[ Post a Response | Puzzles Forum ]
Posted by markr on 8/8/2009, 4:55 am, in reply to "Fairer Method"
If the sets of opponents are disjoint, then comparing sums of opponents' scores isn't necessarily fair. Consider a group with strong offenses and weak defenses (high scoring) and a group with weak offenses and strong defenses (low scoring).
My suggestion would be to mix up play as best as possible (don't form round-robin type groups) and use an ELO application to rank your players. I think this approach takes a more global view of the data instead of looking at specific individual cases. I've used such an app to rank players in a squash league. It works rather well unless you have disjoint sets (two or more sets of players who only played games within their sets).
I used the program (free) from this site:
http://remi.coulom.free.fr/Bayesian-Elo/
It's quite easy, but it's set up for chess; so you have to use a particular format for the data. I'd be happy to share a sample data file I used for squash if you're interested.
Message Thread:
« Back to thread