Author: Peter McKenzie
Date: 00:50:00 02/10/99
Go up one level in this thread
On February 10, 1999 at 01:54:25, KarinsDad wrote: [big snip...] I'm not totally sure what you're saying below, but I'll try to answer anyway :-) >Ok. So in a given position (ply 0), there may be 5 captures (maybe even 10) that >are real losers (ply 1). For each of these, I would quickly find a response (ply >2, e.g. pxq). To do a null move, I would have to check about 36 reponses (on >average, not endgame) by white followed by 36 responses by white (i.e. all legal Ok, so far we have : Ply 1. Q moves somewhere stupid, Ply 2. pxQ You then talk about doing a null move, but it is not clear to me when you want the null move to be done. In this line, I think the important null move will come at ply 4. So we might have: Ply 1 (white) Q moves somewhere stupid, Ply 2 (black) pxQ Ply 3 (white) try all moves here Ply 4 (black) try Null Move first (then try other moves) Ply 5 (white) try all moves Stop HERE if doing a 7 ply search (using R=2). So with any luck, the Null Move at ply 4 will hardly ever be refuted which means that in that branch you save yourself a whole **2 plies** of searching. I personally think that Null Move Pruning is the most significant computer chess algorithm to have been discovered in the last 10-20 years. >moves at ply 3 and all legals moves as if it was still white's move at ply 4) >for a total of 36 + 1296 or 1332 generated legal moves (37 calls of the legal >move generator) and 1332 evaluations (but possibly not a full evaluation). > >If any of the "ply 3" or "ply 4" null moves lead to check, do you have to search >on? And what do you do with the "ply 3" check where you are doing null move, so You don't try the null move if you are in check, does this answer your question? >that black cannot move out of check before the "ply 4" null move? Do you just >ignore it? > >In a situation without the check (not knowing the answer yet), you pruned 10 >lousy moves (although there was no guarantee that qxp, pxq didn't lead to mate >in 3 or more) and there are 26 left which are also null prune checked, but they >do not fail, so their children can be searched and pruned normally. Correct? > >KarinsDad > >> >>IE with a full-width search, there are _many_ captures that are losers. Null >>move reduces the depth below these moves and make dismissing them quick and >>easy... >> >>>KarinsDad :)
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.