Sunday, November 6, 2016

Who are the great clutch hitters of modern times?

Many of the finest baseball minds believe that clutch hitting doesn't really exist; that it is an illusion created by a confluence of chance and small sample sizes. Before I started to research the subject, I was completely in that camp, and was willing to go even a step further. I theorized that if clutch hitting did exist, I would be suspicious of it. If a player is capable of hitting better in clutch situations than he does at other times, why doesn't he just perform like that at those other times? Is he too lazy? Unmotivated? I was, however, willing to concede that clutch hitting might have an opposite - choking. If some ballplayers choke under pressure then, by definition, those who do not choke are better clutch hitters, in which case clutch hitting does exist, albeit as the absence of a negative, rather than the presence of a positive.

In order to evaluate all those hypotheses, it is necessary to have facts and a definition of clutch hitting. Perhaps someday a sabermetrician will dig into each player's career at-bat by at-bat, identifying every moment that represented a clutch situation, and assigning a weight to each of those moments by the degree of pressure, finally determining who the greatest clutch hitters are. For now, we have something tangible, simple and transparent that can stand as a surrogate for that missing information: performance in the late innings of close games. Not only are these moments reasonably identified as clutch situations, but all the data are readily available at baseball-reference.com, so any of you can easily verify any claims I make here.

The statistic does have some weaknesses:

1. Clutch data is not directly comparable to non-clutch data, because pitchers behave differently in the late innings of close games. Moreover, pitchers alter their behavior even more for great hitters like the ones studied here. If a pitcher is ahead or behind by ten runs, there is no special reason to pitch around the great Joe DiMaggio, but if the moundsman is protecting a one-run lead in the ninth inning, it's a whole 'nother kettle of crawdads. The most obvious result of this disparity is that great hitters draw a lot more semi-intentional walks in the late innings of close games. No professional hurler is going to allow DiMaggio to see one he can drive, if that can be avoided.

2. Adjustments are necessary in order to compare players from different eras. In the era before pitch counts, when starters were expected to finish the game, hitters in general performed better in the late innings than they did early in the game, probably because they were often facing a tiring starter. That has gradually but inexorably changed in the past fifty years, so that hitters in general now perform much worse in late/close situations than they do overall. Instead of facing a worn-down starter, they are usually facing a fresh, strong arm, occasionally one chosen specifically to face them because of the match-up history. While Ted Williams might have been facing a tired Bob Lemon, throwing his 150th pitch in the 9th inning and struggling to reach 85 MPH, Joey Votto in the same situation might be staring down the barrel of a fresh Aroldis Chapman bringing his first pitch of the day at 104, and typically striking out fifteen batters per nine innings.

Baseball is a tradition-bound game. Quite astoundingly, it took baseball's managers nearly a century before they even started to realize that it was not wise to let starting pitchers finish so many games. Even the very best and very toughest pitchers in history tended to flag as the game wore on. Let's take a couple of dramatic examples: no modern pitcher was better at completing a game than Warren Spahn (382 CG, tops by a mile among post-WW2 pitchers), and nobody was tougher than Bob Gibson (in his best years he completed 80% of his starts). Yet they both flagged in the late innings. Their strikeouts declined by about ten percent. Batters' OPS production increased by about five percent overall, slightly more in close games.

Opp. OPS
Innings
1-6
Innings
7-end
Innings
7-end (close)
Decline
in 7-end
Decline
in 7-end (close)
Bob Gibson
.613
.642
.654
4.7%
6.7%
Warren Spahn
.643
.676
.678
5.1%
5.4%






K/9
Innings
1-6
Innings
7-end

Decline
in 7-end

Bob Gibson
7.4
6.7

-9.4%

Warren Spahn
4.6
4.0

-15.0%



Pitchers work harder than batters, and thus tire sooner, but in the past they would often remain in the game, thus giving hitters an advantage as the game progressed. But those were the old days. Batters can no longer count on improving their performance in the late innings. As the game has evolved, starters get pulled earlier, and relievers become more specialized. Perhaps the only thing that pushes a starter to the end of the fifth inning today is the old-time tradition that the starter needs five innings of work to get the win, but even that could be challenged some day. As the analytical management style continues to dominate the game, baseball is loosening its grip on the traditions which are unsupported by empirical evidence. Sometime in the future, perhaps sooner than we expect, baseball franchises with nothing to lose will experiment with three-inning starters, or even with having nine men pitch an inning apiece. The first time such an experiment produces a dramatic improvement in the win column, old-time baseball's floodgates will open, and the modern world will come pouring through. For now, however, let's consider what has happened so far.

