A probability-based cut-off is used to stop investigating branches of a search tree that are so unlikely they can be effectively ignored. This occurs when planning under uncertainty including in game playing. In real life we know there are highly unlikely events that may occur, for example, winning the lottery, but these are so unlikely it is not worth including them in practical planning. A probability-based cut-off works well if there are outcomes with very different probabilities, but may be more problematic if there are vast numbers of possbilties, each with low probability; although each is unlikely, together one may be bound to happen.
Used in Chap. 11: page 156