Author: Robert Hyatt
Date: 17:37:37 10/20/98
Go up one level in this thread
On October 20, 1998 at 09:29:45, Jari Huikari wrote: >On October 20, 1998 at 09:03:46, Jari Huikari wrote: > >>What about only having higher value for beta cutoffs in Ply=2. Won't it find >>then exact values of the best replies for opponent and so with quite a low >>cost values for legal moves? Two plies minimax, the rest alpha-beta. Can it >>cost many plies? > > Forget it... I have drunk too few coffee this morning! > >Oh course it will take 19 times longer. But would this give any better >move ordering if used only in searchdepth=2 in iterative deepening? The >move order could be slightly better for deeper lookaheads? > > Jari You ask two questions without realizing it, so I will rephrase and answer each implied question separately: (1) will a 2-ply (or whatever shallow depth) search done like this (minimax at the root to produce a true score for each move) produce decent move ordering? The answer is yes. I did this in Cray Blitz as a root-move ordering strategy as well as a debugging strategy (ie why is Rb8 better than Rc8? with true scores for every move you can answer that, even though it takes a bunch of time). (2) will such a search be better than what we do today? Not really, based on some comparisons I did when starting development on Crafty. In some cases it would be better, in others, worse. I chose to not do it due to the extra complexity of managing this... But I'd suggest anyone working on a chess program try it, if for nothing else than the nice debugging facility it gives...
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.