Leveling the playing field: fairness in AI versus human game benchmarks