Author: Ricardo Gibert
Date: 23:33:18 01/10/01
Go up one level in this thread
On January 10, 2001 at 17:50:40, Dann Corbit wrote: >On January 10, 2001 at 07:26:35, stuart taylor wrote: > >>The question is all in the heading. I mean with hardware of about 450 mhz. >>upwards (till 1.2 ghz?). >> This question is an ofshoot of Uri's comment that Rebel does better with more >>time. >> If you want to analyse a move for 2 hours, which program would have seen most >>(of what is important and relevant, and consequently play the strongest move) in >>those 2 hours. Or longer? > >Not only do we not know the answer to that question, we'll never even come close >to knowing. Never? That's a long time. You're forgetting about Moore's Law, which can effectively convert hours into minutes. Also, someone with access to a large LAN could do the test. Besides, if the programs compared are competitive at normal time controls, probably the one with the lowest average EBF is strongest at the much slower time controls. Naturally, a test is the only way to know for sure, since a lower EBF can have it's drawbacks. > >Calculate the time it would take to play 100 games between 2 engines. Then, >multiply that by 30 high quality engines. With 100 games, you will have an >enormous error bar. So 1000 games will be more like it to find a really >accurate answer. > >At two hours per move, a computer won't play a whole lot better than at 40/2. >That's because of the exponential nature of chess. By 40/2, most engines have >pretty well hit the wall. You will probably get less than two additional plies >by running two hours. If you let it run 24 hours, maybe another ply or two. > >I would suggest asking correspondence chess players. Maybe Robin Smith can tell >us what he thinks about them, since he has faced many computer opponents. Such >an evaluation would be subjective, but a lot better than a wild guess.
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.