Improve win predictions for 1v1 teams
antl3x opened this issue · comments
First of all, congrats and thanks for the great repo!
In a scenario that Player A has 2x the rating of Player B, the predicted win probability is 60% vs 40%. This seems strange.
players = [ [Rating(50)], [Rating(25)] ]
predict_win(teams=players)
[ 1 ]: [0.6002914159316424, 0.39970858406835763]
If I use this function implementation, I get 97% vs 3% which sounds more reasonable to me.
Maybe the predict_win
function has some flaw?
@nthypes This seems to be side effect of that function being generalized by us to accommodate for more than 2 teams. I just tested that function with openskill benchmarks and it does seem to perform better (by around 2%) on 1v1 teams. But overall it doesn't seem to affect the benchmarks by much. We actually created our benchmarks against the win_probability
function proposed in that thread to compare against trueskill. As expected the ratings will get more accurate the lower sigma gets. If that 2% matters to you, feel free to create a pull request to add it as a special case for 1v1 teams.