Author: Robert Hyatt
Date: 05:59:23 12/06/99
Go up one level in this thread
On December 06, 1999 at 02:26:00, Bertil Eklund wrote: >On December 05, 1999 at 18:32:19, James Robertson wrote: > >>On December 05, 1999 at 05:06:57, Bertil Eklund wrote: >> >>[snip] >>>If your program can play for about 2500 in match-play it should probably play >>>about 2600 in tournaments and this is the usual way to achieve an established >>>rating. Humans play probably around 75-125 elo lower during a tournament, >> >>There is no way to see if humans are playing 75-125 ELO lower in a tournament >>because you can't play a tournament and a match at the same time and compare the >>results. Humans may get tired after 7 rounds, but then they would get tired in a >>match too. DB-Kasparov match #1 didn't seem to show this trend, though. And >>Fritz lost more points in the last half of Frankfurt than in the first half. :) >> >>In other words, all this is conjecture and [maybe] wishful thinking on the part >>of us programmers. >> >>James >> > >Hi! > >Extremely strange conclusions from you, the Rebel test is so far single games >compared to a normal tournament of say 9 to 13 rounds. It´s well known that >players often are tired in the later rounds and sometimes do horrible mistakes. >That´s not a big problem in the same pool of tired players but I guess it >shouldn´t affect a program. > >Regards Bertil > > >>>because they can´t prepare as much and are much more tired in round seven than >>>during the first round. Increments is also a very big advantage for the human. >>>During this Internet-play, the human have double increments because of the slow >>>transmission of moves. If Baburin had been in the serious time-trouble (against >>>a computer)he was, during a normal game he had probably lost. >>> >>>Regards Bertil >> >>[snip] The opposite is also true. I have seen many players get _better_ as a tournament wears on... not worse. I think that in a match vs a computer, the computer will do worse as the match gets longer, because the opponent will learn its weaknesses and they won't be repaired.
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.