Is the Greed ladder score based on just win / loss ratio, or does it include the time taken to win as well?
It seems like some solutions will get you a higher win/loss ratio while others will get you a lower ratio but faster win times.
1 Like
Hi Traitor,
The final ranking will be sorting by wins and losses, but the current leaderboards use a Bayesian approach.
You can see the paper here that our ranking system is based off of, and my implementation.
Basically, it tries to gain an estimate of your solution’s strength. It’s similar to the method that Halo uses for ranking players (TrueSkill). It basically pre-calculates the estimated odds that you will beat another player, and then once it knows the actual outcome, adjusts both players’ scores.
2 Likes
Will final score be ranked based on wins and losses of the most recent algorithm, or the cumulative sum of all attempts?
Final ranking will be based on wins and losses of your last submitted code during our exhaustive top-humans-vs.-top-ogres pairwise matching round. We’ll take maybe the top 200 or so from each leaderboard, and each will play all of the top 200 on the other leaderboard.