Author: Will Singleton
Date: 09:17:33 01/05/00
Go up one level in this thread
On January 05, 2000 at 04:21:58, David Blackman wrote: >On January 04, 2000 at 15:46:56, Will Singleton wrote: > >>On January 04, 2000 at 08:36:08, Tijs van Dam wrote: >> >>>Glad to see that GKJunior is at least in one list not at the bottom :( Sigh! Do >>>you guyes have so much spare time, or have you been working on your proggies for >>>years? >> >>All of the programs are at least several years old, at least the ones that play >>regularly. Often new programs will come in, play for awhile, then drop out. I >>suspect the issue of time is paramount. One cannot compete effectively without >>a long period of trial-and-error, learning and fixing, trying new things, trying >>old things, etc. It just cannot be done in a few months, or even a year. And >>that's just to get a halfway decent program that will beat most humans >>regularly. To go the next step, and get to the point where you beat most >>*programs* regularly, that either takes serious time or divine inspiration, not >>sure which. >> >>Will > >KnightCap by Andrew Tridgell was playing at mid to strong club player strength >after only a couple of days work on the chess engine. A couple more days work >and it was stronger than most club players, perhaps in the expert range. After >the huge amount of work it took me to get Desperado to that standard, i was a >bit surprised at that. A couple more days work and KnightCap was beating >Desperado easily and giving Crafty a tough game. All this was spread over about >a year though, since Andrew is normally very busy on other stuff. > >Look out Tiger, Fritz, Crafty et al. Maybe Andrew will find time to do a couple >more days work on KnightCap. After appearing on ICC a few years back, KnightCap (TDChess) improved it's blitz rating rapidly (a few weeks) to about 2400-2500. It has stayed there ever since. The intent of KnightCap was to implement TD learning, and not to develop the engine. Therefore, I don't know what engine he used as the basis for his development work, but I'd assume it was some freely available source. If your goal is to work solely on the eval, then you do not work on the engine. In fact, to work on the engine would invalidate the results of the experiment. So, I would say that your point about improving the program at a rapid rate is misplaced, because since it appeared on ICC no changes to the engine took place. The changes were limited to the effects of TD learning on the eval. btw, what happened to Desperado? Does it play on ICC or Fics? Will
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.