Author: Dave Gomboc
Date: 17:01:01 05/06/99
Go up one level in this thread
On May 06, 1999 at 19:37:02, Peter Kappler wrote: >On May 06, 1999 at 17:17:10, Luis E. Alvarado wrote: > >>I do not understand why the BLITZ ratings of IM and GM masters on ICC are so >>inflated. Is this normal? Why some GM with ELO ratings betwen 2500 and 2600s >>have BLITZ ratings in the 3000s. It appears that these rating are biased by mor >>e than 400 points on the top end. > > >ICC ratings are unreliable for a variety of reasons: > > >1) Cheating. Many players use computer assistance (wihout saying so). This >causes their ratings to fluctuate wildly. > >2) Non-random opponent selection. I know players who will avoid an opponent >whose rating is currently below their true strength. Likewise they will try to >players that they think are overrated. Or they will find an opponent who >doesn't match up well with them and play them over and over. My experience is >that a few of the Crafty operators are the worst offenders... > >3) Anti-computer specialists. This is really just a special case of #2. These >people play a very slow, quiet game that is difficult for computers, or they >search for holes in the opening books of automated computer accounts and exploit >them. My program, Grok, has recently suffered from this... Thank these people. They are doing you a favour. Every time they bust your opening book and you have to fix it, that's one less hole for someone to hit you with in a real tournament. >4) Varying time controls. This may sound silly, but the fact is that some >people are particularly good at certain time controls. If a 3 0 specialist >happens to play a 3 0 match against another player who is used to a slower time >control like 5 3, the result will probably be skewed in favor of the first >player. > >I could go on and on, but I think these are the main problems. Obviously #4 is >pretty minor, but #1-3 cause big problems with the efficiency of the rating >pool. Since the computer-related problems tend to have more impact at the high >end of the rating scale, I think the ratings are particularly unreliable there. I would say that #4 is also very important. For instance, if I play 5 0, my rating will drop to the 1800s. When I play 2 6, I can hold high 1900s or low 2000s. If I play 2 12, I can even get over 2100. Of course, it's much tougher to find people to play 2 12 than 5 0. 2 12 gives lots of chance for someone to operate a computer, and 5 0 doesn't, so it's easily understandable. >--Peter Dave
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.