Seirawan on Kramnik vs Deep Fritz game one

The chess duel Man vs Machine, Vladimir Kramnik vs Deep Fritz is being staged from November 25 until December 5th. It is sponsored by the RAG AG, one of Europe's largest energy companies. The venue is the National Art Gallery in Bonn, Germany. Schedule:

Game 1:	Saturday	25.11.2006	15:00 h
Game 2:	Monday	27.11.2006	15:00 h
Game 3:	Wednesday	29.11.2006	15:00 h
Game 4:	Friday	01.12.2006	15:00 h
Game 5:	Sunday	03.12.2006	15:00 h
Game 6:	Tuesday	05.12.2006	15:00 h

There will be full live coverage on the Playchess server, as well as on the official site and a number of partner sites.

Game one

Analysis by GM Yasser Seirawan

A few introductory comments to the latest Man versus Machine challenge. In the pre-match press conference, Vladimir Kramnik opined that because the computer hardware speeds were getting faster all the time, allowing programs to see an ever increasing number of ply, the window of opportunity for making such contests was closing. He hoped that it would not close on him!

Yasser Seirawan analyzing with the Fritz Team (Mathias Feist, left, and Alex Kure)

Finding a sure-fire win for Kramnik

Speaking on behalf of Deep Fritz 10, Matthias Wüllenweber, expressed an opinion I had not heard before. He was answering a question as to why Fritz 10 had not entered into the latest computer competitions such as Turin 2006? Matthias explained that preparing for a match against the world champion as opposed to a computer opponent was not the same thing. That a lot of preparation was necessary as the opponent was so different! The programmers live in fear that the human player will shut down the position and lock up the pawn structure. The program, unable to calculate long forcing sequences, plays without understanding and the programmers are humiliated. Such a situation is really a problem when playing Vladimir Kramnik but is not a problem when playing against computers. Against computers a different kind of preparation was necessary and the team of Fritz 10 felt it was best to use their preparation time for the human match and not a computer challenge. It was quite an irony to hear this side of a familiar discussion amongst humans.

Now to game one. In his matches with Deep Blue, Garry Kasparov was quite cagey with the White pieces. In order to avoid certain defensive lines he would play: 1.Nf3, trying to steer the computer into the type of (closed) game that he was aiming. In contrast, for game one at least, Vladimir just gets on with business.

Vladimir Kramnik - Deep Fritz
Man vs Machine (1), Bonn 25.11.2006 – Catalan Opening [E03]

1.d4 Nf6 2.c4 e6 3.g3. Repeating the Catalan Opening which he used to good effect in his recent match with Veselin Topalov. In principle, an excellent opening choice against the computer as the strategic positions that arise are extremely subtle. Without making an obvious error the Black side can be put under long term pressure that continues even into the endgame. On the other hand, there is much to be said for: 3.Nc3 Bb4 4.Qc2, just planning to win the two Bishops and hope that the computer will allow a slow buildup thereafter. Happy thoughts of his last round win over Peter Leko in Dortmund 2006, is the type of ending Vladimir might try to repeat.

3...d5 4.Bg2 dxc4! While there is nothing wrong with: 4...Bb4+ 5.Bd2 Be7, and an eventual ...c7-c6 playing the Closed Catalan, there are dangers present in these lines. White might get a queen-side clamp and the computer might not realize that a certain moment a central break is required. By the text, the computer insures at least a semi-open game and the danger of being dominated in a closed position passes by at an early stage.

5.Qa4+. In general, most Catalan players are not afraid of losing the c4-pawn and often prefer not to recapture it at once. However, sacrificing material against a human is one thing, against a computer Vladimir might never get the pawn back.

5...Nbd7 6.Qxc4 a6 7.Qd3. Up to this point, Vladimir could see the computer's monitor as the players were still in "book" or rather opening theory. After the text, Deep Fritz or Fritz 10 was out of book and the monitor was turned away from Vladimir. However, a few moves later, the game had transposed back into opening theory and Vladimir could again observe the computer's monitor. The text sidesteps a quick ...c7-c5 and ...b7-b5 which would come with tempo. White's early Queen moves aren't particularly effective in terms of playing for an opening advantage.

