Author: David Mitchell
Date: 11:14:33 03/01/04
Go up one level in this thread
On February 29, 2004 at 16:29:15, Joshua Haglund wrote: >On February 29, 2004 at 12:48:06, Uri Blass wrote: > >>On February 29, 2004 at 12:20:03, Joshua Haglund wrote: >> >>>On February 29, 2004 at 03:20:30, Matthias Gemuh wrote: >>> >>>> >>>> >>>> >>>>My engine has a very unstable and unreliable search. >>>>Sometimes I think I implemented what you are proposing. >>>>My opinion is that the blindness introduced by such tricks would really hurt. >>>> >>>>/Matthias. >>> >>>yes! it wouldn't do the best search if it didn't reach great depth where f = 3; >>>If your engine searches deep try ply > 12. If it doesn't try ply 8, etc... :) >>> >>> >>>I have reason to believe a person will gain atleast 1 ply in the same amount of >>>searched time. >>> >>>example >>>time = 30; >>>ply 7 = 4 seconds. >>>ply 8 = 13 >>>no more plies reached. >>> >>>// with idea. >>>time = 30; >>>ply 7 = 1 second >>>ply 8 = 5 >>>ply 9 = 20 seconds >>> >>>Maybe this would be good for long time controls? Skip shallow and go to deeper >>>lines. >>> >>>Thanks for your reply, >>> >>>Joshua Haglund >> >>I can only say that I do not understand your idea. >> >>If you suggest to do selective search in the first plies then it seems to me a >>bad idea because you may miss important moves. >> >>It is more logical to be selective in the last plies and programs do it for >>example by qsearch but even then I do not see why do you use fixed number of >>moves and the number of moves that you search should be dependent on the >>position. >> >>prunning illogical moves is something important to do and programmers know it >>so it is not new information. >> >>Uri > >I thought about doing a selective search also... pretty much the same thing. > >If you skip time wasted looking at first several plies, it'll get to a greater >depth in less time to look for good moves. > >Thanks for your reply, > >Joshua Haglund Joshua, your idea is a great way to get deep, and very fast. Unfortunately, the well you are digging will not have any water in it, no matter how deep, or how fast, you dig down. :) You can't throw out several shallow moves, and go onto the deeper parts remaining, and have success with Alpha beta. Every shallow move you throw away has it's _own_ deeper moves, which may lead to the best positions available. You will _never_ recover from that loss by simply going deeper with the remaining moves into the search tree. In a game like Tic-Tac-Toe, you can do that, sometimes. In chess, you can't recover. Try it and you'll see for yourself. dave
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.