Single Comment

With "proper citeria" I meant characteristics that can be formulated as unambiguous rules, rather than vague (but possibly very accurate) expert opinions like "this looks pretty good". Of course a lot of progress has been made in extracting knowledge from experts that the experts themselves are not sufficiently aware of that they could formulate it algorithmically. Especially in Shogi this has been done a lot. <p> Note that pruning in modern Chess programs is hardly ever done (except very close to the tree leaves), for exactly the reason you say. The emphasis is more on reduction, i.e. moves classified as less promising will be searched less deep as the more promising moves, but their depth will increase without limit as the depth of the search increases, just at a slower pace. So that the knowledge gained from the search can be used to re-classify them.