2003 MLS RPI

[Foosinho]

Some of you may (or may not) know that I used to run a computer ranking of pro soccer teams in the US a few years back. I quit due to the time required to maintain the database and generate the standings, but I'm thinking about getting back in the game since I've got my own webserver now.

At any rate, since it's nearly Selection Sunday for the NCAA tournament, I got to thinking about the RPI, and decided to run one on the 2003 MLS season. This may be slightly off - I did this by hand in about 20 minutes. But here it is, sorted by RPI score.


W-T-L Points RPI SOS SOS Rank
Fire 15-8-7 53 0.5306 0.4965 8
Quakes 14-9-7 51 0.5193 0.4867 10
Revs 12-9-9 47 0.5116 0.4988 6
Metros 11-9-10 40 0.5040 0.4996 5
Wizards 11-9-10 42 0.5000 0.4944 9
United 10-9-11 39 0.5000 0.5056 3
Crew 10-8-12 38 0.4965 0.5063 2
Rapids 11-7-12 40 0.4948 0.4988 7
Galaxy 9-9-12 36 0.4884 0.5012 4
Burn 6-5-19 23 0.4548 0.5121 1


Not too surprising that Eastern Conference teams got a boost - they didn't have to play the Burn as often! I'm not entirely sure of the utility of this - but I thought it might be interesting. Thoughts?

[mellon002]

I'm always interested in the RPI rankings in college basketball. How did you make those calculations?

[Foosinho]

My understanding of RPI is that it's 25% your win percentage, 50% your opponent's average win %, and 25% your opponent's opponent's average win %. I think it goes two deep to help "connect" teams. It seems like it'd be less useful in such a connected league as MLS, but it does provide some accounting for schedule strength.

[Kevin in Louisiana]

I think it's almost entirely useless. The better you are the worse your strength of schedule is. I think RPI is useful for college sports, where you control your own schedule, but in the pro ranks it loses its value.

[mellon002]

I wonder how it would look if you took out the games head to head so that your RPI isn't punished for beating a team.