Saturday, December 1, 2007

Bill James must have been watching waaaay too much Sportscenter...

Because there is no other explanation for this piece of trash he penned for Sports Illustrated.

As most people know, Bill James is the legendary Baseball Statistician who helped revolutionize the way fans, teams and (some, but not many) journalists look at the game of baseball. So what in the world is James doing writing a statistically un-sound article on the merits of clutch hitting?

I'll be the first one to admit it: I've been watching waaaaaaay too much ESPN lately. It's mainly because I got this Sick-Ass HDTV. I keep ESPN on basically 24/7 because it's the only sports channel in High Def. Although I love to watch me some sweet ass sports highlights in HD, I've felt stupider with every passing week since I got my new TV. I find myself arguing with perfect strangers at the grocery store about how Tom Brady is clearly the greatest QB ever and that A-Rod isn't worth the money because he hasn't won a World Series. The way I've been acting is the reason I believe that Bill James has been glued to the couch 24/7, in his underwear with a bag of Funyuns rested on his belly, watching ESPN. Because honestly, I never, ever EVER thought I'd see Bill James sucking off David Ortiz for being "clutch." And that's even taking into consideration that he works for the Red Sox.

Back in the early days of sabermetrics, when dinosaurs roamed the American League Western Division, we made a very fundamental mistake. A friend of mine wrote an article asserting, essentially, that clutch hitters don't exist. At the time, we lacked any real ability to study the issue. We didn't have access to play by play of the games. No one could plausibly assert that clutch hitting did exist, because we couldn't document it without access to the game accounts, but Dick Cramer had finagled access to a couple of seasons of old data, studied the data and concluded that it didn't. There was nowhere for the discussion to go.

First of all, I personally don't buy into the fact that baseball players are robots who will always hit in a manner that matches their overall statistical profile. Guys get streaky. Sometimes is seems like they know what's coming and can always get contact and usually a hit. Other times players can psyche themselves out into swinging at bad pitches. I'll even buy into the fact that some players hit worse under lots of pressure. After all they are humans. But I will never, ever, ever believe that there are players who unequivocally hit BETTER when under more strenuous situations. It simply doesn't make sense.
It was about seven years after that before we began to have access to play by play, long before the data began to come on line, the discussion had stalled out at the assertion that clutch hitting did not exist.

The discussion has been premised upon an assertion, rather than flowing from the question itself. What I have been trying to do for the last couple of years is to back up, define a clutch situation, begin accumulating data, and gradually go down the other path.

Some people find this confusing. "Why are you publishing this clutch data," they will ask, "when you don't have any reason to believe that there is such a thing as a clutch hitter?" But that's the thing: We're publishing the data because we don't know.

This has all been done before. People have tried to shit all over A-Rod because supposedly he's the kind of guy who hits home runs exclusively when the Yankees are up 10-2. Bill Simmons did the same thing with JD Drew this year. He talked about how un-clutch and terrible Drew was, despite Drew's relatively productive season. I mean it was no mystery that Drew underperformed based on his career numbers, but Simmons made him out to be the devil because "he always grounded into double plays when we needed a hit." Larry, on this very Blog, pointed out Simmons' terrible reasoning. For example, Drew had terrible numbers in the batting with the bases loaded statistic. But he performed close to his season averages in the "close and late" category.

The problems with typically used "clutch" hitting metrics is going to be the same problems you'll have with any new clutch hitting metric that comes out. They depend on subjective paremeters and small sample sizes.

The other question everybody asks now is "How do you determine what is a clutch at-bat?" I'll have to stiff you on that one for right now. I'll explain it generally and leave the details for some other time.

"Clutch" is a complicated concept, containing at least seven elements:
1. The score
2. The runners on base
3. The outs
4. The inning
5. The opposition
6. The standings
7. The calendar.

Now what the hell is going on here? James seems to be implying either that

1) readers of SI are too stupid to understand the criteria for "clutch hitting"

or, the more likely scenario

2) James cherry picked his own criteria for clutch hitting that amounts to nothing much more than the notoriously flawed "close and late" batting stat.

Sometimes people look at things like batting average with runners in scoring position, batting average with runners in scoring position and two out, batting average in the late innings of close games. Those things are all interesting, but Tampa Bay playing Texas in April is not the same as San Diego playing Los Angeles in September.

Yeah tell that to the 2007 Brewers or Marlins. Both teams, out of playoff contention, relished the role of spoiler and pulled out all the stops to beat teams like the Padres and Mets, respectively. What about players on non-contending teams who feel pressured to bat well at the end of the season to get that extra few million on their contract extensions? Are these players also not under pressure?

We made up a system giving weight to each of these seven factors; not saying it's perfect, but you have to start somewhere. Baseball's most famous clutch hitter is David Ortiz, so let's start with him. The Big Papi's batting record in clutch Situations, over the last six years.

This is the first "proof" by James that Clutch hitting exists.

Let's see here, lemme pull up some Ortiz stats....


Clutch Hitting: 50 AB, 1.025 OPS
Regular Season: 412 AB, .893 OPS

Clutch Hitting: 75 AB, 1.016 OPS
Regular Season: 448 AB, .961 OPS

Clutch Hitting: 62 AB, 1.139 OPS
Regular Season: 582 AB, .983 OPS

Clutch Hitting: 72 AB, 1.172 OPS
Regular Season: 601 AB, 1.001 OPS

Clutch Hitting: 46 AB, 1.170 OPS
Regular Season: 558 AB, 1.049 OPS

Clutch Hitting: 89 AB, 1.042 OPS
Regular Season: 549 AB, 1.066 OPS

Not only is Ortiz super duper clutch, but he's also really damn good at hitting ALL OF THE TIME.

So it appears as if Ortiz has been, on average, about 100 points higher in OPS over the past 6 seasons in so-called Clutch situations. There are, however, several problems here. First off, James doesn't reveal what exactly defines "clutch." He lists what criteria he uses to define "clutch," but let's be honest here: it's a completely subjective term with subjective parameters. And what good will this new stat be to us? How can you possibly define clutch so that it gives us a good idea of added value in terms of extra wins? The idea of Sabermetrics is to take as big of a sample size as possible, in order to eliminate natural fluxuations in the wonky, sometimes random, game of baseball. Then we can turn those samples and correlate it to number of runs, and therefore wins, a player contributes. This new clutch statistic seems to serve no real purpous besides patting James' best buddy Ortiz on the back for being SO FREAKING AMAZING.

Furthermore, it's almost as if James has never heard of the statistical phrase "sample size." I don't have a direct link, but I remember reading something James wrote a few years ago. Basically, he said that there have been players who have had massively productive seasons (like a 200 point increase in OPS) that were primarily a product of luck. He cited the usual suspects, like an unusually high BABIP, as reasons for the spike in performance. Fine, whatever, that's fair and probably a good argument. But the problem here is that James' explanation of "luck" was used on players who had like 650 PAs over the course of the season. But now, James expects us to believe that Ortiz is a pure clutch badass based on 450 PAs, cherry-picked by an unknown formula, over the course of six seasons.

James goes on to list some other players, including Juan Pierre, that preliminarily show that "clutch DOES matter." Frankly, I don't buy it. James has done a lot of important things for baseball statistics, but this one is highly suspect right from the get-go.

No comments: