2/26/03 - BCE #144

I've added number 144 to the BCE section, another battle of bishop vs. pawns. In one of the notes, Fine rejects an alternative drawing method. The note is a bit strange since if it was so bad for the White King to go to the b-file then instead of 3...b5, Black could make White take the pawn with 3...b6+ 4. Kc4 b5+ 5. Kc5 b4 etc.

2/23/03 - Man vs. Machine

The latest man vs. machine match concluded a couple of weeks ago in New York with world #1 Garry Kasparov and world computer champion Deep Junior ending in a 3-3 tie. Since the conclusion of the match we have heard from both sides and I think it is a good time to reflect on what it all means.

First, the games themselves. Unlike in his celebrated 1997 match with IBM's Deep Blue, where Kasparov tried to play the so-called "anti-computer" style, this time Kasparov remained true to himself. This was evident in game 1 when after 1.d4 d5 2.c4 c6 3.Nc3 Nf6 4.e3 e6 5.Nf3 Nbd7 6.Qc2 Bd6 instead of quite moves such as 7. Bd3 or 7. Be2 he lashed out with Shabalov's sharp gambit 7.g4 . Getting into a tactical slugfest is not supposed to be how to play against a computer but by the time the computer was out of its book, it was clearly in a heap of trouble and Kasparov scored a fine victory.

Game 2 was a small turning point. Kasparov played the Sicilian (although for some reason he avoided the Najdorf) and got a nice position out of the opening. In the middlegame he sacrificed an exchange for an attack, but this was cut short when he reversed his intended move order and allowed Junior to sacrifice its queen to obtain perpetual check.

Game 3 was the major turning point in the match. Again Kasparov played very agressively with White with another early g4 against the Slav. After about 15 moves, Junior's position looked very suspicious, but remarkably there was no knockout for Kasparov. Kasparov maintained an attack as compensation for a pawn, but in time pressure he miscalculated and instead of going for a repetition, he blundered and lost.

This seemed to really throw Kasparov off form for the next game. He again had a promising position, but this time, instead of going for several promising looking pawn breaks, he decided to keep maneuvering. Eventually, Junior found a plan, and Kasparov had to struggle to the draw

Game 5 was the shortest of the match, but also the best. After 1. d4 Junior's programmers apparently decided that the Slav wasn't getting the job done, so they switched to the Nimzo-Indian 1...Nf6 2.c4 e6 3.Nc3 Bb4 4.e3 0-0 5.Bd3 d5 6.cxd5 exd5 7.Nge2 Re8 8.0-0 Bd6 9.a3 The last two moves are somewhat confusing to me, althought they have been played before. Black retreats his bishop without being kicked by a3 and then White kicks the ghost of the bishop with a3. 9...c6 This is a somewhat rare position. I only had one correspondence game in my database that had reached this position. There, White continued 10. Ng3 where I had an additional couple of games that got there by transposition. Kasparov 10.Qc2 seems to be a new move, and I think it is safe to assume that Junior was on it's own when it unleashed the stunning

10...Bxh2+!? Usually when a computer makes this sort of sacrifice it is accompanied by some announcement of mate in 15 or something. Here, it looks not only totally speculative, but a first sight, just unsound. I would have loved to have known what was going through Kasparov's head at this moment. I gave the diagrammed position to Shredder and asked it to show its top 10 moves. At depth 13, Bxh2+ doesn't even show up. (Shredder thinks Black is slightly better and its top 10 moves are Qc7, Nbd7, Na6, Be6, a6, h6, a5, Bc7, h5, and b6). After 10...Bxh2+ Crafty likes White by nearly 2 pawns at depth 13. 11.Kxh2 Ng4+ 12.Kg3 Qg5 After a couple of forced moves, Kasparov began consuming enormous amounts of time and finally bailed out for a draw 13.f4 Qh5 14.Bd2 Qh2+ 15.Kf3 Qh4 16.Bxh7+ 16. g3 was the way to try and play for a win, but it is very double-edged. 16...Kh8 17.Ng3 Nh2+ 18.Kf2 Ng4+ 19.Kf3 Nh2+ 1/2:1/2

Hansen's recent book on the e3 Nimzo

I think the results of games 3 and 5 weighed heavily on Kasparov's mind in the final game. It looked like he was going to lay it on the line when he finally trotted out the Najdorf variation. Junior surprising played a quiet variation with 6. Be2 and was clearly out of its element. Kasparov obtained a model Sicilian position after some very strange moves by Junior. I don't really know how to comment moves like 11. Bf3, 14. h3, 17. g3, 18. Kh2. Kasparov played the stock sacrifice ...Rxc3, but had overlooked the idea Be3-c1-a3. This seemed to spook Kasparov and even though Black was still better, he probably couldn't ignore the memory of game 3 and the shock of game 5 so he offered a draw, which was accepted.

The media coverage for the event was outstanding. All of the major news sources, both online and offline, had numerous articles on the event. Personally, I enjoyed the commentary by John Fedorowicz on Chess FM. This was the first time I had listened to Chess FM, and I really like it. Unlike the chess servers, where you have to constantly be in front of the machine to read all the comments, here you can do something else or play over variations on a real board while listening. If you haven't checked out Chess FM, I would encourage you to give it a try for the Linares supertournament, which kicked off yesterday. The cherry on top of the sundae was the live television coverage on ESPN2 of the final game. Jeremy Schaap did a very professional job as the host, and GMs Yasser Seirawan and Maurice Ashley were also very good in their commentary roles. They also managed to recap much of what was currently going on in the world chess, so I think even people who only casually are familiar with chess would have enjoyed the coverage. It was only unfortunate that the game ended with an anticlimactic draw. I hope ESPN off of future events, I didn't hear how the ratings were.

I suspect many people were disappointed with the drawn result in the final game. What bothers me most about it is the way draws are decided by committee for the computer in these matches. I can understand that the computer is going to need help in deciding its opening book (the alternative would be to turn off the opening book entirely, but against a GM, I think most computers would just get crushed under those circumstances), especially so it wouldn't lose the same way in each game. However, deciding when to resign and when to accept or decline draws is another matter entirely. It was widely reported that there were very detailed rules for this match (although I haven't been able to find them on the internet anywhere), but I think some change is necessary. While most of the human commentators agreed that Black was better in the final position of Game 6, according to MIGs report Junior's evaluation was positive by about half a pawn and it was only the GM on Junior's team (Boris Alterman) who convinced the programmers to accept the draw. I saw time and again listening to the Fed's commentary, where he would say that Junior was just losing (in games 3 and 5 especially) and the computer proved him wrong. Of course, the computer does not make perfect evaluations, but if it thinks it is better by half a pawn (which is pretty significant in computer terms) then I think the game should have continued. In game 4, play went on much longer than it needed to because the computer evaluated in was a full pawn up. In that game, the position was even and the machine didn't show any plan to make progress so they eventually agreed to the draw. It's a delicate situation, perhaps they should just eliminate the draw by agreement in man vs. machine matches and make them play out positions like that in Game 4 to the 50 move rule. Or possibly only allow the program to agree to a draw when its evaluation is zero or negative. What especially bothered me in this situation is that GM Alterman is a much, much stronger player than the programmers. (Programmer Amir Ban gives some very interesting annotations at the Chess Base site. One of my favorites is from Game 1: "7. g4 Huh? What's this ? I've never seen such a move in a queen's pawn opening 7...dxc4 Book move. Okay! Boris is on top of this"). What happens in a position like this

The programmers see the large negative evaluation the machine is registering and want to resign. The GM steps in and points out that White can reach a fortress by 1. Rxf6+ Qxf6 2. Rxb4 followed by shuffling the rook back and forth between f4 and h4. OK, the machine would still have to find this line, but it certainly won't find it if the programmers stop the game. (If you are wondering where I got this position from, it is a possible variation from the end of the 6th game of the Kramnik-Fritz match)

Inevitably, comparisons with Deep Blue arise. Unfortunately, no such comparisons can be made. Deep Blue played 6 public games and was dismantled. No logs were ever made public. At this point in time, I would have to call the whole Deep Blue project a waste of time as far as computer chess is concerned. Amir Ban's light annotations of Deep Junior vs. Kasparov are much, much more than we ever saw from Deep Blue. Some might argue that Deep Blue was better because it evaluated more positions and won its match against Kasparov. On the first point, I'll only comment that quantity is no match for quality otherwise computers would be much better than humans at chess. On the second point, Deep Blue won 2 games vs. Kasparov. In one of those, Kasparov resigned in a drawn position (I think it is safe to say that Deep Junior would have won game 5 if Kasparov had decided to resign after Bxh2+ or do you think Amir Ban would have said "No, Garry, we decline your resignation, our evaluation reads zero"). The other game Deep Blue won was when Kasparov played a known opening error. Deep Blue had a won position without ever making a single move on it's own. Put Deep Junior or any other reasonably strong program or player on the White side after 8...h6? and I think they wouldn't have much trouble bringing the point home. I don't mean to belittle Deep Blue's achievement, chess is about taking advantage of the other guy's mistakes. Chess is also about proving yourself over the long haul. Six games just makes Deep Blue a tiny footnote in the history of chess.

2/8/03 - Land of the Sky

I played in the annual Land of the Sky tournament last weekend in Asheville, NC. Despite a considerably weaker field than in years past (I was seeded third in the Open section after GMs Wojtkiewicz and Becerra), I only managed to equal my 3-2 score from last year. Things got off on the wrong foot in round 1 when I lost with the White pieces to Daniel Tapia. I built up a close to winning position with an extra pawn, but left myself with much to little time (only about 10 minutes) to make the time control on move 35. After 21...fxe5

My first intention was to try to eliminate the bishop pair with 22. Ng5 threatening Ne6. Instead, I decided to control all the open files with my major pieces with 22. Rfd1 This is probably a poor plan since it takes away protection from the f2-square 22...exf4 23. Rxd5 Qe7 24. Bc4 and this move unprotects Nf3 24...Kh8 25. Re1? The correct idea is to keep Black's light squared bishop from coming to the long diagonal with 25. Nc5 Black can then force a roughly equal ending with imbalanced material after 25...fxg3 26. hxg3 Rxc4 27. Rxc4 Qe2 28. Rd2 Qxf3 29. Qxf3 Rxf3 25... Qa7? with 25... Bb7 26. Rxe7 Bxc6 White would have to shed at least an exchange 26. Nc5 Now, this idea is too late, better is 26. Qc5 with advantage to White. The problem now is that Black has b6 under control and was able to bring his rook into the attack after 26... fxg3 27. hxg3 Rb6 28. Qa4 Rbf6 with a big advantage. Plus, I was under a minute on the clock here. I managed to reach the time control, but at that point was dead lost and resigned. After being behind on the clock every round at Clarksville and now this debacle, I told myself that, if nothing else, I needed to play faster.

I bounced back with a quick win against Ronald Dennis in the next round. He made several mistakes in the opening and after 15. Nc3xBb5 (better was 15. Bxe5 Rxe5 16. Nxb5 Nxd5, but that position is no fun for White either)

The floodgates poured open with 15... Nc4 when all of Black's pieces are working together beautifully 16. Qc2 preventing ...Qa5+ with 16. Qc3 exposes him to tactics along the long diagonal with 16...Nxe4 17. Qxc4 Nxg3+ 18. Kf2 Nxh1+ 19. Rxh1 Be5 -+ 16... Qa5+ 17. Nc3 Nxd5 18. Nge2 Nxf4 19. Nxf4 Nxb2 20. Nfd5 Na4 21. Rc1 Nxc3 22. Nxc3 Qxc3+ [0:1]

In round 3 I had a very interesting game with White against James Mottonen. He played a piece sacrifice early in the game, but it looked like it was only temporary after 12...Qxc6 attacking both my bishop and queen.

I had a very similar position to this a couple of years ago against Lee Maring. In that game, White had not yet castled, and the Black pawn on e6 was on g7. In that position Bd5 was immediately decisive. Here, the same move 13. Bd5 sets Black the most problems, but he still may be able to claim the advantage. 13...Bc5+? I expected instead 13... Qb6+ 14. Kh1 c6 15. Rxf5 exf5 16. Be6+ Kb8 17. Qf1 when I thought the 2 pieces for the rook and Black's damaged pawn structure gave White a small edge. Instead he has the interesting move 14...Bg4 with the idea of meeting 15. Rxf6 with 15... Bg7 16. Rf4 Bxc3 17. bxc3 Rxd5 (17... Bxe2 18. Bxb7+) White can regain his pawn after 18. Qe1 Bf5 19. Ng3, but with the bishop vs. the knight and the damaged White pawns on the queenside, Black should be much better. White has various other possibilites, but I haven't really found anything where Black doesn't come out with at least an extra pawn. 14. Kh1 exd5 15. Rxf5 If he could hold his pawn structure together, Black might have compensation, but the only way to save d5 here is 15...d4 but White has the simple 16. Nxe4 Qxe4 17. Rxc5 with an extra piece.

The next morning I had Black against Aleksandar Stamnov. I dropped a pawn against his middlegame attack, but thought I might be stirring up some resistance with my passed d-pawn in the ending after 36. a5

36...Qd5 possibly 36...Qd4 was better. At least it carries the threat of 37...Re2. Instead, I played to pressure his queenside pawns. 37. Rc3 Rd8 A better choice was to try to resist in the ending after 37...Qxb5 38. Rxd3, but this ending should be pretty easy for White. Instead, I had miscalculated and after 38. b6 Qxa5 39. b7 Black is lost in all variations 39...Kh7 39... Qb4 40. Rc8 Qxd2 41. Rxd8+ Kh7 42. b8=Q covers f4; 39... Qa7 40. Rc8; 39... Qb6 40. Rc8 Kh7 41. b8=Q; 39... Rb8 40. Rc8+ are all straightforward for White 40. Rxd3 and White soon won.

In the last round I had White against Todd Trawer. I think I had a solid edge in a queenless middlegame with 2 bishops and a queenside pawn majority. I wasn't completely happy with all my last half dozen or so moves before the diagram since his pieces had become much more active than they had been and had put my queenside into a small bind. Still, after 25. Rc1 I think White is still quite comfortable.

My plan was to use potential pins on the c-file to get the queenside pawns moving. He helped me out with 25...Nc5?! 26. Nxc5 Rxc5? 26... Bxc5 27. Bd3 and Black is going to have problems holding his a-pawn. 27. Bb4 Rc7 28. b3 Nxa3 28...Bxb4 29. axb4 and he is still pinned. 29. Rxc7 better is 29. Bxa3 with a whole piece. 29... Bxb4 30. Bd3 He resigned after we reached the time control at move 35 although neither of us was in time pressure. I thought it was a bit strange since it wasn't completely clear to me how to win. Black's pieces are completely in a bind, but if, for example White plays his king to b2 and rook to d1 intending Ra1xa3 Black can answer with Be7 stopping Ra1 because of Bf6+. I don't doubt that White is winning. Most probably White can use zugzwang to pick off one of the kingside pawns and then create a pased h-pawn, but I still thought he might play a few more moves.