Grandmaster tournaments and matches are much more varied today than they were throughout most of the 20th century. The “old-fashioned” events had slow time limits and players rarely had to play more than one game a day. More recent tourneys are often played at a much faster overall pace, frequently have sudden-death blitz playoffs after a relatively slow start, and may involve computers or humans-plus-computer as entries. One of the most interesting new varieties has consistently attracted the top grandmasters in the world to its venue – the annual Amber tournaments in Monaco or Nice during March or April. There the contestants play two games a day with a single opponent, one at a rapid speed (25 minutes for the entire game, with a bonus of 10 sec for each move made) with a standard chessboard and pieces to move in front of them, and the other at basically the same speed with both players “blindfolded”, in the sense that they enter their moves on a computer keyboard but can see only a blank chessboard and their opponent’s last move on the monitor facing them.
The Amber tourneys allow an eventual comparison of each player’s world ranking at blindfold chess with his or her ranking in rapid chess or in chess at the traditional slow speed (“classical”, FIDE-rated games). Would the FIDE rankings of grandmasters correlate best with their blindfold play or with their rapid play, and would players’ rankings in blindfold and rapid chess differ significantly? Supposedly obvious predictions about these correlations might prove false if data were available to test them.
Elmer Sangalang of the Philippines volunteered to calculate ratings based on the 2,376 games played in the rapid and blindfold modes over all the 18 Amber tourneys that started in 1993, including the most recent event in March of 2010. Sangalang was the editor of the 2nd edition of Arpad Elo’s “The Rating of Chess Players, Past and Present”, published in 1986, which extended and corrected material in the first edition. Now retired, Sangalang worked mainly as an engineer, actuary, and applied mathematician. He has been a consultant for FIDE on the ELO rating system since 1984.
It was not an easy job to collect complete scoretables for every Amber tourney but ultimately Sangalang was successful and he could include all games from the blindfold and rapid halves of those events. On the other hand, FIDE ratings appear regularly every 2 months and he waited for the publication of the May 1, 2010 ratings and rankings to have the most recent results available for his analysis.
His method for calculating the Amber rapid and blindfold ratings followed the standard ELO procedure (Method of Successive Approximations). The calculations began by assigning every player an initial rating of 2600, to keep the numerical values completely independent of players’ different FIDE ratings. Starting with the players’ actual FIDE ratings seemed less reasonable and would bias the results in favor of the more highly-ranked individuals. So all the numerical ratings for the three groups presented below (Blindfold, Rapid, and FIDE) are independent of each other and cannot be compared in terms of their numerical values, that is, one cannot conclude that, say, Anand’s FIDE rating of 2789 means that he is better at slow chess than rapid chess (rating of 2688) or blindfold chess (rating of 2667). However, the rankings of the players (from 1 to 29) have no such limitations or restrictions and a comparison of these in the three groups is entirely justified. To increase the statistical reliability of the results, only players who participated in at least two Amber tourneys were included below, a total of 29 competitors.
Here are the results for the three types of play. We reiterate that each of the three sets of data are independent of each other, and the numerical values of the ratings cannot be legitimately compared. Before looking at the results, readers might like to guess, for example, whether FIDE rankings would correlate best with rankings in blindfold play or sighted rapid play.
|Ranking||Name||Number of |
|20||Vallejo Pons, Francisco||4||2531|
|25||Van Wely, Loek||12||2503|
|Ranking||Name||Number of |
|20||Van Wely, Loek||12||2534|
|22||Vallejo Pons, Francisco||4||2515|
|16||Vallejo Pons, Francisco||2703|
|21||Van Wely, Loek||2653|
After all the above rankings had been tabulated, statistically-determined correlations were calculated for each of the three possible pairs of comparisons: Blindfold vs. Rapid, Blindfold vs. FIDE, and Rapid vs. FIDE. Somewhat surprisingly, the FIDE rankings correlated most strongly with the Blindfold rather than with the Rapid rankings, even though both the FIDE and Rapid results involved games played with sight of a chessboard and the Blindfold games did not. All the different correlations were highly statistically reliable, but the strongest one was between FIDE and Blindfold; the next highest was between FIDE and Rapid, and the weakest was between Blindfold and Rapid. For those readers who are familiar with correlational techniques in statistics , the FIDE vs. Blindfold correlation for player rankings was +.84, for FIDE vs. Rapid +.76, and for Blindfold vs. Rapid +.72.
It is intriguing to speculate as to why a player’s world ranking (FIDE) in regular, “classical” chess would correlate best with his or her blindfold ranking, rather than with his or her regular rapid play. We offer one possibility and we welcome other suggestions from readers: Players may well be more cautious or careful in blindfold play than in rapid play with sight of the chessboard and thus try riskier lines of play in the latter, leading to more variable outcomes. (Recall the advice of world-class blindfold players like Alekhine who recommended that one “keep it simple” when playing without sight of the board). The fact that in the Amber tourneys the correlation between the Blindfold and Rapid conditions was relatively low (+.72) would be consistent with essentially the same kind of argument. At any rate, and speaking more loosely, you can predict a grandmaster’s FIDE ranking better from his Blindfold ranking than from his Rapid ranking.
We thank Mr. Sangalang for his careful and extensive work making the above calculations. Readers with questions or critical comments should send them to him or us via the “Comments” boxes below this blog. All of them will be published and answered.