Author: Ratko V Tomic
Date: 08:59:53 11/21/99
Go up one level in this thread
> Now I play in e-mail tournaments and have 2515 rating. ALL MY PARTNERS > use top engines without any understanding what's going on the board. ... > Unfortunately I play against 99% postmen. Till now I finished 20 games > and I won 14 and 6 draws. Not too bad result against Fritz and Hiarcs, > what do you think ? 17 out of 20. ... > So, please be realistic. The level of all top engines in correspondence > games is around 2400. The rating difference for the 17:3 result is 280 points, which from your 2515 would put computers at the very long (postal) time limits at 2235. That looks closer to the real rating at those limits than 2400 that you suggest. At these time limits humans rarely make the kind of cheap oversights which are the bread and butter of the computers' strength against humans, especially at blitz. I agree with your judgment on general overestimating of the programs' strength in the serious chess games. Even without going to the postal time limits, but at the tournament time controls and a large number of games against the same human player, the program's rating would drop gradually over several weeks from the 2400-2450 (which may be a valid for the first few days of such match) down to around 2300, where it might stabilize (unless the human is particularly motivated in which case it may go even lower). Only at the fast time controls would the program keep its IM/GM range ratings. On the other hand, in a few years perhaps, if someone manages to complete and implement the Botvinik's Pioneer algorithms for chess, you will feel lucky to get 3 points out of 20 against that species of programs, at any time limits. >Now I have Fritz 5.32, Hiarcs 7.32, Junior 5 and Nimzo 99. I didn't buy Rebel >because I don't like to use DOS programes. It is a bit of annoyance, but in case of Rebel it may be worth the little extra trouble. From my play against the same programs you tried, plus Rebel, Rebel has the most solid strength among them, the least subject to the rating drop over time against the same human opponent.
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.