Author: Ed Trice
Date: 12:53:22 06/22/04
Go up one level in this thread
This is essentially the "search vs. knowledge" discussion in microcosm. Just remember (for the layman) that the game tree that a chess program explores branches out at roughly a factor of 3 if your software uses all of the contemporary software tricks out there. So, a machine that is 3x as fast as you would allow it to search 1 ply more deeply, on average. A machine that is 9x as fast would outsearch you by 2 plies. Now "way back when" even a 1-ply advantage in search depth at tournament time controls was almost universally won by the deeper searching engine. Ken Thompson published a very brief, but very influential paper on this in the 1980's. It could be argued that this one paper may have influenced more programmer to go the "fast dumb" routine (as opposed to designing a "smart" program that would search slower.) I also seem to recall Dr. Hans Berliner once stated that if he disabled all of his 'intellegence' in Hi-Tech, which allowed it to search faster than the smart version, the "Low-Tech" version won. There is also now an interesting inflection point occuring in the search-knowledge debate. Did not Rebel win one game (originally supposed to be a match of several games) with Crafty, offering Crafty odds of 100:1 in search depth? While one game is a very insignificant sample (indeed, the smallest sample size possible) I think that result should indicate that the quality of knowledge, if designed intelligently, can be very, very useful! I am not sure what is in the Rebel leaf node evaluator, but it must be very good if it could oust Crafty at 100:1 time odds. I guess this was a very long way of saying that, perhaps, all of the emphasis on faster hardware could be undermining the real core of the software's greatest asset: its evaluation function. > > Chess Tournaments. > > If I were the author of a chess program I would want the very best hardware > to run my program when competeing against others. > > But I am not author of any chess program.My only interest is to know which > program is the best.And this is not evident since chess programs when playing > against the competition do not necessarally use equal hardware. > > > A simple question.Why is it not oblgatory that chess tournaments require > participants to operate using the same hardware. > > I have don'd my flack jacket and run for cover...........Alan
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.