AI vs. AI tournament - Swiss Hybrid-Elimination Tournament
Okay, let's do this tournament - each game I will pick Mongols, and pack up into the corner to not interfere.
I will report a Hardest vs. Hardest AI battle, and update the tournament brackets with the outcome.
At the end of the tournament, I will use the win-loss matrix to calculate the Maximum Likelihood statistical estimator of relative ELO scores, using the parametric statistical model given by the Logistic ELO model.
Tournament Rules - Swiss Hybrid Elimination Tournament (SHET)
- There are n=3 tables, labelled from "top" to "bottom". Each contestant starts on the top table, and each loss moves you down one table.
- Each round you pair up and play. Continue until one winner remains!
The Buchholz pairing algorithm is as follows:
- Assign the "dominance order" lexographically - that is, primarily by the number of direct wins minus losses, then secondarily by first-order dominance wins minus first-order dominance losses, then by second-order wins minus second-order losses etc. On round r, go no further than the r-th dominance, after that, revert to some pre-determined seed/preference rank.
- If an odd number of contestants remain, then you award a bye to one of the players who has the fewest byes in this tournament already. Choose the "top" player by the definition in Clause 1.
- To pair the remaining contestants, you take the "top" player, and pair him/her:
a) Primarily with an opponent they have played the fewest times already in this tournament.
b) Among those options, choose the LOWEST ranked player with the same number of accumulated losses.
c) If no suitable players are on the same table, search in the lower tables in descending order of preference. - Delete the paired contestants, and repeat this process until pairing is complete