7...c5. Annoying. Okay, the text is a perfectly normal move but Vladimir could hope that his opening novelty might induce the computer to take another approach such as playing: 7...Bd6, aiming for ....e6-e5 which is not a particularly good idea. 8.dxc5 Bxc5 9.Nf3 0–0 10.0–0 Qe7 11.Nc3.

Thus far the game has followed along what could be described as normal channels but the first real question has been laid before Deep Fritz 10: Would it play 11...b6 or 11...b5 and why? I'm not sure which of the two is best. On the one hand an argument for 11...b5 can be made that Black takes control of the c4-square and keeps the b6-square open. Also, after 11...b5 a future ...b5-b4 might be a useful tempo. On the other hand, a future a2-a4 might undermine the Queenside. A difficult choice and an intriguing moment into the insight of how Deep Fritz 10 plays.

11...b6. 12.Ne4! An excellent moment to clear the board of some pieces. White cannot become complacent. It is clear that after ...Bc8-b7, Black will bring a rook to the d-file and suddenly White could find his queen harassed. The text preempts the potential for such problems. Naturally, 12.Nd4? Ne5!, would only result in problems for White.

12...Nxe4. Further insight into how Fritz 10 plays. The text shows a prejudice for Bishops. A reasonable alternative was: 12...Bb7 13.Nxc5 Nxc5 14.Qa3, when White's Queen is safe from further harassment. White has won the two Bishops but Black's harmonious development gives a comfortable game as well. 13.Qxe4 Nf6.

14.Qh4! Following the KISS principle (Keep It Simple Stupid). Against a human player the line: 14.Qxa8 Bb7 15.Qxf8+ Kxf8, with two rooks for the Queen deserves consideration. After all the c and d-files are both open and offer good avenues for the rooks. However, if we continue the line several ply further for example: 16.b3 e5! 17.Bb2 e4 18.Ne5 e3!, the human player starts to feel uncomfortable. There are some dynamic factors lurking in Black's favor and the rooks have hardly had a chance to flex their muscles.

One possible thought to consider is whether Deep Fritz 10 had seen this line and judged that in case White opted to win the two rooks for a queen, that in this case, a supporting b6-pawn to the c5-Bishop was better placed.

14...Bb7 15.Bg5 Rfd8 16.Bxf6 Qxf6 17.Qxf6 gxf6 18.Rfd1.

On the one hand, Vladimir can be pleased. He has managed to channel the game into the quiet waters of an ending. Best of all Black has doubled f-pawns which gives rise to a certain optimism. On the other hand, Vladimir has no advantage whatsoever. Vladimir is "gambling" that Deep Fritz will not play the correct move...

18...Kf8? The gamble is successful. Most human players would have played 18...Bxf3! and offered the world champion a halving of the point. Vladimir correctly understands that the computer would look at the wonderful open-range b7-Bishop, see the puny short-hopping f3-Knight and refuse a trade. While Deep Fritz can see 18 ply deep and even more in certain positions, it is unable to deduce that the f3-Knight has the potential of being a superior piece to the c5-Bishop.

19.Ne1! After this move, White has a serious endgame advantage. Of course from this moment to the end of the game Deep Fritz 10 was showing that the game was always equal or perhaps at most a 0.18 advantage for White. In actuality, of course, White's winning chances are excellent.

19...Bxg2 20.Kxg2 f5 21.Rxd8+ Rxd8 22.Nd3 Bd4 23.Rc1. In my day, players like Ulf Anderson (Sweden) and Zoltan Ribli (Hungary) made an excellent living from such endgame positions as White. Time and again they would demonstrate that White's control of the c-file was more important then Black's d-file control. That Black's king-side structure was a serious liability and that given time; White's King could play a more active role in the endgame then its counterpart.

