Exactly because evaluation is not good enough. If evaluation were good enough, a one-ply search would be sufficient to win any game. But Evaluation is not that good, so the score of a deeper search is usually different from the static evaluation. From this it follows that the evaluation score at the leaf node should be expected to change on deeper search.George Tsavdaris wrote:Why it is a fundamental mistake?
If evaluation is good enough why should we care about how did we got there? And even if eval isn't good enough anyway, why we should care about the path anyway?
And when it changes for the worse, you might not want to go there anymore. In that cae you get stuck where you are. And if that is in a position with a poor evaluation, you are of worse than when you are in a position with a good evaluation. So when you have the choice, it would be wiser to take the path through good intermediate positions, rather than the path through bad ones.
See that you don't get into trouble, then you don't have to get out again. Especially if the path you had in mind to get you out of trouble is unreliable, and might later turn out to be a dead end.