Author: David Rasmussen
Date: 05:09:18 03/16/01
Go up one level in this thread
On March 16, 2001 at 02:48:19, Uri Blass wrote: >On March 16, 2001 at 02:29:59, David Rasmussen wrote: > >> >>If searching one more iteration takes 6 times longer than the previous, you just >>do the simple math between iterations to estimate whether you should continue >>searching or not. But as I've said: I wouldn't recommend this, as it can't >>really be used in real life. If such a program were to participate in a >>tournament, it would probably lose often because of it's poor time allocation >>strategy. > >It is not clear. > >Stopping at the end of the iteration and deciding when to stop only based on the >time that you finished the iteration and the time limit of the game(except cases >when you are in danger of losing on time) is not the best strategy but I do not >think that you can earn more than 50 elo from a better strategy. > I was talking about just doing this plain and simple with no regard to time trouble. In that case, you are in great danger of losing on time, because the branching factor is not constant. I think that a good time allocation strategy is worth a lot more than 50 elo, if you compare this simple strategy with a Crafty-like approach. But I agree that different, not too simplistic, time allocation strategies, as found in different top programs, is not worth more than 50 elo.
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.