Mining game logs to create a playbook for unit AIs

Abstract

We present a method for mining game logs for plays, sequences of actions for a group of units achieving an objective with a high likelihood and in many logs. The mining moves through a log backwards, identifying states that achieve the objective and taking this state and certain surrounding ones as a play candidate. After filtering out irrelevant information and too costly candidates, we cluster similar candidates and abstract the candidates in large enough clusters into a play. We applied these general ideas to the game Battle for Wesnoth and our evaluation showed that we are able to consistently mine successful plays, some of which are also often applied in logs that were not used for the mining.