Author: George Tsavdaris
Date: 15:00:51 07/21/04
Go up one level in this thread
On July 21, 2004 at 13:17:26, George Sobala wrote: >It has been claimed that a REAL super GM would also have struggled to get a good >TPR against such lowly opposition as Shredder faced in Argentina. > >To investigate this further, I posed the question: how would Kasparov have >performed? > >To do this, I have looked at all games played by Kasparov against opponents with >ELO 2200 to 2500 from the year 1990, as recorded in Chess Assistant Hugebase, >with updates. > >His results are: > >+57 =23 -2 > >Assuming average opponent ELO is 2350 (I cannot be bothered to calculate it >exactly and this is about right): > >TPR = 2632, 95% CI = 2538 to 2683 > >Haha! Kasparov, like Shredder-8, is undone by weak opponents! > Yes, if someone has a 2700+ FIDE-ELO rating, this normally means that he would have a tournament performance rating of about the same as his FIDE-ELO, if he plays on a tournament with some players. But it is a fact that although this is true for tournaments with players that have a FIDE-ELO of more than 2600, when the FIDE-ELO decreases and it is about 2300-2400 it is VERY difficult to keep your FIDE-ELO on the same level. Instead it is almost a rule that the 2700+ player would have a TPR much lower. That is why high rated players avoid playing in those tournaments. The reputation of their high rating would be in danger. So the fact that Shredder had a TPR much lower than the excpected it is not something that can make us conclude anything about the FIDE-ELO of Shredder. Also some other reasons for the not best TPR of Shredder is that it used a book that is commercially available. Also the machine it was running was not the best available. There are now available quad's or 8x machines. And also the TPR performance of Shredder in a tournament of 2.431 mean FIDE-ELO, doesn't say anything about the expected TRP of Shredder in another tournament of 2.700 mean FIDE-ELO. I bet the second TPR would have been much higher and Shredder could even win the tournament. Also the behaviour of computer's ELO is strange. They can win a 2700+ player and they can draw a 2000 player. This comes from their HUGE weakness to play well at closed games. If one player succeeds to create a closed game then the game can be drawn by him. And this weakness comes from their incompetence to plan. Until humans(programmers) find a way to solve this computers will play 3000+ at open game, 2700+ at semi-open games, and 2400- at closed games.
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.