Improving Determinized Search With Supervised Learning In Trick-Taking Card Games