Sabermetrics: What is clutch? Who is clutch? Is it real?

The Rockets' championship teams gave Houston the nickname "Clutch City" in the 90's. That was a play on the Houston Chronicle headline "Choke City," which accompanied some earlier home court losses in the playoffs. So goes the narrative of heroic clutch players vs. chokers.

Sabermetricians are skeptical of narrative driven concepts and attempts to separate the mythology from the reality. The existence of clutch skills among pitchers and hitters in baseball is controversial, even within the sabermetric community. Although opinions may not be unanimous, I think it's fair to say that sabermetrics places less emphasis on the predictability of clutch skills than traditionalists. This article will look at the concept of clutch performance from a sabermetric perspective, and then show how the Astros have started the season from a clutch perspective.

What is Clutch?

Clutch usually is defined as the extent that players change their performance, for better or worse, in high leverage situations. High leverage situations are the most critical moments in games, based on number of runners on base, the score, the number of outs, and the inning of the game.

How is it measured?

There is not universal agreement on the best way to measure clutch performance, which may be one reason that disagreement arises related to the research of this topic. There are numerous types of Runner in Scoring Position statistics, as well as "close and late" statistics. Sometimes the various RISP stats point in different directions for a particular player.

I prefer the fangraphs Clutch statistic, which is part of the Win Probability Added (WPA) family of statistics; the Clutch statistic can be found under Win Probability on the player page. The clutch stat compares the player's WPA in high leverage situations to the player's performance without any weighting for leverage. Keep in mind that this metric compares a player to himself. It doesn't necessarily tell you whether a player is better or worse in clutch situations than another player. For example, Bobby Abreu and Darwin Barney are among the highest ranked players on this metric in 2011. Their clutch ranking is much better than Albert Pujols (who has a negative clutch metric on a career basis, by the way). But that doesn't mean that they are better hitters than Pujols in clutch situations; indeed, most people would want Pujols rather than Abreu or Barney to bat in high leverage situations. Abreu and Barney would have to raise their normal performance by a lot, in clutch situations, to even come close to Pujols' normal performance.

Does Clutch Exist?

No one questions that clutch performances exist. Most MVPs for a league or World Series involve clutch performances. The controversial question is whether clutch performance is a repeatable skill. Clutch performances change the distribution of hits, walks, and outs, thereby allowing players to outplay their underlying average statistics and teams to over or under perform their WAR and/or Pythagorean W/L record. So, repeatability versus randomness would be something we want to know. Some teams make important decisions, from batting orders to player acquisitions, based upon clutch performance, and those decisions may not be consistent with sabermetric findings.

A substantial body of sabermetric research concludes that clutch hitting is not a perceptibly repeatable skill. (For some reason, clutch hitting has been studied more than clutch pitching.) Some more recent studies have found that clutch hitting skill can be identified, but that either the effect is small or the number of players who have repeatable clutch skill is very small (less than 2%). "The Book: Playing the Percentages in Baseball" concludes that clutch hitting skill exists but that the impact is very small and not statistically significant. The existing studies have illuminated the issue, but it remains unsettled. Some sabermetricians, such as Bill James, still consider clutch skill to be an open question, because of the data and methodology obstacles. Sample size is always a problem for clutch hitting studies; as Tango's blog points out, roughly a level of 7,600 clutch plate appearances is required for normalizing the results, meaning that it is practically impossible to predict a given player's clutch hitting.

Tango's blog suggests that clutch hitting has a very small impact, approximately half the effect of typical left-right platoon impact, and should not affect team decisions. Interestingly, baseball management decisions may have affected clutch hitting performance, but not by identifying clutch hitters. There is some evidence that the increased use and specialization of relievers has reduced clutch hitting performance over the years. The Cybermetric blog has many interesting articles on clutch hitting. Among the study findings there that caught my attention: experienced hitters may have better clutch results, and power hitters may tend to be less clutch than other hitters.

My take on this: clutch hitting or pitching skill may exist, but if you see extreme clutch statistics, either positive or negative, expect regression toward the player's normal capability. We consider Astros on the next page.

Astros' Hitters With Clutch Reputation

Carlos Lee is the current Astros' hitter with a reputation for clutch offense. He is the only current Astros' hitter who appears in the Top 10 clutch statistic for the previous four year period. You may be surprised to know that two former Astros from that period also appear in the top 10: Geoff Blum and Pedro Feliz. Despite Lee's high ranking, 2011 was a negative clutch year for him. Although Lee has a cumulative net positive clutch metric for his career, 7 of his 14 years are negative, providing an illustration of our inability to predict clutch performance.

Astros Hitters in 2012

Based on the game thread chatter, Astros' fans complain about the Astros' lack of clutch hitting. If one just looks at the RISP stat, this complaint seems misplaced. The Astros have the fourth highest batting average with runners in scoring position in the majors, .302.. But the fangraphs' clutch statistic is more consistent with the fans' complaints. The Astros' team fangraphs clutch ranking is 27th, ahead of only the A's and Royals. First, it's clear that a statistic which more accurately reflects leverage. like fangraphs' clutch, is more consistent with the impact of clutch hitting on critical situations. Second, if you've been paying attention, you know that we should expect the Astros' clutch hitting to regress, which may mean that more hits should start falling in during critical situations in the future. As you know, the 2012 stats are subject to huge sample size issues; so the clutch stats below (excludes pitchers batting) are shown only for amusement.

Lee + 0.42

Bogey +0.21

Schafer +0.14

Bixler -0.04

Snyder -0.04

Gonzalez -0.09

Martinez -0.13

Maxwell -0.16

Lowrie -0.18

Buck -0.19

Downs -0.28

Castro -0.30

Johnson -0.32

Altuve -0.53

Does it surprise you that Altuve has been the least clutch, and Lee has been the most clutch?

Astros Pitching

Pitchers probably have a greater ability than hitters to exhibit clutch skill. Pitchers can cruise along until they reach tough situations, and then raise their velocity or bear down in throwing better breaking pitches. On the other hand, other factors, like pitching from the stretch and over-throwing may have a negative impact on pitchers' clutch performances. Wth the same amusement caveat, above, here are the pitchers' clutch stats:

Harrell +0.14

Happ +0.12

Weiland +0.11

Abad +0.10

Myers +0.01

Norris -0.05

F-Rod -0.07

Wright -0.13

Cruz -0.23

Lopez -0.33

Carpenter -0.34

Wandy -0.37

Lyon -0.83

Noticing the negative result for Wandy, I'm reminded that pitchers don't accumulate positive clutch scores if they are never threatened in a game. In Wandy's case, his best game was last Sunday, and the outcome of the game was never in doubt.

As we've mentioned previously here, the shutdown and meltdown stats are useful metrics for relief pitchers. These results also may tell us something about clutch relief pitching, since they are associated with WPA performance. Myers and F-Rod lead in shutdowns with 2 each. Lopez and Cruz have 1 each. F-Rod leads in meltdowns with 3. Lopez and Wright have 2 meltdowns. Cruz, Abad, and Lyon have 1 meltdown each.

Any surprises here?