Halfway through last year’s Qatar Masters Open, 23-year-old American Sam Shankland found himself across the chessboard from Gadir Guseinov, a fellow grandmaster from Azerbaijan. Shankland quickly advanced his pawns and knights. Guseinov, 28, paused for long stretches to think, but each exchange only made his situation worse. He gave up on the 38th move.
Guseinov may have appeared outclassed, but Shankland won the game before it had even started. His secret: “Computer analysis, looking for new moves in well-known positions,” he said.
Computer-processing power has advanced so rapidly that today’s grandmasters have pretty much given up beating even general-purpose PCs at chess. But recently, computers have become indispensable training tools for top players.
The Shankland-Guseinov game exemplifies technology’s pervasive role in a centuries-old pursuit. When Guseinov put himself in a position Shankland had researched, Shankland knew exactly what to do.
Into the Computer Age
Playing at a world-class level in chess involves strong concentration, visualization, pattern recognition and memory. These are areas where computers excel.
In the 1980s and 1990s, chess was a great testing ground for high-profile “man vs. machine” matches. Eighteen years ago, IBM supercomputer Deep Blue made headlines by eking out a win against world champion Garry Kasparov.
Using 30 parallel processors, Deep Blue was an 11.38-gigaflop computer, meaning that it could handle about 11 billion operations per second. A Mac Pro with a six-core Intel Xeon E5 processor, by contrast, performs 2.2 teraflops, or about 2 quadrillion operations per second.
Today, even a run-of-the-mill chess app on a smartphone is strong enough to routinely beat anyone below master level.
An average chess program on an average computer can evaluate millions of moves per second, according to Kenneth W. Regan, a computer science and engineering professor at the University at Buffalo (SUNY). Shankland estimated that he can evaluate about 150 chess moves per minute.
Machines are “wipe-the-floor-with-us better,” said Regan, who has been researching computer-aided cheating in chess for the last nine years. “No grandmaster has challenged a computer on even terms since 1995.”
Man and Machine
Computers’ general chess-playing dominance over humans masks an important distinction between the two parties. Computers are excellent at evaluating positions where tactical considerations are paramount. But the best grandmasters are still stronger at evaluating certain strategic positions.
In chess, strategy is more like literary analysis — something computers don’t do as well.
Tactics, however, refer to sequences of moves involving concrete threats like capturing pieces or attacking the king. The answer can be calculated — and computers are great at calculation.
Shankland said this explains “why computers can just demolish grandmasters.”
To help prepare for tournaments, grandmasters employ a team of human analysts, or “seconds,” to analyze common opening sequences with the help of chess software.
Shankland’s team uses a free open-source chess program called Stockfish.
Before competing against Guseinov, Shankland recalled, he had assigned one of his seconds to analyze moves to play against the Dragon variation of the Sicilian defense, an opening setup in which the arrangement of black’s pawns resemble the mythical beast.
The team member set the initial position of this defense within Stockfish, then switched into analysis mode.
Chess programs often identify powerful moves known as novelties that have not been played in recorded grandmaster games. During weekly conference calls, Shankland’s seconds present potential novelties, and his team tries to poke holes in the analyses.
Ensuing discussions, Shankland said, help him memorize opening powerful sequences like the one he used to combat Guseinov.
“By move 20 [of that game],” Regan said, “Stockfish said Shankland had a big advantage, and it persisted throughout the game.”
Could Shankland have known in advance that Guseinov would play the Dragon variation? Not for certain, but he easily could have determined that it was one of Guseinov’s favorite setups.
In addition to analyzing opening lines in chess engines, grandmasters often search databases of recorded games such as ChessBase, which stores more than 6 million games.
They can research their future opponents’ preferred opening lines, as well as run statistics on various positions they’ve played, to identify their stylistic tendencies and flaws.
There is some truth to the old sports platitude “to be the best, you have to play the best.” Before the prevalence of sites that enabled online play, it was harder for talented young players to develop their skills against more experienced opponents.
“When I was kid in Brooklyn, if I wanted a game, I had to go to the park and play the hustlers,” said grandmaster Maurice Ashley, 49. He said the Manhattan Chess Club was only an option for those who could pay the membership fees.
“That’s how you cut your teeth,” He said. “Today’s kids can go online and have fifty options for finding a strong opponent.”
Talented players, Ashley said, can get challenging practice “any time, any day,” improving their game faster than ever.
According to calculations garnered from World Chess Federation rating lists over the last 15 years, the average age of the game’s top 100 players has dropped from 34 to 31. The current world champion, chess’ highest-rated player ever, is 24-year-old Magnus Carlsen of Norway.
Shankland, who was recently ranked 84, would love to reach that top level. With the help of his team, he searches for shortcomings in his game as diligently as he searches for those of his future opponents.
“Some people say you should just build on your strengths,” he said. “But my approach is to improve broadly. At this level, you can’t have any targetable weaknesses.”