Coming into the last round of the Alekhine Memorial, Boris Gelfand led Viswanathan Anand, Levon Aronian, Mickey Adams and Maxime Vachier-Lagrave by half a point. Gelfand faced Anand in the last round, played it safe with the white pieces, and Anand drew without much trouble. That eliminated one rival and made it such that no one could catch him unless he or they won their games.
Unfortunately for Gelfand, Aronian defeated Vachier-Lagrave in a generally impressive game, finishing with some very nice tactics. Through the first 28 moves, everything was going smoothly, and had Aronian played 29.d6 he would have been well on his way to a clean victory. Instead, 29.Rxb8+ was a mistake, and after 29...Rxb8 30.Rxa7 Bxc3! 31.Rxd7 Rb4 Vachier-Lagrave had reached an objectively drawn position. This is a pure abstraction though, and would remain so as long as White's d-pawn was alive and dangerous. After 32.d6 Rxc4 33.Be7 some care was required.
Black's best move would have been 33...f6, immediately eliminating all the dark-squared mating nets around Black's king and allowing it to join in the fight against White's d-pawn. Maybe there's a sly trap both players thought was winning, but I'm not seeing it. Instead, it just looks like a relatively straightforward draw, e.g. 34.Ra7 Kf7 35.Bd8+ Ke6 and now White seems to have nothing better than 36.d7 Rd4 (36...Rf4 first is an interesting finesse, threatening mate starting with 37...Bd4+. White plays 37.g3 and only then Black's rook goes to d4. The point is that after the same moves given in the 36...Rd4 line, the presence of a White pawn on g3 makes it easier for Black to liquidate the kingside and draw. Remember that he can give up everything he has for White's kingside pawns, his bishop included, and then draw with his king parked in the a8 corner.) 37.Bb6 Rxd7 38.Rxd7 Kxd7 39.Bxc5, when White's outside passer won't give him any serious winning chances.
Vachier-Lagrave played 33...Kg7 instead, and while it wasn't losing it kept him in danger. For one thing, it keeps the king away from the d-pawn; for another, it doesn't yet save the king from possible mating nets. After 34.Ra7 Black had to play 34...Re4!, and it still seems that he should hold the game. White can promote: 35.d7 Rxe7 36.d8Q Rxa7, but it appears that Black has a fortress, despite the presence of White's a-pawn. Of course if it's exchanged for Black's c-pawn the result is a dead draw, so let's see what happens if White tries to keep it: 37.Qd5 Bf6 38.Qc4 (threatening to start making progress with a2-a4) 38...Ra3! Now White's only winning idea is to bring the king over to b1, so the queen can go to c1 to push Black's rook away. (And even that is just a first tiny step.) This plan is incredibly slow, however, and Black has many ways to deal with it - just pushing the kingside pawns, for instance, easily generates sufficient counterplay.
Unfortunately for Vachier-Lagrave (and Gelfand and their fans), but fortunately for Aronian (and his fans and for those of us who can appreciate the aesthetics of his winning combination), Black played the losing move: 34...Rd4(?). It looks like an obvious blunder, but Black had a nice trick in mind. After 35.d7 Rd1+ 36.Kf2 Vachier-Lagrave played 36...c4!, a move with not just one but two points. The first, obvious point (though not so obvious when you have to think it up several moves in advance) is that if Aronian promotes (36.d8Q??) then Black saves the game with 36...Rxd8 37.Bxd8 Bd4+ and 38...Bxa7. But the really brilliant point was that if Aronian had played something obvious like 36.Rc7(?) Black has a de facto perpetual check! 36...Rd2+ 37.Kf3 (37.Ke1?? Rxd7+ and it won't be a perpetual; Black will simply win) 37...Rd3+ 38.Kg4 h5+ 39.Kf4 (39.Kh4 Bf6+ 40.Bxf6+ Kxf6 41.Rxc4 Rxd7=) Rd4+ 40.Ke3 Rd3+ etc.
But Aronian was up to the challenge, and played the only clear winning move: 37.g3! This eliminates the perpetual, allows Black to play ...Bd4+ if he wants (which he doesn't, as 38.Ke2 Bxa7 39.Kxd1 is a trivial win). The remaining moves were pretty simple, and in the final position White plays 43.Ra8 and starts pushing the a-pawn, with our without first interpolating Rc8.
The reason all this was bad news not just for Vachier-Lagrave but for Gelfand as well is the same reason why the last round of the London Candidates was a triumph for Magnus Carlsen despite his loss: tiebreaks. The same one, in fact, that cost Kramnik in London: it was number of wins of that determined the official winner of the event: Carlsen then and Aronian now. As then, and even before then - I've expressed similar complaints going back to the introduction a few years ago of chess tournaments with 3-1-0 scoring - I object to privileging a win and a loss over a pair of draws. Going +1, -1 doesn't show that someone played better chess or more enterprising chess than the player who drew twice; it doesn't even necessitate more fighting spirit. (Look at some of Kramnik's events here, or Nakamura's marathon draws in Zug.) It's impossible to discern anything about a game's quality just by knowing that it was drawn. What one does know, however, is that decisive games contain mistakes. So we know that the player who went +1 -1 made at least one error, and we don't know that his win was of particularly high quality. Maybe his opponent made a gross blunder in a perfectly good position.
It's also true that the draws might have been fightless and dull while the decisive games were dazzling and daring. It could be, but the point is that we don't know this a priori without looking at the games, and the games hadn't been played when the decisions about tiebreaks were being made. I'd prefer to skip out on tiebreaks altogether, either having co-champions when possible or a playoff when necessary. If tiebreaks are necessary though, I'd propose eliminating the "most wins/losses" tiebreaker or putting it much further down the list. (And why isn't head-to-head the first tiebreaker? It wouldn't have affected anything here or at the Candidates', but when it is relevant how is that not the most obvious and natural way to distinguish the players? Still another idea, aiming for objectivity over the kinds of dumb luck rewarded by the Sonneborn-Berger tiebreak: what about factoring in something like Ken Regan's Intrinsic Performance Ratings, both for the player's moves and his opponent's? It's not perfect, but it at least tries to isolate the most relevant factor: the quality of a player's moves.)
Rant over. In other games, Kramnik was successful today where he wasn't yesterday, this time winning the 7-hour game. Adams (his opponent) was doing fine for a long time, but a couple of loose moves between moves 30 and 40 got him in trouble. 33.Ne2 would have been better than 33.Nf1, but the bigger culprit was 36.Nd5? Adams must have missed or underestimated Kramnik's 36...f4! It's antipositional and ugly as sin, but it sets up the threat of ...c6, exploiting the knight's lack of squares to win the b-pawn. (Note that ...c6 needed to be prefaced by ...Be5, as White could have met 37...c6? with 38.Nf6!=.) From there it was a long, hard grind, and while Kramnik in general handled the ending extremely well and was a deserved winner, he seems to have erred on moves 61 and 65. I'm not 100% sure that Adams could have drawn even then, but at the very least Kramnik endangered the win.
In both cases Adams returned the favor; the first with inaccuracies on moves 62 and 64; the second with 70.Rf5. I'm not sure Kramnik is winning after the immediate 70.Rf8, e.g. 70...Rd3+ 71.Kc2 Re3 72.Nf6+ Kg6 73.Ng4 Rg3 74.Ne5+. White's setup is incredibly effective: the f- and g-pawns are frozen and the poor Black king can't go to its otherwise ideal square, h5, on account of Rh8#. The h-pawn has a little freedom, but it's limited. Continuing a bit: 74...Kg7 75.Rf7+ Kh6 (75...Kg8 76.Kd2 h3 77.Rf6 eventually comes to the same thing) 76.Rf8 h3 77.Kd2 Kg7 78.Rf7+ Kg8 79.Rf6 may just be drawn. When Black's king goes to the 7th rank, White plays Rf7+ and then goes to f6 or f8 - whichever rank is opposite Black's king. If there is a win in there, it's not easy to find. Anyway, Adams missed this chance and played 70.Rf5(?), after which Kramnik only had to find the simple but nice finesse 70...Rd3+! and only after 71.Kc2 Rg3. With the king on d1 White could capture and draw, but with the king on c2 it's an elementary win for Black. Adams played a few more moves, and then resigned.
Nikita Vitiugov and Ding Liren slugged it out in the Anti-Saemisch Gambit line of the King's Indian. For a while Vitiugov looked like he would be able to keep the material and win, but he never quite figured out how to extinguish his opponent's activity and the game finally ended in a draw. There are various improvements available to White, but the last chance to keep winning chances was with 32.Rc1 rather than 32.Rd1. After 32.Rd1 Rc4 followed by doubling on the 2nd rank, the game was equal. The difference is that if Black goes for the same plan with 32.Rc1 Rd4 White has 33.Rc6. In the 32.Rd1 Rc4 version, 33.Rd6 is ineffective due to 33...Bd4+, when White is lucky that he can still draw. In the 32.Rc1 Rd4 version, 33...Bd4+ is illegal, so White is winning. 32...Rd4 isn't forced, but White can still fight for the full point.
Finally, the game between Peter Svidler and Laurent Fressinet also finished in a draw. Fressinet was better most of the way and probably could have pushed a bit more, but in general it looked like the players were happy to vacuum up the board and draw at move 40 - which they did.
1-2. Aronian (first on tiebreaks), Gelfand 5.5
3. Anand 5
4-8. Vitiugov, Fressinet, Kramnik, Adams, Vachier-Lagrave 4.5
9. Ding Liren 3.5
10. Svidler 3