Friday, May 20, 2011

Trying to make sense of it

As a follow up to yesterday's post, I ran the numbers for playoff, LCS and World Series teams against teams that miss that cut off for 1998-2010 (1998 chosen because it was the first year with full expansion.)

The pattern stayed the same - the playoff/non-playoff divide had playoff teams winning 56.53% of the time, while the LCS/non-LCS divide has LCS teams winning 56.00% of the time. After thinking about it, I'm a bit more comofortable with why this is happening.

It's going to seem like a simple explanation, but the reason this happened is because the non-LCS population of teams is better than the non-playoff population of teams. Playoff teams that lose in the divisional round have an overall win percentage of 54.93% in the following year, and have a winning percentage of 50.83% against LCS teams.

(I will note that if you compare LCS teams to non-playoff teams, LCS teams win 57.05% of the time, which is the phenomenom that I was expecting to see.)

I still have reservations about the testing method that Levitt used - baseball win percentages generally speaking range between 40% and 60%, so I don't know if a comparison of straight win percentages is the best method to show the comparison. But since there isn't really an equivalent win/loss for the random pairings that Levitt used, I don't know how to solve this. As well, there is a related argument to be made that MLB teams are already divided into varying ranges of elite talent teams, while the poker population tested contains talent levels that range from elite to poor. The argument basically is that if you put an MLB team up against, say, a beer league team, you would expect the MLB team to win 99% of the time, if not 100%. In the poker population, you already have the equivalent of MLB teams against beer league teams, so you would expect the random pairing win percentage to be higher.

I still wish to test the basketball playoffs, to see if there's something there. But I am glad to understand why cutting at a highter threshold ends up providing a lower win percentage. Sometimes it takes doing a bit more work and a second look to understand better what's going on.

Labels: , , , ,


Post a Comment

Links to this post:

Create a Link

<< Home