Kasparov crushes x3d fritz in game three to tie the match

Nov. 16, 2003 – A beauty opened the game and the game was a beauty. Miss New York City, Katie Horn, was at the New York Athletic Club to make the ceremonial first move of game three for world number one Garry Kasparov.

The singer and music educator had no trouble with "Knight to f3" to start the game. This was the same move that Kasparov used to begin game one of the match. Then there was a little confusion among the commentators when X3D Fritz didn't reply instantly.

The short pause wasn't due to technical problems but to give the photographers a chance to snap away at Kasparov and the VIPs. When such commotion only disturbs one of the opponents it's not fair to have the clocks running while it's going on.

[ Full analysis of game three ]

Garry Kasparov, X3D CEO Eliot Klein, Miss NYC Katie Horn

The stakes could hardly have been higher in game three. A win was the only way Kasparov could hope to win the match. Another loss and the fourth game would be only a formality. A draw and Kasparov's back would be against the wall in the fourth game with the disadvantage of the black pieces.

In some games computers play chess so well you wonder how humans have any chance at all against them. Then they have games like this one and you wonder how humans can ever lose! Garry Kasparov completely dominated X3D Fritz in game three to bounce back from his horrible blunder in the second game and even the score with one game left to play.

X3D Fritz looked completely confused almost from the beginning. The opening moves of the game created a closed position with very little active play for the powerful pieces. In such positions the human ability to make long-term plans becomes far more effective than the machine's ability to calculate variations.

Left: Kasparov enters, but it was X3D Fritz that slept through the game.

Chess programs like X3D Fritz look at each position and evaluate the best move by looking as deeply into the future as they can. A powerhouse like X3D Fritz running on an Intel quad Xeon 2.8 GHz machine like this one can "see" as deeply as 20 ply, or half-moves in a few minutes by searching over four million moves per second.

(We say "half-move" because in chess notation we record each pair of white-black moves as one move. E.g. 1.Nf3 Nf6 2.c4 e6 is two moves, but four ply or half-moves when considering how far ahead a computer is looking.)

Despite that incredible power, this is only a dent in the depth and breadth of the game of chess. In a typical middlegame position there might be 20 possible moves for one side. Then the other side will have 20, then the other side another 20... You can see that the number of possibilities quickly grows into the hundreds of millions. Of course most of these moves won't be good ones, so X3D Fritz's intelligent program sifts out the best lines and focuses on those.

With all of that, X3D Fritz plays fantastic chess, but only in most positions. There are still some chess positions that show that computers don't really play chess at all, they just simulate it very, very well! Game three was one of these.

When the center of the board is blocked with connected pawns the power of the pieces is severely muted. They can't break through the pawns on their own, other pawns must be brought into the fight in order to make any progress. This often requires a sequence of moves that goes beyond the depth of the computer's search before anything good starts to show up in its evaluation.

Game 3, position after 14.Nb3

This position, with X3D Fritz to move with black, is where the game left chess and stepped firmly into (bad) computer chess. A strong human would look for a way to get play against the massive pawn chain in the middle of the board. White is dominant on the queenside (left side) and Black has a space advantage in the center. Black's pieces won't be effective against the compact white pawns, so more pawns will be needed on the front lines asap.

All the commentators were shouting for something like 14...Ne8 followed by pushing the f-pawn with ..f5 and ..f4 to open lines on the kingside. But that would be a bit slow, and would also run into further blockading when White played g3 to control the f4 square. So Black would then need to push its g-pawn too, in order to rip open the pawns and counterattack.

That's where we come into the second problem X3D Fritz had here. According to X3D Fritz creator Frans Morsch, one of the pieces of knowledge in its program is that moving the pawns in front of its king is bad. In general this is true because those pawns protect the king from danger. Here, however, it is essential to move them to make any progress because White is crushing Black on the other side of the board.

So X3D Fritz was basically helpless. It couldn't see far enough to know that its king was in no danger or to see that White was going to push through his humble a2 pawn to victory. You don't have to be a Grandmaster to see this simple plan for White, you just have to be human! All X3D Fritz could do was play the bizarre 14...Bd6 and then move the bishop right back after 15.Rb1 Be7. That's the sort of thing that gives chess programmers sleepless nights and there were many more such moments to come in game three.

Kasparov kept things closed up tight. He would never want to open up the game when X3D Fritz's power of calculation could be decisive. The machine made useless move after useless move with its pieces, never touching its pawns at all. Kasparov easily warded off the occasional threat while consolidating his forces on the queenside. By the time he evacuated his king to safety "it was all over" according to Kasparov.

As the commentators – and Kasparov – had foreseen early on, White slowly but surely advanced on the queenside and only crashed through when his forces were at their maximum power. He had no need to hurry since X3D Fritz had no idea it was even in danger and continued to shuffle aimlessly on the other side of the board. Almost to the end its evaluation of the game gave White only a tiny plus score. This while all the Grandmasters had long since buried the poor silicon warrior.

The X3D Fritz team stepped in and stopped the carnage on move 45 in the position below. X3D Fritz's evaluation was still only around a pawn and a half in White's favor, far from a decisive advantage in its judgment. But the machine was showing no sign of figuring out what to do and the humans in charge of the machine wanted to put a stop to what had been a thoroughly humiliating experience.

Game 3, position after 45.Rb3. X3D Fritz resigns.

In the final position there is nothing Black can do about Kasparov's plan to double his rooks on the a-file and invade. Soon his extra b-pawn will promote to a new queen, or force the win of a decisive amount of material. For example: 45...Ne8 46.Rba3 Nc8 47.Nb4 Rab8 48.Ra8 Bg5 49.Rxb8 Rxb8 50.Ra6 Bd8 51.Qa4 Ne7 52.Ra8 Rxa8 53.Qxa8 and the b-pawn can't be stopped.

Smiles all round in the post-game press conference. Kasparov, GM Maurice Ashley, Paul Hoffman, GM Yasser Seirawan.

A contented Kasparov came down for the post-game press conference with a broad smile on his face. "I think we can see that computers still have much to learn from us!" he declared. "There are many, many positions on the board that computers just don't understand. It's a matter for our technique to create these positions where the machine is totally hopeless."

Kasparov added that he hadn't come in to game three to win at all costs, but to "play two more good games of chess from beginning to end" to recover from his terrible losing blunder in the second game. Noteworthy was the lack of comment from Kasparov on his X3D virtual reality surroundings. Not only is he getting used to playing top-level chess in this environment, but winning is a powerful balm.

When asked about his plan for the final game of the match on Tuesday, Kasparov said, "I hope I can play good chess. It's very hard to make any predictions, but I'm in a very good mood now! In a good mood, with fighting spirit, I hope I have a good chance."

Frans Morsch, the creator and programmer of X3D Fritz, was optimistic about his creation's chances in game four and resigned about the painful loss. "We've been here before, often. We know what the problem is, it's just that we do not know how to fix it. [In game four] we have white and more influence on the opening. It will be more difficult for Garry to lock up the position, so we have good winning chances."