THE BOOK--Playing The Percentages In Baseball

I agree with your point that if you pick out Angel in advance for his reputation, that isn’t cherry picking.

However, I’d argue that you have to pick out Angel in advance, not just for being a poor umpire, but for being a *racially-biased* umpire.  Did he have a reputation for that, in particular? 

Or, if he had a reputation for being a poor umpire in general, but he was a HUGE outlier on the racial-bias list, there might be something there, too; you could hypothesize that he’s just so I-don’t-give-a-crap that anything is possible.

But his observed racial discrepancy is pretty moderate, not out of line with the other umps.  So, in my opinion, it’s too big a leap from “bad umpire” to “racially biased umpire” without more corroborating evidence.

I am not suggesting that he is racially-biased, anymore than I think CC’s belly is strikeout-biased or David Wells’ belly is walk-biased.

I am suggesting that the identity of the player or umpire is overwhelming the effect of the belly or race.  And these studies, rather than attribute the effect to the belly or race, should instead attribute the effect to the persons themselves, intrinsic to them as unique as fingerprints.

CC and David Wells happen to share the same belly size, and Angel and whoever happen to share the same Hispanicness.  To therefore attribute belly size or race to something that is based on such a tiny number of persons in that universe of belly or race, I find, is unsupportable.

Angel is a bad umpire.  And it is possible that bad umpires are racially-biased.  Angel happens to be a bad Hispanic umpire.  If 50% of your Hispanic population is a bad umpire, that looks bad for Hispanics.

I agree completely that the effect should be seen as affecting only certain umpires.  But I’m saying that the fact that Angel is a “bad” umpire should not also be taken as evidence that he is also a racially-biased umpire. 

As I wrote, there IS statistically-significant evidence of bias.  And because there are so few minority umpires, the significance could be caused by even only one of them being biased. 

And Angel is as likely a candidate as any, but I’m not sure I want to argue that because Angel is problematic in other areas, he’s more likely to also be race-biased. 

To me, racism is a much more serious charge than incompetence, and I hate to connect the two.

It isn’t cherry picking.  It’s just good methodology.  Groups with an N=2 are going to be unstable, so a little individual examination of the two is a good idea.  Imagine if we based our conclusions about all Americans’ ability to hit a fastball and just happened to base them on me and Albert Pujols.

Phil/3: I’m suggesting that you don’t have race bias at all.  What we have is bad umpiring.  Just like we don’t have belly-bias at all.  What we have is great control pitching.

That we happen to notice that the two fat guys happen to be great control pitchers doesn’t have any causative link.

That we happen to notice that the two hispanic guys happen to have a different than average effect on race doesn’t mean we have any causative link.  It’s just pure coincidence.

Right, as Pizza said, creating a universe where the two actors are Pujols and anyone else will automatically confer a bias to Pujols’ partner that he simply does not earn, simply because we were able to create that universe.  Say, anyone born on Nov 11, 1979 (or whatever Pujols bday is) as showing bias, simply because we have two players in that set, and one is Pujols.

But bad umpiring isn’t the same thing as race-biased umpiring.  If Angel is just a bad umpire, why isn’t he a bad umpire for all races of pitchers equally?  What I am disputing is the “bad umpire ----> more likely to show race bias” connection.

And, absolutely, I agree with you that it could be just coincidence that the two hispanic umpires both happen, by chance, to have seen more strikes from hispanic pitchers.  Absolutely.  However, having said that, the coincidence is still statistically significant!  And, of course, just because something is statistically significant doesn’t mean it can’t be random—for instance, the November 11, 1979 thingy.  But it’s something of a copout to say, “well, it’s statistically significant, but I think it’s just coincidence anyway.”

The fact that it’s just two hispanic umpires is indeed relevant, but that is taken into account by the regression, and it STILL winds up statistically significant.  And so you can’t just point to the fact that there’s only two umpires to argue for coincidence.  But you CAN point out that it might just be one biased umpire.  Which I think is quite possible.  I also think it’s quite possible that it might be Angel.  But I don’t think there is enough evidence to accuse Angel, or any individual umpire.

Here are four possibilities:

1.  Moderate same-race bias among 1-2 umpires
2.  Mild same-race bias among more than 1-2 umpires
3.  Coincidence
4.  Other explanations

Honestly, if I had to pick probabilities for each of these, just as a guess, I’d probably say 25% each. 

For #1, you could argue that it’s not really racism—maybe one hispanic umpire is friendlier with hispanic pitchers, and he’s just helping out his buddies.  Or maybe one black ump secretly roots for a certain black starter.  Whatever.  I mean to include all those in #1.

And, Pizza, I agree that individual examination of the two is a good idea.  And if you examine them, you find one is 6th highest in MLB in hispanic strikes, and the other is 20th or something.  Far, far from a smoking gun.

Right, I am suggesting bad umpire may be more racially-biased inclined.

At the very least, of those academics studying the issue, you are the only one who is even considering the possibility!  And since Pizza is our resident study experts, he immediately noted that if you only have two peas in your pod, an individual examination is a good idea.  Why then don’t the academicians do that?

Imagine if Angel is akin to Gretzky, Jordan, or Ruth, but you didn’t select based on talent, just, selected on whatever, born in Brantford, traded between Redox/Yanks, played baseball in middle of career, then you would point to an enormous bias in basketball players who played baseball at age 30.

The “whatever” is Hispanic umpires.

Slate references Phil, and a link to this blog, from which I will complete the circle:
http://www.slate.com/id/2193031/

Phil summarizes it all for us:
http://sabermetricresearch.blogspot.com/2008/06/hamermesh-umpirerace-study-revisited.html

This is the kind of article that needs to be published at Hardball Times… numbers kept to a minimum, and simply recapping his more involved posts.