Author: Stuart Cracraft
Date: 11:01:47 12/27/05
Go up one level in this thread
On December 26, 2005 at 08:38:13, Roman Hartmann wrote: >Hi all, >recently I had an idea how to reach higher search depths (in the endgame or >progressed middlegame) which keeps bugging me. > >Well, so far it's only and idea and I didn't even try to implement that so far >(don't have hash in my engine yet) and maybe the idea isn't that new at all >anyway as the tablebases are doing something similar in the endgame as well. > >The idea is to reverse the search tree. So after the ordinary search I start >with a mate (or lets say with possible mating positions) and go backwards in the >tree to see if I get a position that the ordinary search has already stored in >the hash. If so I found a mate. > >Maybe that's not of much use in a chess engine but it might be worth a try in a >mate searcher. The advantage is (or at least I'm under that impression) that two >rather shallow trees are checked for identical positions rather than one much >huger tree. > >Any comments on this? > >Roman The issue with this seems to be how you would possibly make the leap from the program's current position to a bunch of "possible mates" without search in order to then verify it with a mate search. That jump is a huge leap in A.I. and *it has not been made* - i.e. for any given position, draw up a set of possible mate positions. Grandmasters can do what you describe but I don't think even Dave Wilkins Paradise program did things like that. Of course, if you make a huge advance, let us know about it - I don't think I am horribly wrong on this assessment though. Stuart
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.