Levitt on Beane Again

Steven Levitt has started blogging over at Freakonomics. Interestingly, one of the first things he’s blogged about is a statement that he made about Moneyball. About a year ago Levitt wrote, “If you look at all the stats they say are so important, the A’s are totally average! There’s very little evidence Billy Beane [the club's general manager] is doing something right.” (Here were my thoughts on the his statement at the time.) In this recent post he further explains what he meant.

The simple point I was making is that the A’s don’t win for the reasons Moneyball implied, i.e that they put a bunch of misfit looking characters on the field who have out of this world on base percentages which somehow leads them to manufacture runs without paying high salaries. The reason the A’s win, year after year, is because they have better pitchers than anyone else.

And in fact, Levitt has a good point. For all the stressing of the importance of OBP in the book, the A’s just were not so hot in this area. I think it’s clear that Moneyball was about a lot more than OBP — see the chapter on Chad Bradford and the discussion of defense — but few people have made Levitt’s point, and Lewis really did hype the OBP stuff. And this paper (by Hakes and Sauer) seems to do a good job showing that the OBP market inefficiency the A’s were exploiting at the time Michael Lewis was writing has evaporated and is now gone. So, the fact that the A’s haven’t been following their own advise on offense, doesn’t really disprove much. OBP is now pricey, so the A’s are looking for and finding new inefficiencies.

Levitt views Beane’s latest business venture as Billy seeing the truth (as Levitt sees it) about his past and the future. Just as the A’s traded away two-thirds of the big three, Beane is acting as though he expects something bad to happen.

Just as the A’s are about to head south, he negotiates a lucrative contract extension and becomes the first baseball GM to get an ownership share (but he doesn’t have any liability for losses, only sharing in the gains!).

That is genius.

I think Beane & Co. are still on board with trying to win more with less, just like everyone else. This makes them a little less special than stat-heads would like to think, but I still think they have some unique methods for doing so that work better than other teams. Such methods led them to draft the big-3 in the first place. Also, I’m not surprised Beane would prefer an ownership share with little risk, and I don’t think that’s much evidence for Beane seeing the end is near. I have fire insurance, but I don’t think my house is going to burn down. Billy Beane is just a typical risk averse investor. Sure, he could have probably gotten an additional share of the team’s success if he risked a little more, but the returns to the success may not be enough to offset the potential losses. He values the potential financial loss of what he has more than the potential gains from the acquisition than if he succeeds. What if Chavez, Zito, and Harden all suffer unforeseen career-ending injuries? Beane would take quite a hit. I’m sure many individuals, even if they expect nothing to go wrong, would also opt in for lower stakes.

But let’s get back to the big-3. Looking at the SPSS, let’s look at the predicted ERAs of the A’s s top four starters from last year.

Name    2004    2005
Harden 3.99 4.15
Hudson 3.53 4.51
Mulder 4.43 4.35
Zito 4.48 4.14

Hudson and Mulder have the highest projected ERAs of the bunch for 2005. Zito, who had the highest 2004 ERA is projected to have the lowest ERA in 2005, with Harden predicted to do about the same. The lowest 2004 ERA belonged to Hudson (oh no!). So, Billy holds on the better pitchers and trades the ones who are predicted to do worse. Now, it my be (in fact it’s quite likely) that by Beane’s projections are different from mine, but it is certainly worth noting that Beane is jettisoning guys who performed worse via ERA projections based heavily on DIPS.

Addendum: Steven Levitt responds to his critics and throws down the gauntlet.

In the spirit of data, the skeptics amongst you should tell me how many games the A’s need to win this year or over the next five years so that they would feel that Moneyball is validated. My own view is that if the A’s win 81 games a year for the next five years, it is more likely that Beane was lucky than good. If they win 97 a year, I’ll happily concede that Beane is the best. Even an average of 90 games
a year and I will acknowledge he is brilliant.

I agree. There needs to be standard, and he’s put forth a very simple and fair one. It’s easy to support Billy now, but maybe not in a few years. I think Beane will succeed, but that’s just my opinion. If he doesn’t do it, I’ll gladly concede that I was wrong. Levitt also points out that the “poor” A’s aren’t so poor.

Oakland used to be a low budget team ($30 million in 2000),
but not anymore. Their payroll in 2004 was above the median
in the American League….So let’s not feel so sorry for them in their pursuit of 82 wins this year.

But this is the best part.

And for all the talk about me not knowing anything about sabermetrics
(I actually do know quite a bit…I will debunk some sabermetric ideas in later
posts if people are interested)

I’m dying to see what he has to say. I think the biggest weakness of the sabermetric community is that it has formed its own set of core “beliefs” that are backed up by poorly done empirical work. I’m not going to point fingers, but I have seen a lot of stuff that is just done wrong. A true system of peer review is very much needed in the sabermetric community. Hopefully, Levitt can shake things up a bit. I think one of the weakest areas of understanding among sabermetricians is aging. Just playing around with the data, it seems to me that players do seem to peak a little later than the gold-standard of 27. I’m not saying this is true, but I certainly have a strong feeling about it. Maybe I need to get back on that.

One Response “Levitt on Beane Again”

  1. Marc Schneider says:

    It’s a very interesting post. I am not competent to discuss the methodology behind sabermetrics, but it always seemed to me that it has become sort of a catechism; throwing out numbers without any real support and then attacking those that dispute the numbers.

    Numbers are an important part of baseball, but baseball players are not simply robots. Human psychology is at play and it seems to me that a lot of the sabermetricians miss this by trying to reduce everything that happens on a baseball field to a statistical formula. That’s not to say that statistical analysis is not valuable and the debunking of a lot of the silly traditional analysis has not been useful, but, after all, it’s still a game. Sabermetrics, perhaps predictably, has become a dogma and like all dogmas may have become blinded to its own limitations.