Author: Uri Blass
Date: 02:38:35 04/14/04
Go up one level in this thread
On April 14, 2004 at 04:34:07, Joachim Rang wrote: >On April 13, 2004 at 23:50:56, Eric Oldre wrote: > >>As those who have read my few other posts have heard, I've been working on a new >>chess engine project. >> >>I've made a lot more progress than i've expected, but still have SO far to go. >> >>In my "Think" function i started with: >> >>continue with iterative deepening until depth x. >> >>then improved to search to time X (where X was a reasonable amount of time to >>spend for one move give the time situation.) >> >>finally i changed to my current strategy: >>1)figure out a sensible time to budget for this move. >>2)keep with iterative deepening until: >> a)i don't think i can complete 1 more ply in the time allowed >> b)if i still hit the time limit stop unless i'm almost done with that ply >> >>The flaw with this of course is that my engine will still spend lots of time >>searching for a move when the response should be obvious. >> >>I was wondering if anyone had any advise for the logic to implement to not use >>all the budgeted time if there is only one move that seems reasonable. >> >>Thanks, >>Eric Oldre (new chess programmer) > > >another idea is to look on the number of nodes it takes to figure out that other >moves are interior. If this is unusually low your best move might be obvious. > >regards Joachim I do not think that it is a very good idea. Suppose that you have 2 moves(A force repetition and B gives you unclear position that you evaluate as slightly better). You will spend only one node to reject A but it is not an obvious move and it is possible that search to bigger depth will cause you to change your mind to A because B is going to get a negative score. Uri
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.