Monday, August 22, 2011

Intrinsic Chess Ratings

Forbes recently had an article on chess, 'Humans Are Getting Better At Chess - Thanks to Computers' (, written by Alex Knapp. His story begins, "A recent study suggests that there are more great chess players now than there ever have been – and that players continue to improve."
Mr. Knapp throws in a thought of his own when he writes, "The authors don’t suggest a mechanism, but if I were to guess, I’d suggest that the reason for this has to do with competitive chess software and online play." Mr. Knapp thinks that if humans continue to 'interface' with programs, " may be that humans catch back up to the best AI programs." I will not live so long...
I clicked on 'recent study' and found a PDF titled, 'Intrinsic Chess Ratings', written by Kenneth W. Regan of the University of Buffalo and Guy McC. Haworth of the University of Reading, UK. This must be the former chess player, IM Ken Regan. I put him into a search engine and found he has a chess page: Gotta be the same guy.
The first sentence lays it out: "This paper develops and tests formulas for representing playing strength at chess by the quality of moves played, rather than by the results of games."
I wondered who was to be the ultimate arbiter of "the quality of move played." My question was answered by the next sentence: "Intrinsic quality is estimated via evaluations given by computer chess programs run to high depth, ideally so that their playing strength is sufficiently far ahead of the best human players as to be a 'relatively omniscient' guide.
There it is. We have reached a point in human evolution whereby 'puters have now become our 'relatively omniscient' guides. I could not help but think of the first paragraph of an editorial by Charles M. Blow, 'Obama in the Valley' dated August 19, 2011 (
"In 1970, the Japanese roboticist Masahiro Mori coined the phrase “uncanny valley.” In the field of robotics, and increasingly in computer animation, it refers to the theory that people feel good about robots — up to a point. When they start to look almost real, but not quite, we experience an eerie and unsettling sense of revulsion."

In the introduction the authors say, "Our work brings new evidence on controversial questions of import to the chess community, with ramifications for skill assessment in other games: 1) Has there been 'inflation'-or deflation-in chess Elo rating system over the past forty years? 2) Were the top players of earlier times as strong as the top players of today? 3) Does a faster time control markedly reduce the quality of play? 4) Can recorded games from tournaments where high results by a player are suspected as fraudulent reveal the extent to which luck or collusion played a role?
The paper, especially page seven, contains a plethora of equations. If you had showed it to me and told me they were equations for a magnetoplasmadynamic propulsion system from aliens from planet Zud, I would have said, "OK". Fortunately, they sum it up by saying, "We conclude that there is a smooth relationship between the actual players' ELO ratings and the intrinsic quality of the move choices as measured by the chess program and the agent fitting." Got that? It continues, "The results also suppost a no answer to question 2. In the 1970's there were only two players with ratings over 2700, namely Bobby Fischer and Anatoly Karpov, and there were years as late as 1981 when no one had a rating over 2700 (see Wee00). In the past decade there have usually been thirty or more players with such ratings. Thus lack of inflation implies that those players are better than all but Fischer and Karpov were. Extrapolated backwards, this would be consistent with the findings of (DHMG07), which however (like some recent competitions to improve on the ELO system)are based only on the results of games, not on intrinsic decision-making."
Whoa! Has the level of chess play risen to the point that ALL 2700+ players of today are better than ALL players of a generation ago other than Bobby Fischer and Anatoly Karpov? That is saying a great deal.
I recall GM Andy Soltis writing a column whereby he went back a century and compared the moves of the greats of that era with the players of today and found that the old masters made more blunders, and, frankly, I would tend to put more credence in what a HUMAN GM has to say than a COMPUTER.
I cannot help but wonder what a computer analysis of USCF ratings would conclude. Has there been inflation, or deflation? Back in the 70's the average ratting of a tournament player in the USCF was around 1500, give or take. (I believe that was where Prof Elo based his 'average') Today the average rating is around 900. I submit to you that the 'average' player of the 70's was vastly superior to the 'average' player of today. Most players knew then that, if a player made class 'B', he had stopped dropping pieces and could play a decent game of chess. In the first round of the 1980 US Open in Atlanta, a class'B' player upset GM John Fedorowicz in the first round. Consider this game played by two players in the top half of all USCF rated players at a G/30 played here in Louisville at the Monday night tournament where the score is not kept and the game is played without a clock. One of the players is my student. I will not say which to protect the guilty... 1 e4 e5 2 d4 f6? 3 d5?! I will spare you the rest...
I tied for first in the Atlanta Championship in 1974 with a score of 4-1 as a class 'B' player. It was said at the time by many that, "We have a 'B' player as champion!" At the time I was playing actively and working on my game, which needed much work, let me tell you! Although an adult, I was one of those players whose rating had not caught up with with his strength. I won the 1976 Atlanta Championship with a score of 5-0 as a low 'A' player. At that time ratings were not as up to date as they are today. One time the rating system went down for almost a year. There usd to be something called 'bonus points', then they were elimanated. I recall crossing the expert line, 2000, during a time of no bonus points, and my friend, the Legendary Georgia Ironman, Tim Brookshear, took the time to figure what my rating would have been if the bonus points were still being awarded. He came to the conclusion I would have obtained a rating of over 2100. "And you did this while swimming against the tide!" he said. I mention this to give some perspective and to say that, as good as I was, I was nowhere near the level of the top players of today, such as Georgia champions NM Richard Francisco, NM Damir Studen, and FM Kazim Gulamali. At that time I was maybe 500 or 600 points higher rated than the 'average' player. A player today rated 500 or 600 points higher than the 'average' USCF player would be rated 1400 or 1500! What does that say about the current rating pool?

No comments: