Underestimating the Fog

 

A recent Bill James article caused a stir in the statistical community this year. In Underestimating the Fog, James essentially cautions us that the lack of evidence that a phenomenon exists (specifically clutch hitting) is NOT evidence that such a phenomenon does not exist. Before I move on, let me provide a link to the James' article.

 

Underestimating the Fog

 

Do clutch hitters exist? Given James' thoughts on the difficulties of studying the subject, I think real, definitive proof will lie somewhere next to the solution of the existance of God. We may never know. But there is an important aspect here that goes unmentioned in the James piece. If we can not identify if clutch hitting DOES exist, how could we possibly measure with any degree of confidence WHO the clutch hitters are, especially when evaluating a team's roster construction. It would seem that clutch ability could hardly help a team identify whether or not to sign a player or not. In my opinion (and just my opinion), it would seem that clutch ability would play a minisule role in a player's value in comparison to a player's measureable skills.

 

Second, the main issue in which we discuss clutch ability is in the past. In historical discussions, it is irrelevant to debate whether a player had the ability to hit in the clutch. What matters is what he DID in the clutch. James in the New Historical Baseball Abstract argued that players such as Don Drysdale should be downgraded based on clutch performance, while the Bob Gibsons should be upgraded. I agree, with one qualifier. Clutch performance should not necessarily be what we perceive, but something stronger. Reggie Jackson hit extremely well in the World Series, but it is forgotten that he also hit .227/.298/.380 in 163 ALCS at bats. Second, Derek Jeter. I will get into a deeper analysis of Jeter at some point, but when you look at his postseason at bats, he has hit almost exactly as well as he does in the regular season. He has had clutch hits, as you might expect if given 441 at bats. He has also had some poor postseason performances. Just observing a player's clutch ability is not the same as having solid evidence of postseason success.

 

James largley upholds the Voros McCracken theory regarding Defense Independant Pitching Statistics. Regardless of whether you subscribe to the hits = luck theory, this bit of information has proved extremely useful the last few years. Even if hits are not the result of luck, we see that home runs, walks and strikeouts are such strong indicators of pitching skill that they make hits completely unnecessary when evaluating a hitter.

 

The two conclusions that James dismisses due to lack of evidence are the theories that catchers have no influence on a pitcher's performance, and that hot and cold streaks do not exist. James cautions us that it is unwise to dismiss the conventional wisdom unless we have solid proof. In the case of catchers, the small number of innings pitched are not sufficient evidence.

 

On the whole, it is wise to review the whole of the evidence, and always look out for false indicators in our studies.

 

I also posted this at the Bobby Abreu Appreciation Society, and I hope to get that running full steam in the near future.

I always thought the concept of clutch hitting was a symptom of people paying attention to the good and not the bad.

 

People remember the 5 game winning HRs more than they remember the 50 game ending ground outs, so the idea of clutch hitting is born.

The problem with "clutch hitting" is sample size - we take a handful of at-bats in critical situations and extrapolate a lot of analysis that really doesn't have a leg to stand on.

 

That being said, I'm not necessarily sold on "clutch players" as much as I am "players that don't choke under pressure". Many times, you'll see a player like Derek Jeter who hits about the same in those clutch situations as he does in the regular season; some sabermetricians discard that stat and say "See? There is no clutch situations, Jeter hits exactly the same!", but what they fail to realize is what the average guy does in that situation. For example, there may be a player who hits 30 home runs throughout the regular season, but who batted .171 with an .671 OPS whenever there were runners in scoring position and two outs. Once again, you can't really weigh this too much, due to sample size, but I contend that you will see more significant dips than significant leaps in clutch situations, provided you accumulate enough at-bats.

Well, when you expand that to Sosa's entire career, that doesn't really prove too much. Although such a beast may exist. David Justice for example had a .878 career OPS, but hit horribly over the postseason (.224/.335/.382 in 398 postseason at bats).

 

It is difficult to tell because even the leaders in career postseason at bats add up to just over half a season's worth of playing time. And even the data the biggest samples give reveal surprises. Tito Martinez has a career .685 postseason OPS. Others who have seen performance drops include Jorge Posada and Kenny Lofton. None of the top ten in postseason at bats saw drastic performance increases.

 

However, there are several problems with this approach. The lack of sample size causes problems. In addition, there are other issues besides lack of clutch ability that could lead to performance drops. It could be that the pitchers are tougher in the postseason (since the bad teams are weeded out), or even a matter of fatigue. Even more difficult is the statistical principle that in a statistical curve, you will see outliers.

 

From my experience, the more at bats a player tends to collect in "clutch" appearances, the more his clutch performance tends to mirror his career performances.