Author: Colin Frayn
Date: 03:34:44 04/11/00
Go up one level in this thread
>Chess 4.7 was probably around 2000 rating before 1980. 1300 might be about right >for the best microcomputer programs in 1980. (That's 1300 by 1980 standards. I >think the average 1000 player today should beat any microcomputer program from >1980). Remember that computers don't make mistakes. You can get a computer program exhaustively searching 3 ply depth and it still plays ptreety well in shorter time controls and you have to be careful when you're playing blitz. Especially if the positional evaluation is good. >Hardware improvements on their own should be enough to get computers to between >2800 and 3200 by 2010. Hmmm... dubious. Of course the improvement in ELO with processor speed is not linear, but goes more slowly than that. I think that the top supercomputer type program in 2010 will be able to make over 3000 ELO. You have to remember that computer speed probably won't continue to increase at this rate for another 10 years either. There are some pretty large physical constraints beginning to feature in the latest generation CPUs. > However i expect software improvements to be even more >important, resulting in computers stronger than 3500 by then. I disagree. It would take a few stupendously impressive algorithms to generate that sort of ELO improvement. Full 6 man tablebases won't make _that_ much difference, and even enormous opening books are probably still not going to be the deciding factor when playing against human GMs. Perhaps improvements in making computers play well against humans might well cause a few ELO improvement, but I think 3500 is extremely optimistic. I think we'll have supercomputer programs playing as if they were humans with a rating of 3000, but anything more is probably wishful thinking. :) Still.... 3000 would be nice :) Imagine the new middlegame theory...... Cheers, Col
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.