23...e5! More annoyance. The computer refuses to go quietly into that good night. White just wanted to play moves like f2-f4 and Kg2-f3 to really get his positional grippers latched on to the position. The text interferes with this delightful dream by creating counter-play based on ...e5-e4 and also keeps White's King at bay.

Kramnik ponders the position after the move 23...e5!

24.Rc2! Simple chess. Vladimir covers the b2-pawn as well as protecting the d2-square against any invasion. He now awaits developments, perhaps Deep Fritz, with a wide range of pawn moves would weaken its position?

24...Rd5? Whoa! This hyper-active move was certainly a surprise. I was having some doubts about White's winning chances after 24...a5, preventing a possible Nd3-b4-c6 idea. In that case, 25.e3!? e4 26.exd4 exd3 27.Rd2 Rxd4 28.f4! (28.Kf3 f4! 29.gxf4 f5), leads to a likely drawn rook endgame but White after a subsequent Kg2-f3-e3 and vacuuming up the d3-pawn might have something to enjoy. After all the doubled f-pawns have some kind of disadvantage. After the text, I really got excited about Vladimir's winning chances. Now he can force a classic good knight versus bad bishop minor piece endgame with excellent winning chances. A situation that would be well beyond the computers horizon.

25.Nb4! Rb5. In passing, I should note that my Fritz 9 engine rated the text as well as 25...Rc5, as about even. Needless to say, after: 25...Rc5 26.Rxc5 bxc5 27.Nxa6 c4 28.b3 c3 29.Nb4, White, a pawn to the good, wins easily. 26.Nxa6.

Suddenly, there are many pitfalls in the position for Black. For example: 26...Bxb2 27.Nc7 Rb4 28.Nd5 Rb5 29.Ne3 f4 30.Nf5, with both the threat of a back-rank checkmate as well as a plan of Kg2-f3, when we see an example of how White's King can become far more active then its counter-part. Another nice trapping line could be seen after: 26...Bxb2 27.Nc7 Rc5 28.Rxc5!? bxc5 29.Nd5 c4 30.Ne3 c3 31.Kf1, when White's King races to blockade the passed c-pawn.

26...Rxb2 27.Rxb2 Bxb2 28.Nb4. This is one of those moves that I find difficult to explain. At the time, I was convinced that Vladimir should win by: 28.Nc7!, in order to "freeze" the Black King and to quickly bring the knight to the wonderful d5-square. My arguments were quite convincing, at least to myself. Then Vladimir played the text and I was left wondering what the difference was? Now, I'm unable to recall. 28...Kg7 29.Nd5 Bd4.

Thus far we can only express our admiration for Vladimir's sterling anti-computer play (and in general his fine play overall). He is literally at the precipice of victory while Deep Fritz 10 is happily producing 0.00 evaluations. The game is won and this is the moment. At the time of play, I thought the win was simple: White should play a2-a4, e2-e3, march the King to the b5-square and make sure that Black had insufficient opportunity to penetrate on the king-side with his King. Now, of course, I realize that the move a2-a4 is not critical to this line of play and that it might even be helpful if Black can be coaxed into playing ...b6-b5 bringing the pawn closer to its own execution.

30.a4? Missing the direct path to victory. After: 30.e3! Bc5 31.Kf3!, the win is as clear as a sunny sky. White's King is making a beeline to the b5-square, where the b6-pawn is captured and the a-pawn is escorted to coronation. I see no defense: 31...f6, prepares to bring the Black King to the center. Black is welcome to march his King to the g5-square when h2-h3 shuts the door to further progress. White continues: 32.Ke2, executing his plan. Now Black has a choice. He can play: 32...e4, delaying White's King march by a move or not. Keep in mind that: 32...e4, gives up the f4-square, let us judge: 32...Kf7, as the main line. White continues his march: 33.Kd3 Ke6 34.Kc4, and White has achieved his ideal. Black is in Zugzwang. His Bishop cannot leave the protection of the b6-pawn and Black's King is fastened to the protection of the f6-pawn. White's win is simple: he just plays a2-a4 with the threat of a4-a5 when the Bishop is undermined. White wins.

