One chess champion per laptop
New chess engines bring ancient game to unprecedented heights
Search in YouTube for “too weak, too slow” and you will find a video of two young men sitting across from each other at a small table, frantically moving carved tokens on a wooden grid and slapping a clock mercilessly. They are fighting each other to the death, with bravado and gusto, in one of the oldest battlefields known to the human mind: the chessboard. The cocky guy in the green shirt, with the looks of a Viking and the nose of a boxer, is a 22-year-old chap named Magnus Carlsen, who happens to be the strongest chess player to ever walk the earth. The other guy, at the receiving end of Magnus’ Muhammad Ali-esque taunts (“Too weak, too slow! C’mon! What, you wanna play?”) is his close friend and sparring partner, Grandmaster Laurent Fressinet.
Mean as he may sound, the awful truth is that Carlsen is right: Fressinet, and almost everyone else on the planet, is indeed too weak and too slow for him. None of us mere humans stand a chance against him: he is too fast, too strong and too accurate. Less than two months ago, he beat World Champion Vishy Anand in a match without losing a single game. Yet even Magnus, at the peak of his powers, refuses to meet one opponent in a match, notwithstanding the incessant pleadings from chess fans. That opponent is here in front of me as I type, quietly waiting for the champ to accept the challenge. Carlsen won’t budge, and is wise in doing so, because — as he and all other Grandmasters know — even he himself is too weak and too slow to stand a chance against this opponent. In case you haven’t figured it out yet, the opponent I’m talking about is my laptop… and yours!
Back in 1996, when most of the current MIT undergrads were still in the process of being potty trained, Gary Kasparov — the Carlsen of the previous generation — lost his first game ever against Deep Blue, a top-secret, multi-million dollar supercomputer that IBM built using thousands of chess-specific processors with the sole purpose of defeating this one individual. Today, thanks to the increase in computing power of the average computer and to the appearance of a new generation of chess software, there is no longer need of specialized hardware to beat the best human: any decent laptop would maul World Champion Magnus Carlsen in a match.
At the core of this new software are algorithms that evaluate chess positions and calculate variations in order to decide on the best move. These algorithms are called chess engines, and — among the myriad currently available — at least two dozen have an estimated playing strength (or ELO) higher than the best human ever. Seldom have humans reached the rarefied stratosphere of speed and precision where these chess engines fight. Arguably the strongest chess engine of all is Houdini, developed by Robert Houdart. Since its appearance in the chess world back in 2010, Houdini has been widely regarded as the best chess player ever in the long and rich history of the game. A new version is released every year, and each one exceeds the previous one. The current release, Houdini 4, is the de facto gold standard against which all other chess engines are measured.
Humans are not real competitors against Houdini, and for a long time, neither were other engines. But now, for the first time in four years, something else has reached Houdini’s level of play. Not one, but two engines have risen with a legitimate challenge to the alpha dog: Komodo and Stockfish. Together, they made headlines in chess circles when they obtained the two highest scores in the most recent Thoresen Chess Engines Competition (TCEC), above Houdini and many other engines. TCEC is regarded, against the wishes of its organizer Martin Thoresen, as an unofficial world championship for chess engines. The TCEC final between Komodo and Stockfish was a very dramatic event, for many reasons. Not only was it the first time such a final did not feature Houdini, but also it was being played at the same time as the human World Chess Championship was being decided in Chennai, India.
At least for this author, the TCEC final was more exciting than the human championship. Stockfish qualified first for the final without losing a single game, while Komodo qualified second with no loses to Houdini. After a long and hard-fought series of 48 games, Komodo won the final by a narrow margin over Stockfish. In a dramatic twist, Komodo’s main programmer, Don Dailey (who once worked at MIT) didn’t get to see the triumph; he died of acute leukemia on the same day the final started. His partner in the development of Komodo, Grandmaster Larry Kaufman, an MIT alum himself, dedicated the victory to Don. Together with Komodo’s new programmer, Mark Lefler, Larry released the victorious version of Komodo under the name Komodo TCEC, to a fan base that was clamoring for it like teenagers in line for a Justin Bieber concert. In a touching gesture of gallantry and admiration, Stockfish’s team, led by developers Tord Romstad, Marco Costalba, and Joona Kiiski, also released the runner-up version of Stockfish under the moniker “DD,” an open homage to Don Dailey.
The three-way rumble between Houdini, Komodo and Stockfish in the latest TCEC season revived enthusiasm in computer chess and spurred a sort of arms race, as the teams behind each engine prepare stronger versions for the next TCEC, due to start in late January. Larry Kaufman, regarded as an expert on how to evaluate positions in chess, has thoroughly “taught” Komodo the ropes. As a result, Komodo’s play in long games is rock solid, earning praise from world-class experts such as Boris Avrukh and Roman Dzindzichashvili. Not to be outdone, team Stockfish is leveraging the power of open-source to summon the creative power of a dozen developers spread around the world, and is constantly testing new ideas using a cloud-based network of computers volunteered by Stockfish fans, who like the fact that Stockfish is offered free of cost and its code is available online.
There are a lot of expectations from thousands of fans around the world: Will Komodo retain the crown? Will Stockfish prove the superiority of the open-source model? Will Houdini regain and retain the top spot it enjoyed for so long? The answers to these questions remain to be seen. But, regardless of the next TCEC’s result, the clear winner in this new rivalry is the world of chess, which has now access — even for free — to top-class chess analysis tools that were thought impossible just a generation ago.