Tutoring Strategies in Game-Tree Search

Pursuing the idea of a good working knowledge of one's opponent model (OM search), loss-oriented search (LO search) goes beyond OM search in occasionally presenting the opponent with an intentional give-away move.When this model is extended to the tutorial level, a delicate balance must be maintained: giveaway moves are allowed under the important proviso that, on a balance of probabilities, they should go unnoticed.It is acknowledged that the model is possibly too detailed to be realistic and rather naively replaces the stochastic quality (does the opponent recognize a give-away move?) by a numerical value, Yet, the paper provides valuable though preliminary results on an idealized opponent and how to tutor him.