Komodo Coasts To Victory Over Benjamin

Once the world's highest rated chess playing entity, Komodo set the tone in the recent Man Vs. Machine odds match against GM Joel Benjamin by converting a magnificent positional duel in game one.

In the increasing odds that followed, Komodo demonstrated impressive match-strategy artificial intelligence. Komodo circled the wagons and drew the final three games to win the four-game match, held March 22-23, by a score of 2.5 to 1.5.

To attempt a fair match, each game in the match utilized a different type of odds.

1. Five Moves — Benjamin plays White. d4, e4, f4, and Nc3 have already been played.
2. Exchange and Move — Benjamin plays White. The b1 knight and the a8 rook are removed; the a1 rook starts on b1.
3. One Pawn and Two Moves — Benjamin plays White. e4 has already been played. Komodo plays without the f7-pawn.
4. Queen Vs. Two Bishops — Komodo plays White. Komodo starts without the queen. Benjamin starts without either bishop.

Komodo's human opponent, Joel Benjamin, is a three-time US champion. Readers may well remember him from his cameo in the movie, Searching for Bobby Fischer.

Benjamin is also renowned for his role as Deep Blue's mentor in the match against World Champion Garry Kasparov. The Deep Blue team credited Benjamin with teaching Deep Blue how a human thinks about chess.

Benjamin playing a practice game against Deep Blue. Photo courtesy of IBM archives.

Game 1 (Five-Move Odds)

This unique odds format (first tried against GM Hikaru Nakamura) was speculated to provide the human a huge advantage. However, it soon became clear that by keeping the position closed, the machine was able to acquire a playable position.

Benjamin acknowledged having practiced for this game, but sadly said, "It didn’t make too much difference; in the match Komodo played much slower and better than in my practice games."

Benjamin began patiently and seemed to want to sit on his advantage. That quickly proved impossible as Komodo's 10...b5!! produced more wows than Ian Rappoport's Twitter feed.

This stunning sacrifice reminds one of the Benko Gambit or IM Alvis Vitolinsh's creative sacrifices such as the following one in the Nimzo-Indian Defense.

After 10...b5, Benjamin's optimism waned. Komodo's counterplay was evident, and it began to apply pressure to Benjamin's many weak pawns. The mounting pressure on the isolated d4-pawn reminded one of the great victory by Jose Capablanca over Emmanuel Lasker in the 10th game of the 1921 World Championship Match in Havana, Cuba.

Annotations by Capablanca

Despite the pressure on d4, the principle of two weaknesses dictated that Benjamin could not be brought down by pressure on this point alone. Komodo appreciated this and opened a second front along the h-file.

Komodo's positional stalking at least pretends to be slightly more benign than that of its predatory namesake.

As the engine's rooks infiltrated, Benjamin was forced to capitulate and concede that Komodo had been dominant in the game. Benjamin observed that he had "only moved a piece across the fourth rank a few times."

Game 2 (Exchange Odds)

As game two got underway, Benjamin noted that he appreciated Komodo's even speed of movement, which allowed him to think on the computer's time.

This comment was reminiscent of an aspect of Lee Sedol's strategy in his solitary win against DeepMind's AlphaGo. Sedol used much of his thinking time in the early stages of the game and seemed to rely on the computer's thinking time later in the game.

Benjamin's thinking seemed well directed as he was actually able to increase his advantage over the first 20 moves of the game.

Despite his advantage, Benjamin felt stymied and, seeing no way forward, he gave the exchange back to reach a pawn up endgame. Sadly, the pawn advantage proved insufficient, and the game was eventually drawn.

Game 3 (Pawn and Two-Move Odds)

Benjamin cited this game as the best of the match saying, "Komodo found some nice resources to stir up trouble, and I did a good job to hold the draw after the momentum swung against me."

As in game one, Komodo's use of the b-pawn on the 10th move again stunned Benjamin as 10...b6!! offered more material for development.

To quote Benjamin, "Wow -- Is it just playing to completely freak me out. I mean what is that?! I don't... understand... anything... There are no words."

Befuddlement aside, Benjamin responded calmly.

It would have been easy to falter in the ensuing vast complexities. Benjamin did cede his advantage over the next few moves, but careful play allowed him to hold against the silicon monster.

Game 4 (Queen Vs. Two Bishop Odds)

Benjamin and Komodo co-creator GM Larry Kaufman both agreed that the best chance for a human victory was game five. No form of queen odds had ever been offered to the human before, and Benjamin was optimistic that he could win the game and draw the match.

GM Larry Kaufman: Komodo's mentor.

Unfortunately, game four proved to be a striking echo of game two. Benjamin repeated his opening excellence and emerged with his advantage intact.

Sadly, despite the extra queen, Benjamin could again see no way forward, and he decided to give back material to simplify into an endgame.

Excellent play afterward even netted Benjamin a pawn, but the resulting position proved a fortress. After a few further moves, the position began to repeat, and a draw was tallied.

Final Match Standings

Post-Match Interview with GM Joel Benjamin

Benjamin was kind enough to answer a few questions about the match.

Q: Did any of Komodo's moves or plans particularly surprise or impress you?

A: Constantly! Its ...c5 followed by ...b6 in game three especially through me for a loop. Mostly Komodo was able to anticipate my plans and stop them… before I realized I even had them.

Q: Would you be interested in a return match? Would you do anything differently in a return match?

A: Sure, I always enjoy these matches. Probably we would have somewhat different odds. Perhaps I might practice a little more, and allot more time to these practice games.

Q: I understand GM Kaufman felt the difficulty of the odds eased through the match. Do you have any thoughts on which odds were easiest for you to play?

A: The five moves was the hardest because the advantage can be fleeting. The other ones felt pretty comfortable, though the pawn and two moves got uncomfortably complicated in a hurry. But they were pretty fair handicaps.

Q: Do you have any thoughts to share with other GMs in future matches with Komodo?

A: Try to stay objective and not get psyched out. If the odds are fair, you have a chance to beat it. You’ll just have to hurdle several key obstacles along the way.

Q: As a member of the Deep Blue team, is there anything in the nature of computer play today (roughly 20 years on) that you think the Deep Blue team might have found surprising?

A: We were able to get our computer to make moves most people thought computers couldn’t play. There were things we couldn’t do at the time, but we still believed they would one day be possible.

Q: Do you have any hindsight you might share with go players who are now experiencing what we chess players experienced 20 years ago?

A: Computers have built-in advantages in calculating that we cannot possibly match. They haven’t taken away our understanding of chess and have in some ways enhanced it. Chess is still a great game for people to play and the same will be true of go.

All games with commentary from Benjamin are available for replay at twitch.tv/chess. Here is Benjamin's commentary on the first game.

Watch live video from Chess onwww.twitch.tv

If you like this  masochistic format (and who doesn't?!), be sure to check out IM Daniel Rensch playing Komodo in various crazy blitz odds games on his #ChessMonday show Man vs Machine. Watch every other Monday at Chess.com/TV.

A Man vs. Machine historical archive summarizing previous matches is available here.