Because this main line is so convincing, Black would likely have to try a line combining ...e5-e4 and even ...b6-b5 as well but I view the ending as hopeless. Let us check: 30.e3 Bc5 31.Kf3 f6 32.Ke2 e4 33.Kd2 b5 34.Kc3 Kf7 35.Nc7 Bb6 36.Nxb5, let us pause to give thanks. White has achieved our favorite living condition, he is a pawn ahead. Black can continue with: 36...Ba5+ 37.Kd4 Be1, however, this counter-play is far too late. White ignores the loss of the f2-pawn and with: 38.a4, Black can give it a rest.

30...Bc5 31.h3? This is a real groaner. I think the text throws away most of the winning chances that were left. I was still wildly optimistic about White's chances after 31.Kf3! One point is to trick the computer into playing an innocuous move like: 31...Kg6? After which White plays: 32.e3!, and he is back in business as his King goes to the b5-square as we have seen. If Black plays: 31...e4+? 32.Kf4 Kg6 33.Ke5 gives White the active King and a simple win. The main question is: 31.Kf3 f6 and can White still win? Black is now prepared for ...Kg7-f7-e6 booting away the overpowering d5-Knight. My attention had been focused on: 32.e4, inviting the line: 32...fxe4+ 33.Kxe4 Bxf2 34.Kf5, playing for the simple recapture of the f6-pawn. White appears dominant. Therefore, 31.Kf3 f6 32.e4 Kg6, when we come to the major difference between this line and the game continuation, White can provoke Black into playing ...f5xe4 under favorable circumstances: 33.Ne3!, now after: 33...fxe4+ 34.Kxe4, White has achieved what he wants, an active King. While after: 33...Bxe3 34.Kxe3, White has a winning King and pawn ending.

31...f6 32.f3 Kg6.

33.e4?? The final unbearable mauling of this favorable ending. Although my optimism was seeping away I still felt that Vladimir had not tossed away his whole advantage. Now was the time for 33.e3! awaiting developments. If Black falls asleep, then the plan of Kf1–e2-d3-c4 will force a rude awakening. White's dream is revealed after: 33.e3! h5 34.Kf1! Kf7 35.Ke2 Ke6 36.e4, now the difference in the game is crystal clear, Black's King on e6 is not offered any king-side counter-play: 36...fxe4 37.fxe4 f5 38.Kf3 Bd4 39.Ne3! fxe4+ 40.Kxe4, and White's King has reached the square of milk and honey.

33...h5! Whoops. Now White finds himself in Zugzwang. If his King was on e2, then Ke2-d3 and just wins. Unfortunately, White's King has to stay close to the e4-square as: 34.Kf1 fxe4 35.fxe4 f5, and White can't maintain his light-square blockade. Vladimir now went into a long pause and was horrified to realize that he had done the one thing that he was trying hard to avoid: Leaving a king-side pawn weakness.

34.g4 hxg4 35.hxg4 fxe4 36.fxe4 Kg5 37.Kf3 Kg6 38.Ke2 Kg5 39.Kd3 Bg1!? A bit of a surprise. Most of us would have played: 39...Kxg4 40.Nxf6+ Kf3 41.Nd7 Bd4, and awaited a draw offer. 40.Kc4 Bf2 41.Kb5 Kxg4 42.Nxf6+ Kf3 43.Kc6 Bh4 44.Nd7 Kxe4 45.Kxb6 Be1 46.Kc6 Kf5 47.Nxe5 Kxe5 ½–½.

Click to replay and download the analysis

The game nears its end

Kramnik takes the last black pawn with 47.Nxe5 Kxe5...

...and then offers the draw.

The post-game press conference was an unusual experience. Vladimir was asked if he felt he had realistic winning chances and he answered that he thought so but was not sure where he had gone wrong, whereas the Fritz team thought the game was equal throughout. At least I could still comfort myself with the thought that, yes, humans can still compete with computers and the window has not been fully closed.

Vladimir Kramnik and Mathias Feist, the operator of Deep Fritz in the post gam press conference