What's the "best" algorithm to pair teams in TrueSkill given only the mu and sigma values of players in each team? I'll be working with even teams but would prefer a method that works for all teams. Right now I'm summing the (mu - 3sigma) ratings but I want something that's more likely to generate even matches.

Edited 8/7/2015 03:40:13

The best way depends on what you want to achieve. The mu's give the skill level, so you'd want to look at the mu's first. However, if your goal is to give people with few games (high sigma) lower ranked opponents first, then you should consider mu-3*sigma (rating). Something in between could be mu-sigma.

Additionally, if you want new players to get an accurate rating quickly, you want to match players with high sigma's with players with low sigma's. This way, the sigma of the new player will decrease quickly (but it isn't so nice for the player with the low sigma necessary).

Edited 8/8/2015 14:49:51