lance Berkman 83 361 1.081 70 297 .748 .333 1.45 Richie Sexson 68 312 1.056 94 406 .830 .225 1.27 Jorge Posada 87 374 1.055 55 214 .706 .349 1.49 Magglio Ordonez 85 375 1.105 75 299 .699 .406 1.58
An assessment of some of the top hitters last year by OPS reveals the comforting fact that most of them were clutch hitters to one degree or another. Barry Bonds‘ ratio of 1.24 indicates that he was 24% better in clutch situations, which while good, is 20 points behind Albert Pujols‘ 44% improvement. Perhaps awarding Bonds the MVP was premature.
Stathead darlings Brian Giles and Vladimir Guerrero both rate relatively poorly, with ratings below 1.10, making them less than 10% clutch, which probably explains why they’ve been toiling for small market clubs like the Expos, Pirates, Padres, and Angels.
The best overall hitter who was also a choker by this measure is Ken Griffey Jr., whose clutch performance was 5% below his performance in less important games. Though Griffey has a long history of being a good clutch performer (clutch ratios in the 1.20-1.50 range up through 2001), he’s been a choker in both 2002 and 2003 in limited playing time. Perhaps age and injuries are catching up to him.
The one truly stellar clutch performer on the list is perhaps a bit of a surprise: Richard Hidalgo, who improved his OPS 98% in clutch games. That ClutchRatio of 1.98 was sixth-best for players with at least 100 PA in clutch & non-clutch games last year. Maybe the Astros are closer to getting their money’s worth than many realize?
Top 20 ClutchRatios for 2003 NAME G PA OPS G PA OPS DIFF RATIO --------------- ---- ---- ------ ---- ---- ------ ----- ----- Paul Konerko 69 262 .949 65 233 .449 .500 2.11 Jose Reyes 29 130 1.078 40 162 .516 .562 2.09 Troy Glaus 46 188 1.078 45 179 .529 .549 2.04 Brian Schneider 53 202 .922 53 175 .455 .467 2.03 Greg Myers 60 215 1.105 60 154 .552 .552 2.00 Richard Hidalgo 74 309 1.247 67 276 .629 .618 1.98 Gary Bennett 45 181 .779 50 157 .396 .383 1.97 Travis Hafner 33 127 1.157 57 197 .594 .564 1.95 Jason LaRue 46 185 1.026 70 252 .535 .491 1.92 Karim Garcia 34 129 .955 38 133 .504 .450 1.89 Jeromy Burnitz 59 246 1.027 67 259 .563 .464 1.82 Bernie Williams 73 335 .932 45 186 .512 .420 1.82 Randall Simon 57 207 .973 67 224 .536 .437 1.82 Jeff Conine 77 338 1.018 72 308 .561 .456 1.81 Paul Bako 33 112 .809 33 101 .451 .358 1.79 Brad Wilkerson 70 303 1.081 75 299 .607 .474 1.78 Preston Wilson 69 308 1.146 86 353 .651 .495 1.76 Steve Finley 75 299 1.098 72 283 .625 .473 1.76 Damian Rolls 44 182 .874 59 222 .499 .374 1.75 Juan Uribe 38 168 .925 46 175 .529 .395 1.75
Forget the cliches–if you really want a player who gives 110%, Paul Konerko is your man. His improvement of 111% in the clutch truly demonstrates that while his season was a disappointment to many, he was still able to produce when his team won.
The rest of the list contains a variety of names both prominent and underappreciated. Bernie Williams and Steve Finley have had their share of success on teams with postseason successes. And Jose Reyes looks even better when you focus on the games he performed well in. Trusty veterans like Jeff Conine, Troy Glaus, and Jeromy Burnitz appear on the list as well. But how about the unsung clutch heroes, like the Clutch Catching Triumverate of Brian Schneider, Greg Myers, and Gary Bennett, who occupy three of the top seven spots on the list? Indeed, part-time catchers like Paul Bako and Jason LaRue occupy several spots on the list, their lack of playing time concealing their talent for hitting well in victorious efforts.
Looking back over the period we can compute ClutchRatio for (since 1972), the top season with 100+ PA in clutch/non-clutch situations was in 1977 by Tom Grieve, while with the Rangers. He followed a season in which he was in the top 10 in the AL in HR with splits of .304/.344/.514 in wins, and .112/.163/.122 in losses for an astonishing ClutchRatio of 3.00, or a 200% improvement in clutch games (wins). Grieve never came close to those levels again, (and indeed most players with a high ClutchRatio were unable to maintain that level of performance over several years), but his 1977 peak is impressive nonetheless.
If we’re willing to lower our standards just a little further in this discussion, we can find several even better examples of high ClutchRatios, headed by Russ Morman‘s 1986. As a rookie first baseman with the White Sox, he must have turned the manager’s head by hitting .431/.476/.667 in his 82 plate appearances in those clutch Chicago wins, versus .103/.194/.103 in 98 PA in non-clutch losses. His 3.84 ClutchRatio edges out Doug Flynn‘s 1975 (3.82) and Francisco Cabrera‘s 1991 (3.77) as the top clutch seasons with at least 50 PA in clutch/non-clutch games. Cabrera would validate his reputation as a true clutch player by delivering a game winning hit in the 1993 NL playoffs.
Baseball statheads have been working for decades trying to detect and identify clutch hitters, but now we can see what we want to see in the data. By recognizing that winning is clutch, and losing is choking, it becomes almost trivial to see that most players who reach the majors possess better stats in games their teams win. We’d expect this for a variety of reasons. The weeding-out process of high school, college, and minor league ball should eliminate those players who aren’t interested in producing when the game is in hand. This confirmation of replacement-level theory, and the way it dovetails with conventional wisdom about the existence of clutch hitters in general, may one day be seen as one of the crowning achievements in sabermetrics.
Thank you for reading
This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.
Subscribe now