Sunday, October 9, 2011


Let me start with the punchline:

If you've ever run a regression in your life, you should definitely see the movie Moneyball.

This is a sufficient but not necessary condition for liking it - it's a very good movie anyway. The combination of Michael Lewis for the original book and Aaron Sorkin for part of the script is a pretty damn compelling one.

It's the story of the Oakland Athletics, and how the general manager (Billy Beane, played by Brad Pitt) got a great team on a tiny budget by following the advice of Peter Brand (played by Jonah Hill), who uses a data-driven approach to identify undervalued players and better strategies to winning games.

The whole movie is like catnip for econ nerds. The main guy who shakes things up is this fat, awkward young guy who studied economics at Yale. And even better, the obligatory montage for any movie seeking to convey computer programming (closeups of lots of numbers, statistical looking outputs) featured code running in Stata! I was at this point thinking Ha ha, I run those regressions too! I could be that fat econ nerd. No wait - I am that fat econ nerd!

But the phrase 'catnip' above is deliberately chosen, in the sense that non-statistical people are likely to see it differently. The key dynamic is that Brand/Hill has the underlying regression-based strategy for winning games, and Beane/Pitt is the general manager who sees the promise in the idea, and implements the it against the wishes of the establishment. So which of the two is more crucial to the process? Both are are necessary, but which part you emphasise depends on how you view the world.

Here's how IMDB describes the movie:
The story of Oakland A's general manager Billy Beane's successful attempt to put together a baseball club on a budget by employing computer-generated analysis to draft his players.
In other words, to the non-statistical public, the role of Brand/Hill is limited to 'computer-generated analysis' -  this is a story about the general manager with the vision. The statistical guy is the wonk who runs the numbers in the background. Same as in finance: you need the quants to do the analysis, but you need the visionary portfolio manager to know when to implement it.

The stat nerd views it as follows: 'Screw that! If I could only pick one of them, the data guy knows the strategy to run, and without him the vision guy is toast. Ulysses Grant won the Civil War for the North, not Abraham Lincoln'.

The CEO type rejoins: 'Lots of people have visions for how to run baseball, and a large part of being a general manager is knowing who to listen to. If you can't get yourself into a position of authority to actually make the decisions, your strategy is useless. Lincoln had to go through seven generals before he found Grant.'

(In an ironic twist for the nerds, in real life Paul Podesta was the assistant GM, but he didn't want his name used, in part because he objected to being portrayed as a pure stats nerd. So he became 'Peter Brand'.)

And in fairness to Beane/Pitt's character, he does come to understand and embrace the strategy, and we're also shown that he understands the general problem even before coming across Brand/Hill. Specifically, how can teams with small resources and budgets hope to compete with the far better funded Yankees?

The answer is screamingly obvious to econ types - stop trying to buy the assets that everyone agrees are the best, for which you'll almost certainly overpay, and start buying the most underpriced assets that get you the same output. I've written about this before in the context of stocks:
In the language of the common man, you're better off buying a crappy but underpriced company than a solid but overpriced company.
As it turns out, this is as true in baseball as in any other business.

The tragedy for Oakland, of course, is that any strategy that exploits mispricing will be most effective when few other people are doing it. The best chance for this actually getting them a World Series was the first year it was tried. The more people think the strategy is successful, the more it gets copied, and the less of an advantage it brings you. True to form, the Oakland Athletics have still not won a World Series since the strategy was started..

Update: The one scene that the movie didn't show is the one where the strategy initially isn't working, and Brand/Hill is seen furiously re-checking his analyses late at night and muttering to himself, 'Oh $#**, I really hope I didn't make a coding error or run a badly mis-specified regression.' Because I guarantee you that that happened at some point.

No comments:

Post a Comment