Chess (#673)

Levon Aronyan took the 20th and last Amber title with a score of 15.5 points from 22 games. Carlsen was second with 14.5, and Anand, third with 13. Grischuk and Ivanchuk tied for 4-5 with 11 each. Everybody else had minus scores. Aronyan took the blindfold with 8.5 (from 11) while Carlsen scored a record 9.5 in the normal rapids. It's a pity this unique event is ending. One hopes somebody will take up the cause of blindfold even if it's unlikely that another sponsor will be as open-handed as Oosterom.

Vladimir Potkin leads the European Chps with 6 from 7 rounds. He's followed by 19 players who are tied at 5.5 each. There's plenty of time for twists and turns in an 11-round Swiss, given 76 GMs rated above 2600.

The ratings system itself has been the subject of much debate. In the 1960s, Arpad Elo sought ways to assess strength; he wanted to compare players across generations and to predict performances. Of necessity, his system had to be simple.

Elo works well and it is the foundation stone of the entire structure of tournament chess with titles, appearance fees, pairings, etc, built upon it. But it has flaws. For one thing, steady rating inflation makes historic comparisons dicey. Another problem - when there are extreme rating differences of 400 points or more, the formula has errors versus actual results across millions of games.

Increased computing power makes it possible to implement much more complex formulas. In late 2010, DeLoitte sponsored a contest to develop a new, more accurate rating system. Contestants received "training" datasets of 65,000 games with player names and results redacted. Their rating system were tested against another dataset, and contestants were allowed to tweak parameters.

Yannis Sismanis, an IBM researcher won, ahead of some 250-odd teams from 41 countries. As many as 39 systems were notably more accurate than Elo. Yannis' system was 6.6 per cent more accurate in predictive ability. In the second stage, Fide is sponsoring a contest with the top finishers and offering a much larger dataset to find a new "practical chess rating system". One leader in this contest is "Pragmatic Theory" - the two-member combine that won $1million developing a movie-preference rating system for Netflix.

The Diagram, Black to Play (White: Ter-Sahakyan Vs Black: Potkin, Euro Chps 2011) is a good illustration of Potkin's aggressive tactical style. 25...b4! 26.axb4 Qd4 27.c3 Bxg5 28.Qxg5. The queen diversion to g5 sets up the following thunderbolt 28...Rxc3+!! The simplest line is the acceptance with 29. bxc3 30. Qxc3+ Bc2 Be4 31. Rd2 Qa1+ but white tried 29.Bc2 Rxc2+! 30.Kxc2 Rc8+ 31.Kb1 Be4+ 32.Ka1 Qxb4 33.Qe3 Rc5 34.Rd3 Rb5 (0-1).

Devangshu Datta is an internationally-rated chess and correspondence chess player