Author: Tony Werten
Date: 03:38:24 08/30/02
Go up one level in this thread
On August 30, 2002 at 05:21:32, Omid David wrote:
>On August 30, 2002 at 02:57:20, Steve Maughan wrote:
>
>>David,
>>
>>>I implemented all the mechanisms that i think i'm supposed to--nullmove, hash
>>>move first, then two killers, then winning static exchange moves sorted by
>>>expected material win, then moves sorted by their history score. I sort moves
>>>at the root with each iterative deepening.
>>
>>Try swapping Killers and SEE wins - I think that's the problem i.e. generally
>>wisdom says that the move order should be:
>>
>>1) Hash Move
>>2) Winning SEE captures
>>3) Killers
>>4) History
>>5) Losing SEE captures
>>
>>Regards,
>>
>>Steve
>
>There are interesting possible options for the priorities, take a look at:
>
>\bibitem{schaeffer89} Schaeffer, J. (1989). The history heuristic and alpha-beta
>search enhancements in practice. \emph{IEEE Transactions on Pattern Analysis and
>Machine Intelligence}, Vol. 11, No. 11, pp. 1203-1212.
>
>\bibitem{schaeffer83} Shaeffer, J. (1983). The history heuristic. \emph{ICCA
>Journal}, Vol. 6, No. 3, pp. 16-19.
One paper is 13 years old, the other 19. Don't rely on such old papers too much.
In that time searchdepths were pretty low compared to nowadays. Deeper
searchdepths much historytables quite useless.
I've thrown it out of XiniX and didn't find a negative consequence. Only a
positive, my program ran faster.
Tony
This page took 0.01 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.