Author: Vincent Diepeveen
Date: 06:13:27 11/05/01
Go up one level in this thread
On November 04, 2001 at 17:15:35, Hans van der Zijden wrote: >I entered the Open Dutch tournament without any hope of scoring a point. I would >have been satisfied with one not too bad game. The reason for this is that I >just started programming on Gadget the last 5 months I think. When Gadget >reached a depth of 6 ply in about 90 seconds average, I started thinking about >implementing quiescence search. This proved to be too difficult for different >reasons (one being that I am very lazy). Eventually I had to decide to play with >the old version. Apart from not having QS I also lacked hashtables, permanent >brain, nullmove, some necesary chessknowledge and who knows what more. I >practised against the Kasparov GK 2000 (about 1850 elo). Gadget on 90 minutes >for the whole game and the GK 2000 on 1 sec/move. Gadget lost 3½-½. And now I >had to play 2300+ programs calculating double the number of plies Gadget >reached. But it only got severely beaten by Fritz (mate in 23) and Kallisto II >(mate in 31). Most other games it was only slowly smothered. Against Diep Gadget >last a pawn on move 20 but then it took Diep 24 more moves before he won the >second pawn. Against Ant it was a clear draw after 74 moves but then Gadget took >a poisened pawn and lost after move 93. Gadget even won a game against another >debutant. What happened? Am I dreaming? Search depth is not holy that's what you're seeing here. If you get outsearched loads of ply, that's not important in most positions. Only a few positions it's important to search deeply. More important is to ask: what depth do i need in order to beat the others? Kallisto played 1.e4 against you and got a very aggressive Petroff against you. So there are more tactics there. Diep played 1.d4 Nf6 2.c4 e6 3.g3 d5 4.Bg2 dxc4 5.Qa4+ Nbd7 6.Qxc4 a6 7.Nf3 b5 Not exactly aggressive. Fritz played 1.e4 against you, AMAZINGLY gadget played Ruy Lopez here which soon became a very aggressive line. Crafty 1.d4 so that was a long game. Morphy was actually smartest by playing a gambit, but morphy played some bad moves in opening and then it was not so easy to win it soon. Nevertheless still 42 moves is quick enough considering move 42 was the mating move. In short there were 3 programs that opened 1.e4 against Gadget and they all won pretty quick. The others that played 1.d4 needed way longer. Some even 1 hour for a single move to decide if i have heart the rumours well... :) Nevertheless it's obvious that chess is not about search depth. Yes you need a certain search depth to not lose everything somewhere in a game based upon tactics. Obviously that's more than 6-7 ply, but i would be amazed if a 10 ply search with all kind of extensions turned on, like i used to get this tournament on the 800Mhz machine, whether this was 'deciding' games. More logical to assume is that better knowledge works better than another ply. Best regards, Vincent
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.