The drop-off in late inning performance has been quite dramatic over the years, and has not yet bottomed out, as shown in the following table. The table uses American League data only, and only since the inception of the DH, in order to remove pitchers from the equation. (They bat less frequently in the later innings, and thus pollute the data).

American League OPS, all players

Late and Close
Overall
Difference
1973-1978
.701
.706
-0.8%
1979-1987
.718
.731
-1.8%
1988-1997
.714
.743
-3.9%
1998-2014
.718
.755
-4.8%
2015-2016
.691
.737
-6.3%


The good news is that we know all of this, and can therefore establish a reasonable expectation of how any given hitter should perform in late/close situations, given his overall level of performance. The OPS of Ted Williams, for example, is expected to be about 1% better in late/close situations than his overall number, while Mike Trout's late/close OPS is expected to be about 6% worse than his overall career norm. These estimates vary based upon the actual years in which the player appeared, and while they are estimates, work very well in general. Making the appropriate adjustments, I estimated that my selected group of players would have a cumulative OPS of .901 in late/close situations, and they actually produced .899.

Why did I use OPS? It is the most easily and readily available surrogate for run production. As the pioneer sabermetrician Bill James has pointed out, OxS (on base times slugging) actually correlates more closely to runs than OPS, but OPS is available and transparent, so it's the best user-friendly metric. It's also much easier for all of us to use and manipulate. Most people who are at all interested in numbers can calculate that a .350 OBP plus a .550 SLG equals a .900 OPS, but not many people can multiply two three-digit numbers in their heads.

There is one remaining question, and it's an important one. Is clutch hitting properly defined by actual performance in late/close situations, or by the improvement (or lack of decline) in performance? That is more a matter of opinion than analysis, so I'm going to place both into a sortable table. You can decide which assumption suits you.

I think that's probably enough disclaimers and caveats. We should probably postpone any additional verbiage until you've had a first look at the results and a chance to tinker with them.

Here are a few notes necessary to read and interpret the data:

  1. The OPS+ in clutch situations is a crude estimate based on the OPS ratio. That is to say that if a player's OPS was 6% better than expected in clutch situations, I estimated that his clutch OPS+ was also 6% better than his overall OPS+. Why bother with this? It's necessary to convert OPS to OPS+ to make the data more meaningful, otherwise you would have no way of knowing that Dick Allen's clutch .925 is actually quite a bit better than Stan Musial's clutch .990!


  2. The far right column is the ratio of actual clutch performance to expected clutch performance, multiplied times 100. In other words, a score of 111 means the clutch performance is 11% better than expected.


  3. You will note that some of the items in the far right column are in red bold type. These are the only ones in which the difference between the distribution of hits in non-clutch and clutch situations is statistically significant at a 95% confidence level. In other words, if Mickey Mantle's non-clutch plate appearances were divided into random samples the size of his clutch appearances, his actual clutch performance would be above the 97.5th percentile of those samples, or would have less than a 5% chance of occurring in that population. Similarly, Ernie Banks' actual clutch performance would be below the 2.5th percentile. In other words, except for those six players with red ratios, you're just studying the table out of curiosity. It's nice to know that the player you love performed 4% better than expected, and the guy you hate fell 6% below, but none of those facts are statistically significant. That's like comparing a .305 hitter during a single season to a .295 hitter: one number is better cosmetically, but does not represent significant evidence of better hitting.


So what have we learned?



Well, first and most obviously, we have learned that it doesn't matter how you sort the data when it comes to the number one spot. It's Mickey Mantle. The Oklahoma Kid was the best hitter in clutch situations, and he was also the guy who improved the most from his normal performance. Here's how it breaks down:

Mickey Mantle per 550 at-bats

singles
2b+3b
HR
walks
avg
in clutch situations
98
33
46
140
.321
in non-clutch situations 99
28
35
114
.294


As you can see from the data, Mickey profited greatly from pitchers' fear. In non-clutch situations, the Commerce Comet walked a lot - 114 times for every 550 at-bats, but with the game on the line in the late innings, that number rose to a superhuman 140 because no pitcher in his right mind would intentionally give Mickey a pitch to hit. He also had the advantage of being a switch hitter, so opposing managers did not have many options when it came to relief pitchers. When he made contact, Mantle did not hit significantly more singles, doubles and triples, but his homer production increased more than 30%. Basically, he sat and waited, willing to take a walk if there was nothing in the zone, but swinging with all his considerable might if the pitcher released one he could take downtown. In those late inning pressure situations, The Mick transformed from a .294 hitter with a seasonal average of 35 homers and 114 walks into a .321 hitter with 46 homers and 140 walks!



