Author: Dann Corbit
Date: 15:48:03 05/02/03
Go up one level in this thread
On May 02, 2003 at 17:09:07, Geoff wrote: >Hi > >I would appreciate some advice on how I should procede in improving the chess >program I am working on. > >Currently it is very similar to TSCP. I will explain what I have changed and the >results I have got. > >Mod1) >Changed the Alpha Beta search to use use an Aspiration Window >This did give an improvement, reducing the number of nodes searched to get to a >given depth in most positions. >I did read that this could give search instabilites and might even cause lock >ups or crashes, I have not noticed this yet however > >Mod2) >Added a simple mobility function to the Evaluation. >This took my nodes per sec down from 500k/s to about 350k/s as I recalculate >every move for every piece again in the eval function. The playing strength >marginally decreased, probably due to the severe hit on speed ;-( > >Mod 3) >Added a save of 2 killer moves at each ply and re-sorted moves before searching >This made hardly any difference in reducing the nodes searched to get to a given >ply. I assumed this lack of effect was due to the fact that it has already got a >history[64][64] array built into the code > >I noticed that the good amateur programs get down to a depth of about 9 or 10 >where my program only manages 7 or 8 in a given time, so I guess I need to look >at things like Null moves and other complex search tricks ? > >Any advice as to what further mods I could do to gain the most improvement ? >Simple mods preferred first ;-) 1. Hash table 2. Move ordering (based primarily on hashing, so do step 1 first) 3. Null move. (make/unmake is the tricky part)
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.