The Reason for the Benefits of Minimax Search

Abstract

Since there existed no convincing theoretical explanation for the usually observed benefits of minimax search in practice, we investigated two instances of a class of tree models which are based on the concept of quiescence. (This way the strict separation or static and dynamic aspects in prac tical programs is modeled.) We performed Monte Carlo simulations, enhanced by analytic results. The behaviour of these models in our studies gen erally corresponds quite well to observations in practice (especially that of the model based on the more restrictive definition of quiescence). Hence, we found empirical evidence for an earlier conjec ture, and these results can serve as an important step towards understanding the reason for the benefits of minimax search. 1