The other three players who had significant increases in their performances with the game on the line are a more interesting group, in a way, because they may not be the players you expected. When I began this study I thought the great clutch hitters would be players like Reggie Jackson, Derek Jeter, and especially David Ortiz, who always seems to deliver in the big moments. It turns out that none of those players were especially good at stepping it up in the clutch. The other three great clutch hitters are Jackie Robinson, Roberto Clemente, and the Toy Cannon, Jimmy Wynn. Jackie Robinson was especially productive, actually tying Mantle for the highest percentage increase in critical situations. Two other players almost made the list: Tony Perez and Yogi Berra.

Note that these three men stand out based on their ability to improve in clutch situations, not their actual OPS+ in those instances. Ted Williams performed substantially worse with the game on the line than he did in other situations (it was almost a statistically significant difference, but not quite), while Jimmy Wynn performed substantially better, but a manager would still prefer to have Williams in the batters' box, because 94% of Ted Williams is still a lot better than 108% of Jimmy Wynn. In fact, 94% of Ted Williams is still better than anyone else in the clutch, except Mickey Mantle.

Clemente basically improved with singles. His power and walks improved a tiny bit, albeit on a small base, but his batting average improved from .313 in non-clutch situations to .341 in the clutch.

Jim Wynn, surprisingly, also improved with singles. His sum total of EBH plus walks actually declined a bit, but his batting average jumped from .244 in non-clutch situations to .279 in late/clutch appearances.

I am assuming that Clemente and Wynn were just smart, unflappable hitters who bore down and concentrated on meeting the ball in those late situations. Jackie Robinson, on the other hand, just plain got fierce:

Jackie Robinson per 550 at-bats

singles
2b+3b
HR
walks
avg
in clutch situations
128
36
24
97
.341
in non-clutch situations 117
37
14
81
.306


Like Mantle, his homers and walks improved in clutch appearances. Unlike Mantle, his singles production also increased. When the chips were down he turned from a .306 hitter with 14 homers and 81 walks into a .341 hitter with 24 homers and 97 walks. With the game on the line, he played like a man with something to prove.

Which he was.

I really don't want to dwell on the players at the bottom of the list, because I prefer to celebrate greatness, but I would be remiss in failing to see that Ernie Banks, whom I greatly admired as a player and a man, really fell apart in the clutch. On the surface, the smiling, affable, easy-going Banks seemed like the kind of man who would maintain his composure at all times and be unlikely to fold under pressure. Obviously, he had a more complicated psyche than fans realized, because he regressed into a very ordinary player in the late innings of close games. In the chart below you can see that he was roughly equivalent to Mickey Mantle in non-pressure situations (except for Mickey's walks, of course) but that there was a massive gulf between them with the game on the line in the late innings. Ernie's power production evaporated, while Mickey's exploded.

Per 550 non-clutch at-bats

1b
2b
3b
HR
BB
avg
Ernie Banks
92
25
6
31
42
.279
Mickey Mantle
99
23
5
35
114
.294







Per 550 clutch at-bats

1b
2b
3b
HR
BB
avg
Ernie Banks 93
20
2
22
60
.250
Mickey Mantle 98
26
7
46
140
.321


One more thing:

Amazingly enough, as I was studying this subject, I found that the greatest clutch hitter of modern times, at least in terms of improvement from non-clutch to clutch at bats, is somebody you may not remember at all. The reason he does not appear in the main table is that he had only about a thousand at-bats in his major league career, but he had an absolutely astounding improvement in late/close situations. The batter is Joe Lefebvre, who played for the Yankees, Padres and Phillies in the early 80s.



The table below says it all:

Joe Lefebvre per 550 at bats

1b
2b
3b
HR
BB
avg
in non-clutch situations
87
24
6
14
69
.238
in clutch situations 119
36
8
23
74
.338


Lefebvre once won a triple crown in the minors (Eastern League, 1979), and a lot of people must realize that he knows a thing or two about hitting, despite his mediocre major league stats, because he is still in the game to this day, at age 60, currently working on the staff of the major league Giants (image left), having spent some thirty years as a hitting instructor at various minor league levels, first with the Phillies' organization in the late 80s, then with the Yankees from 1990 to 1996. The years in those organizations are encapsulated in the images below. He finally landed in the Giants' camp in 1997, eventually becoming an assistant hitting coach for the big league club in 2013. As of 2016, he has left that position to perform various scouting duties for the Giants.



If Lefebvre had appeared in the main table, his ratio of actual clutch OPS to expected clutch OPS would have been 129 - although nobody actually on the table had a ratio higher than 111! That ratio was so high that the results were statistically significant despite a relatively small sample size. So hats off to Joe Lefebvre, the .238 hitter with moderate power who underwent a metamorphosis into a slugger with a .338 batting average and a .970 OPS when the game was on the line, making him arguably the greatest clutch hitter in modern baseball history.

Who knew?