Komodo Beats Nakamura In Final Battle
It was the battles of the Alamo, Thermopylae, and Hoth all rolled into one, and the result was the same — valiant resistance but ultimate and seemingly inescapable defeat.
Chess.com's Man vs. Machine contest pitted U.S. number-one GM Hikaru Nakamura (Elo 2787) against the most powerful chess-playing entity of all time, Komodo (Elo 3368) in a four-game odds match.
One could not have asked for a better representative of humanity as Nakamura swashed and buckled his way to three difficult, but ultimately confident draws in the first three games of the match.
It was only in game four that he faltered as Komodo acquired its best position of the match, and Nakamura found himself unable to hold back a tide of positional pressure.
Komodo won the game and the match (2.5 - 1.5) after a lengthy 58-move struggle.
GM Hikaru Nakamura -- AKA The Chosen One.
Nakamura is no stranger to heroic performances against chess engines. In 2008, he won a much-celebrated blitz game against the top engine of the day, Rybka, by exploiting the engine's failures in closed positions, and its desire to win at all costs. Those interested in a history of Man vs. Machine battles can find a summary by Chess.com's Peter Doggers here.
These days the top chess engines are considerably stronger than Rybka was in 2008, and most easily exploitable failures of chess understanding have been eradicated.
Komodo -- the brainchild of Don Dailey (who died in November of 2013), GM Larry Kaufman, and Mark Lefler -- is now universally recognized as the strongest chess-playing entity on the planet. It eradicated any question to that effect by defeating the number-two engine, Stockfish, by a seven-game margin in the 2015 TCEC Superfinal.
The 2015 TCEC Superfinal produced a 53.5 - 46.5 match victory for Komodo over Stockfish.
For this match, Komodo played on a 24-core, Xeon machine run by Kaufman.
No human can currently hope to survive on equal terms in a match against Komodo. Odds must be given. In this match, Komodo offered Nakamura pawn and move, pawn, exchange, and four-move odds. All games were played at a time control of 45 minutes with a 15-second increment.
Komodo's chess guide, GM Larry Kaufman (photo from www.uschessleague.com).
In game one, the pawn deficit did not fluster Komodo as it rapidly sharpened the play on move eight with ...c5. By move 15, Komodo was three pawns down, but its pieces were raging. Things looked hairy for Nakamura when he suddenly uncorked the piece sacrifice 24. Nxg5!? saying in the chat, "If I'm gonna lose so be it, but I'm gonna take every pawn I can."
The sacrifice proved to be more than mere desperation as Nakamura navigated his way to an elegant positional fortress in which Komodo couldn't snag Nakamura's final pawns without allowing its own pawn to be captured.
Game 1 (Pawn and Move Odds) — All evaluations by Komodo 9.
If game one was difficult, game two was twice as challenging. The odds were the same, but now Komodo had White!
This was arguably Nakamura's best-played game of the match. Despite the diminished odds, he maintained his advantage past 20 moves into the game, and at no point was Nakamura truly worse.
Komodo made a winning attempt by racing its king to the center and creating a frightening passed d-pawn. However, Nakamura returned his king at just the right time with 48...Kg8! and forced a draw eight moves later.
Game 2 (Pawn Odds)
Nakamura demonstrated extreme precision in game two.
Nakamura noted that "[Game 3] was my best chance to win. I felt that even before the match started."
In the opening, Nakamura outplayed Komodo and increased his advantage to nearly a full two pawns. While he soon returned the exchange, he retained the better position and commenced an attack on Komodo's king that culminated in the piece sacrifice 30...Ng4+!
Komodo had to return the piece a few moves later with 34.Nc2, and Nakamura could have maintained the attack with 34...Bxe3+!! Instead, he regained his sacrificed knight, and Komodo was soon forced to take a draw by giving a perpetual.
Game 3 (Exchange Odds)
Komodo seemed a bit sleepy in game three.
The four-move odds caused general consternation among commentators and chess fans alike. Are such large positional odds more or less favorable to the human than material odds?
Many fans thought these were the largest odds offered, but GM Robert Hess put his finger on the problem in commentary. If Komodo played a closed King's Indian-esque position, the extra moves may mean little.
This is exactly what Komodo did. In the closed position, Nakamura could not convert the extra tempi into anything tangible.
When Komodo played f5 on move 20, everyone knew it would take a magnificent effort to draw. Nakamura defended heroically with only-move after only-move for 30 moves, but on move 50 Komodo's advantage crept to decisive levels and shortly thereafter Nakamura resigned.
Game 4 (4-Move Odds)
The match proved to be hugely popular with casual and serious chess fans alike as it set the record for most live chess viewers of a single game in Chess.com history with 1,800 chess players watching game one simultaneously. Thousands more tuned in via Chess.com/tv and twitch.tv/chess to watch commentary provided alternately by GM Simon Williams, GM Robert Hess, and GM Alex Yermolinsky.
Particularly interesting were the post-match interviews with both Nakamura and Kaufman.
Final Match Standings
Participants | 1 | 2 | 3 | 4 | Final Score |
Komodo (3368) | ½ | ½ | ½ | 1 | 2.5 |
Nakamura (2787) | ½ | ½ | ½ | 0 | 1.5 |
While Nakamura ultimately fell to the machine; he revealed that humans can play at a very high level against computers. His understanding of the fortress was superior to Komodo's in game one, and he outplayed the machine for much of game three. Still, he never came close to a victory.
This begs the question, who is willing to take up the fight against Komodo and finally bring home a coveted match win?