Author: Vincent Diepeveen
Date: 18:57:26 01/09/00
Go up one level in this thread
On January 09, 2000 at 20:37:20, Christophe Theron wrote: >On January 09, 2000 at 20:25:44, Ernst A. Heinz wrote: > >>Hi Whinecent, >> >>>But diep plays without book now and with a 9 ply search or something >>>you won't play the opening good! >> >>Oh yeah, finally even you got it -- that's exactly why searching >>deeper seems to be beneficial to nearly everybody except you. > > >Ernst, this is unfair. > >You are deliberately helping my opponent by giving him excellent advices. > >Stop this immediately or else I have to call the moderators! :) i've done experiments with my draughtsprogram. it gets 30 ply easily at overnight analysis. it gets 12-14 ply at tournament level. I don't see any better moves appearing. of course sometimes pv changes in positions where there are 2 moves the same. those positions aren't interesting. If i add knowledge to the draughtsprogram however then a certain amount of positions. Let's call it i will be played better. Now you directly will argue that a number of positions j will be played worse because of adding the knowledge. If i can take care that i > j, then my program improves right? That's perhaps even for some dudes possible to understand! Yet when that program searches more than twice as deep then it still isn't worth the investment to search deeper than the tactical barrier. However there are some draughtsprograms we play against that do NOT get to that 12-14 ply even. In that case we can kick ass. We also play against a program called 'flits' a lot. despite that we always outsearch it pathetically, Napoleon (name of our program called after 'coupe napoleon') still doesn't manage to win from it. Now 2 new (french) draughts programs have arrived on the scene. One of them made by a very strong draughtsplayer. He has added some knowledge to the program. In contradiction to chess it's of course easier to search deep in draughts. therefore this program searches not undeep. Still napoleon outsearches him everywhere. then we got to a position where our program took a forward post. our program didn't have the right knowledge about it, so we got into a lost position. Afterwards i put this position at the program. At all depths it would play that silly move. Searching simply doesn't work. You do need however to get past a certain depth to see tactical sufficient. I have said years ago that this depth is about 12 to 14 ply. Now most will argue that this is not true, but that it is like 10 ply. I think that this depends upon how many extensions you do, how much you forward prune etcetera. Yet the idea is clear. The fact that you laugh about it now, only says enough about your personalities and little insight in computer games. > > Christophe > > > >>But, admittedly, it is hard to implement a good search that is >>also fast enough to go deep. Hence, I always wondered about the >>name of your program "Diep" which apparently means "deep" in >>Dutch, doesn't it? >> >>=Ernst=
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.