Doesn't this method also (over) simplify? It goes into pseudo explorer mode anytime there is a >0 probability that the presumed worst cast is actually better th

Doesn't this method also (over) simplify? It goes into pseudo explorer mode anytime there is a >0 probability that the presumed worst cast is actually better than the presumed best/better case. Shouldn't there be a threshold to that process, so that the presumed worst case must have at least a (for example) .05 probability of being better before the algorithm gives it a shot?