Thursday, September 04, 2003
Well, spinning up a simulation took a little longer than I thought it would. I tried to model a 14 team league, with two divisions of five teams and one of four. After a while I just took the 1996 NL schedule and used that as a template. I then ran 100,000 full seasons, treating each game as a 50-50 coin flip. This filled a win matrix and then this was sorted by division. The runners up in each division were compared and the team with the most wins was awarded the wildcard.
After 100,000 runs, the mean number of wins for the wildcard team was 87.1, more or less one standard deviation. I'm a little surprised that complicated situation like that comes up so close to the standard deviation.
OK, so I'm good with the wildcard now, historically those have been good teams.
posted by David 9:47 PM
