After 12 draws, Magnus Carlsen is once again the chess world champion

news image

RARELY HAS a tournament involved so little winning. The World Chess Championship of 2018, held in London at The College in Holborn over nearly three weeks from November 9th-28th, began with a series of 12 games played under classical time controls, the traditional slow pace of play. The reigning champion and top-ranked player in the world, 27-year-old Magnus Carlsen of Norway, failed to win a single one of his contests against the challenger, 26-year-old American Fabiano Caruana. Fortunately for the Norwegian, Mr Caruana never reached a checkmate or extracted a resignation either. Each of the dozen classical games ended in a draw, sending the match to a series of faster-paced tiebreakers, starting with a series of up to four “rapid” games, in which players are allotted less than one-quarter the thinking time of the classical format. Mr Carlsen, the stronger competitor in speedier formats, won the first three rapid games to clinch the tiebreaker and retain his title.

To the casual observer, three weeks of drawn games may sound excruciatingly boring. (In the first game alone, the two men played 115 moves over seven hours before reaching an impasse.) But like a football World Cup full of impregnable defences, or a baseball World Series studded with scoreless pitching duels, the chess title match featured two equally matched grandmasters competing at an extremely high level. A game of chess opens in a state of equilibrium, and if the optimal move is made with each play, a draw is all but assured. For a player to win a game, he must capitalise on his opponent’s blunders. Neither man played perfectly—a supercomputer identified a guaranteed, though extremely opaque, path to victory for Mr Caruana in the sixth game, and Mr Carlsen held a sizeable advantage in the final classical game before surprisingly offering a draw—but mistakes were rare.

Indeed, the frequency of draws is a strong indicator of the overall level of play. Among beginners, roughly one in five games end deadlocked. At the elite, grandmaster level, more than half of contests are drawn, a figure that has remained steady for decades. The governing body of chess, FIDE, uses an algorithm called Elo to evaluate players, boosting a competitor’s rating after victories and dropping it after losses, with the magnitude determined by the opponent’s own rating. Among the highest-rated players—those above 2,700 on the Elo scale, a category that currently counts around 50 members—63% of classical games end in draws. Mr Carlsen and Mr Caruana entered their championship match atop the ranking table, with ratings a hair’s breadth apart at 2,835 and 2,832, respectively, and had drawn 18 of their 33 career games against each other.

Mr Carlsen has drawn games at an even greater clip when playing on the sport’s biggest stage. This year’s World Chess Championship was the Norwegian’s fourth, after he claimed the title from Viswanathan Anand in 2013, successfully defended it against the same opponent in 2014, and withstood Sergey Karjakin in 2016. Of the 35 games in the three previous championships (the match was decided after 11 games in 2014), 26 ended in draws, a rate of 74%. Not only are the competitors in these matches among the very best in the world, they also spend months preparing to parry the preferred tactics of a single foe.

Computer chess engines play a significant role in that preparation, and they help explain the increasingly precise play of the world’s best. In 1997, a machine defeated the top-ranked human player for the first time, when IBM’s Deep Blue overtook Garry Kasparov. In the intervening two decades, engines have gotten smarter and processing speeds have increased exponentially. When Deep Blue made its breakthrough, it edged out a player rated by Elo in the 2,800s. Today, Stockfish, the highest-rated engine, is rated above 3,450, more than 600 points above Mr Carlsen, a gap that implies the machine would defeat the human nearly every time. (Google’s AlphaZero engine isn’t rated, but it is far stronger. In December 2017, it demolished Stockfish in a 100-game match, winning 28, drawing 72, and losing none.) While grandmasters cannot use the engines during play, they use computers to evaluate openings and develop strategies tailored to specific opponents.

The result is human chess that increasingly resembles computer chess. Chess.com has developed a metric, Computer Aggregated Precision Score (CAPS), which compares a player’s moves to those of a top engine. Mr Carlsen scores higher than any other champion; his moves parallel those of the engine’s more than 85% of the time, and his overall choices are better than 98% as effective as those of the computer. At this year’s championship, Mr Caruana was nearly as good. A smaller-scale approach to measuring the level of precision relies on the concept of centipawn loss, the difference between a player’s move and an engine’s optimal alternative. (One centipawn is one-hundredth of a pawn, a unit that represents the effects of strong or weak positions on the board.) In the 2018 championship, as well as the drawn games in Mr Carlsen’s three previous title matches, the average centipawn loss was less than 10, a mark that had been achieved only a handful of times in over a century of chess championships. By the same metric of centipawn loss, the gap between the competitors’ performances was the smallest ever in a title match.

But even if an increased draw rate represents a positive step for the quality of play, it doesn’t present the sport at its most gripping at the one time that the world is watching. The tiebreaker, which took place in a single day, made for much more engaging viewing: three speedy games, one after another, and no draws. The faster time controls of rapid chess prevent grandmasters from carefully choosing every move, and the limitations drove Mr Caruana into mistakes that he had avoided in the previous 12 games. Among rapid games between players with ratings of 2700 or higher, only 47% of meetings result in draws. In the even quicker “blitz” format, which was slated to be the second tiebreaker had the competitors drawn even after four rapid games, the draw rate is lower still, at 34% of elite-level games.

Still, the World Chess Championship determines the classical chess title, and it may be a bit unfair to deny the American the honour due to his lapses in a different format. But a decision had to be made somehow, and the suggestions of purists that the match be extended to 16, 18, or more games seem impractical. At the level of Mr Carlsen and Mr Caruana, a draw is the most likely outcome in any single contest, and both deserve ample credit for playing well enough to prevent the other from breaking through. Such acknowledgements are surely little consolation for the American. But as the early favourite to challenge the Norwegian once again in 2020, Mr Caruana now has another 15 head-to-head games’ worth of information to take with him into his preparation to unseat the champion next time.

Read More

Leave a Reply

Your email address will not be published. Required fields are marked *