Author: Tim Foden
Date: 05:07:17 04/24/01
Go up one level in this thread
On April 24, 2001 at 05:40:56, Rafael Andrist wrote: >On April 24, 2001 at 03:57:54, Tony Werten wrote: > >>On April 23, 2001 at 11:12:55, Rafael Andrist wrote: >> >>>In my chessprog, I'm using an Alpha-Beta-Search with an infinite window. After >>>adding a hashtable, only half of the nodes need to be searched, but I get still >>>a branching factor around 9. The use of Iterative Deepening didn't change much. >>>So now my question is what can I do to improve the search? Should I try to >>>improve the move-sorting? Or is it necessary to use other pruning techniques >>>like Nullmove? >> >>The use of iterative deepening doesn't nescesairy help the BF. It's advantage is >>that you don't need to decide on forehand how deep you are going to search. >> >>If you decide to search 10 ply and you run out of time, without ID you have >>nothing, with ID you have the best move of ply 9. > >I know, but I hoped that it helps sorting the moves too. It really does that >too, but only a little bit. > >> >>BF of 9 is a bit too high but not too much if you don't use nullmove. There are >>a couple a reasons why it could be this high: >> >>Are you using killermoves ?(you should, noncaptures only) > >Now I do :) and the BF has gone to 7. > >>Is your evaluation a bit intelligent ? (it should, at least material and passed >>pawns) > >Yes, my eval is little bit intelligent, I get good results in games with fixed >dephts. > >>Do you store bestmoves in the hashtable (you should) ? > >Yes. > >>Do you store bestmove if score is <=alfa ? (you shouldn't) > >No. Why is this bad? > >>How do you orden capturemoves (to start with, use MVV/LVA, most valuable victim, >>least valuable attacker ) >>Have you tried history tables ? (should help at lower depths ) >> > >my move ordering: > >- Hashmove >- 2 Killers (just implemented) >- checking moves >- capturing moves (sorted like you discribed above) >- history heuristic >- castling (if there are moves with equal scores, castling should be favoured) >- moving piece (pawns first, king last) >- square onto which the piece move (center, near the opponent is good) > With capturing moves, you may want to try leaving the losing captures till after all other moves. Also, it is normally good to try the winning/drawing captures before the killers. The current move ordering I use in Green Light (not at the root though) is: 1. hash move. 2. 'winning' captures (uses SEE here, but MVV/LVA is also good) 3. killers (GL has 2, only 1 can be losing capture) 4. 5 of remaining moves ordered by history heuristic 5. rest of moves in the order they occur (losing captures still last) >The history heuristic doesn't seem to work well, I may have to improve this too. > >Thanks > >Rafael B. Andrist > >>Have fun trying out, >> >>Tony >>
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.