Author: Gerd Isenberg
Date: 13:43:52 03/01/06
Go up one level in this thread
On March 01, 2006 at 15:51:29, Robert Hyatt wrote: >On March 01, 2006 at 15:42:18, Alessandro Scotti wrote: > >>Yes I increment "count" every time a move is searched, and "fail_high" for moves >>that failed high. >>I actually did the implementation very quickly in order to participate to CCC >>(here's a quick report assuming you want to practice a little Italian before >>Turin! ;-) http://www.ascotti.org/programming/chess/ccc2005.htm) and with Heinz >>we could see it gave maybe 15-20 elo (at fast time controls). That version had >>validMoves >= 3 and depth >= 3 also. Later I changed both values to 4 and got >>another 15-20 elo. After that I decided to start working on another engine so I >>didn't experiment anymore with these settings. > >OK. I really worded my post badly. Here is what I wanted to say: > >I increment the fail high count by 100 each time a move fails high. I also >increment the tried count by 1 at the same time. Any time another move fails >high, and this move was tried but did not, its tried count gets incremented, >which lowers the "failed high percentage" significantly... > >sort of: > >if failed_high(x) > x.count++ > x.fh+=100 > >if failed_high (y) (later) > y.count++ > y.fh+=100 > if (tried(x)) x.count++ > >Hope that makes what I am doing clearer, whether it is good or bad is not >certain yet... > Ok, now i understand - looks reasonable and sounds interesting. Also a kind of additional move sorting measuring. Since you usually fail high early the additionl loop cost looks negligible. And since you have some plies left to the horizon, it is not performance crictial at all - same for the expensive looking division. Anyway, depending on your percentage you may calculate two increments to do a simple compare of fh > count. How do you index the history table, from-to or piece-to?
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.