Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: About search algorithms and heuristics

Author: Brian Richardson

Date: 07:34:06 05/26/00

Go up one level in this thread


On May 26, 2000 at 09:21:01, Josť Carlos wrote:

>  I'm curious about what search algorithms and heuristics do commercial and
>amateur programs use, for example, null move vs no null move, PVS vs MTD(f) and
>so on.
>  In Averno, I use PVS and AS, and haven't tried null move cause I'm a little
>afraid of losing precision in short tactics, though I see most programs use null
>move.
>  Junior is said not use null move. Do any other of you guys or commecial
>programmers not use null move? If so, what are your reasons?
>  Who of you have tried MTD(f)? Why did you keep/discard it?
>  What other algorithms/heuristics are you using or did you discard?
>
>  Thanks in advance.
>
>  Josť C.

Tinker uses PVS and single-level hashtable (for positions, but not pawns--yet),
used in both full-width and q-search.  Then extensions and lazy eval, then null
moves (variable based on depth and pieces left), then killers, then hash move
(if any), then generates moves.  Next internal iterative deepening (when out of
PV), then PVS (full windwow first move then small window with re-search on
failures), plus futility.

Whew.  The order of these is VERY important (e.g. hash tables before in-check
test), at least for Tinker.  Also, each one gets tweaked.  Each program will
become unique and changes in one area often impact another area and overall
results.  Am now working on faster automated testing process to handle
incremental improvements.

Brian



This page took 0.01 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.