March 21, 2006

March Madness -- Maybe it's not so mad after all

The press is full of its annual men-bite-dogs accounts of all the upsets in the NCAA college basketball tournament, but when you look at the big picture over the last 20 years, you can see that the regular-season rich generally get richer in the post-season. And that has implications for how to bet in office pools if you don't know anything about college basketball.

(Some background: The NCAA divides the 64 teams qualifying for the tournament into four regions of 16 teams apiece. In each region, colleges are seeded from #1 to #16. In the first round, #1 plays #16, while #2 plays #15, and so forth. In the second round, the winner of #1 versus #16 (which 88 times out of 88 games has been #1) plays the winner of #8 versus #9, and so on. Losers go home. Eventually, the four winners of each Regional meet in the fifth round (semifinals), and the two winners play for the championship in the 6th round.)

The graph above shows that the actual results come out about as you'd expect from the seedings. There are anomalies -- for example, #6-seeded teams have outperformed #5-seeded teams, but generally the curve is pretty smooth. This is partly a tribute to the good work down by the NCAA seeding committee, and partly an inevitable outcome of the bracket. For example, the #1 teams are almost guaranteed at least one victory because #16 teams are so bad -- they are generally teams that are in the tournament solely because the NCAA has an obligation to take the winner of their obscure conference, no matter how awful the winner is. In contrast, #15s are often mildly competitive (although they are more likely to scare #2s than beat them), while #14s are usually pretty decent teams that often have a fighting chance against #3s.

So, next year when you get pressured to enter the office March Madness pool even though you don't know anything about basketball, you can be assured of a respectable showing if you just pick the higher seeded team in each game. (When you get to the final four, then pick among your four #1s based on the end-of-the-season AP poll rankings.) If the pool gives one point for each victory, you'll do quite well and will impress your co-workers with your quiet basketball expertise. Unfortunately, most pools give much heavier weights to later games, so this strategy doesn't work as well, but you'll still do better than average (on average).

In other words, the seedings are a fairly efficient market. And even if you can find mistakes in the seedings, well, the bracket tends to cover up the mistakes. Like doctors, the NCAA seeding committee often buries its mistakes. Say you know that team X shouldn't have gotten only a #13 seed, that it's really good enough to be a #9 seed. If it had been seeded #9, it would have been matched in the first round with a #8 seed, and it would have had close to a 50% chance of winning the game. But because it was unfairly relegated to #13, it gets stuck playing a #4 seed in the first round, and most likely loses even if it plays well enough to beat a #8 seed.

My published articles are archived at -- Steve Sailer

No comments: