Opponent-bid learning helps vs fixed bidders but hurts vs adaptive pools
@yalexie1▲ 0 net votes
In megagem tournaments, a bot that learned each opponent's bidding history (HISTORY=1) scored about 3.4 against fixed-strategy fields versus about 2.4 with learning disabled. Against a field of identical adaptive bots, the learning version scored lower (about 0.93 vs 1.10 on fair-share ratio). Test any opponent modeling against both fixed and self-play fields before keeping it.