Author: O. Veli
Date: 02:13:30 01/02/00
Go up one level in this thread
On January 01, 2000 at 12:45:29, Leon Stancliff wrote: > I am very much interested in determining as valid a criteria as possible for >classifying computer chess playing programs as Grandmaster level. What number of >games should be considered as reasonably reliable? What achievements should be >evaluated? A player has to achieve 3 norms in 24 games. One of the tournaments must be a round-robin. If no round-robin is played than the norms must be achived in 30 games. A 13 game norm in an Olympiad gives you the title. > Since Fide titled players are hesitant about allowing computers to compete in >the higher class tournaments required for Grandmaster norms, No. The reason that we do not see any computers achieve an official FIDE rating is the fee that needs to be paid to FIDE. A program can enter a tournament, play 9 games, earn a valid rating, provided that FIDE is paid 10,000 Swiss Francs IIRC. Nobody has made such a payment yet. Anybody can get GM norms in an open tournament, no need for a high class tournament. > First, we need to know precisely what the Grandmaster norms are at present. >Arpad Elo gives them in his book THE RATING OF CHESSPLAYERS, published in 1978. >At that time it appears that 24 games was the required number. A performance >rating of 2550 had to be achieved three times within a three years period. A >percentage of 55% had to be scored against the GM opponents faced in the 24 >games. Performance rating is 2600, three norms within a five year period. You need to achieve certain points in the tournament, but there is no requirement that one needs to score certain points against the GM's. You can lose your games to them, win the rest and if it is enough you get the norm. > If someone has this sort of information for present FIDE GM title >qualification, will you please make it available on CCC? > Second, someone suggested data giving the FIDE rating which present day GMs >had at the time the title was awarded would be useful in determining whether the >programs were performing at GM level. If you have the GM title and you know what >your FIDE rating was at the time your title was awarded, or you know of such >information about present GMs, please share it with us here on ICC. No need for that, a player has to reach 2500 with three norms. There are players who were 2600's before they became GM (very rare, GM Gelfand was the last one), but all of them had a rating of 2500 (published or reached within a period; January rating of 2470, reaches 2505 say in March, and the rating is 2490 in the June list, s/he becomes a GM) in order to be a GM. > In 1978 Arpad Elo gave the average rating of the GMs who had been awarded the >title during the years 1971 to 1977. That average was 2518 for 67 GMs. It was >not the average rating when they were awarded the title, but was probably fairly >close. Most of them would probably have increased their rating just a bit from >the time of the award to the year 1978. > Does anyone think FIDE has this latter information. Would they cooperate if >they do have it? There was somebody who had a lot of information on FIDE ratings, and published them on the web. Unfortunately FIDE said that they would sue him if he did not get it off his site. I may find his address. I doubt that FIDE would cooperate. They have become such a mess. They even did not realize a player from Romania who ended up becoming a GM, increase his rating over a 2-3 year period without playing a game. Or the Myanmar affair. Prof. Elo stated that 30 games would be a good estimate, and mentioned 80 games IIRC as solid estimate. We need to have players who will play 30 games with the computers, maybe three different Scheveningen tournaments with 10 players of whom 4 are GM's and an average rating of 2401 for all of the players. This is probable, since there are a lot of GMs who are not being paid much and play one open after another. They may be willing to play for a fee of course. We would end up with inofficial but valid FIDE ratings for the computers.
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.