Kamsky-Svidler What looked like a coincidence now seems to be a consistent pattern in Kamsky's play! Once again Gata produced a painfully long and tense game. By the end, his opponent was dying of exhaustion, while Gata remained fresh and firm. Peter played very creatively after the opening. His 'knight mill' 24...Nd8!, relocating the knight from с6 to с5, deserves the highest praise. Black lost a pawn, but created strong pressure on e4. Later, however, he incorrectly traded the light-squared bishop. Instead of 27...Bxf5?! Black could continue playing on equal terms by 27...Qb7 or 27...Ncxe4. The text-move gave Black some freedom in the center, but allowed dangerous advance of White's b-pawns. Soon Peter had to give up an exchange. A very interesting spot occurred on the 38th move. Trading the queens on b4 led to a difficult ending. Black could get more saving chances if he dared to sacrifice a knight by 38...Nxf2! I do not say it guaranteed a draw, but the struggle would be a lot tenser. The game could continue 39.Kxf2 e4 40.Qc7! Qxb4, and here, in order to play for a win, White is obliged to find the refined 41.Rb1!, brining the rook to b7 or b8. However I seriously doubt that grandmasters would play all the computer moves exactly - in human chess everything is possible. After Black missed this chance, Kamsky methodically converted his endgame advantage. Svidler built up a solid position with the pawns on e3, g5 and h4, which looked like a fortress, but Kamsky's mating threats dispersed this illusion. However, in the post mortem it turned out that Gata missed the neat 65...Nf2!, which could indeed construct an unbreakable fortress, but Peter was too tired to resist. Overall, Kamsky's victory in this game was quite logical.
Gelfand-Bacrot A head-spinning game. Both players deserve the highest praise. In the Queen's Indian Defense, Black constructed a pawn 'barb' in the center, but fell somewhat behind in development. Gelfand decided to use the latter fact. His opening revelation 14.g4! led to enormous complications, which after incredibly precise play of both sides resulted in a complicated ending with White having three pawns for a knight. How they managed to find all the best moves, is almost inexplicable. This was a demonstration of the highest skill! In the mutual time trouble Boris missed an excellent chance. He had to play 38.Kd3! at once (without inclusion of h2-h4 and Ke5-f5), in order to win the game after the most natural 38...Nc5+ 39.Kc3 Ncxa4+? 40.Nxa4 Nxa4+ 41.Kb4 Nb6 42.Kc5 Na4+ 43.Kс6. Etienne would have to find the only defense 39...Nd5+! 40.Kc4 Nb7! with just a few seconds on the clock. In the actual game Black managed to stop all the pawns in due time.
Eljanov-Ivanchuk Another creative game, in which both players showed some non-orthodox ideas. Pavel stylishly solved the problem with his pinned f3-knight by 14.Qb3! Activity of White's pieces fully compensated for structural flaws. The pressure on Black's position began to grow... And then Vassily launched a nice saving operation. Black bravely accepted the pawn sacrifice: 20...Rxd4!, and then suddenly gave up a queen for a rook and a bishop. In the resulting position computer engines insist that White has a big advantage, but any human player immediately recognizes a fortress. Game drawn!
Leko-Karjakin The players initiated a joint research project on the Chebanenko Slav and came up with important conclusions. White's spectacular novelty 14.Nxd5 did not surprise Karjakin. Black's subsequent play is characterized by computer precision. I was unable to find any improvements for White in the express analysis. A bloody battle in the middlegame led to an equal ending, and Black made a draw from the position of strength.
Mamedyarov-Akopian The logical course of the game that developed around the d6-pawn was shaken by the sharp 18...f5!? This move came as a surprise for Shakhriyar, and the Azerbaijani grandmaster didn't find the nice reaction 19.h4!, keeping Black's minor pieces at the bay. The point is that 19...fxe4 20.Ne1! d5? doesn't work due to multiple attacks on d5 followed by deadly pins on light-squared diagonals. After the exchange on f6 the position simplified (via complications!), and the players entered a heavy-piece ending. The draw by move repetition was logical.
Aronian-Grischuk The Armenian grandmaster made an attempt to win the game on sole technique, without any risk. Many time he succeeded doing this, but... not against Grischuk. The Russian consumed a lot of thinking time, but found all the accurate moves that equalized the game. In the English Opening the players quickly traded the queens and several other pieces, ending up in a simple-looking ending with almost symmetrical pawn structure. White played a good novelty, suggested in Khalifman's opening books. It looks like Grischuk was unaware of it, but he managed to find the best moves at the board. Looking for possible improvements for White, I can suggest 22.Nf3!?, for example, 22...Kg7 23.g4!, and then, if necessary, h2-h4, the pawn goes to g5, and the knight establishes on e5. After White lost a tempo by 22.a3, Black built solid defensive line and held the position. Kasimdzhanov-Alekseev This long game was surprisingly free of obvious mistakes. Apparently, the position after the opening was approximately even, but Rustam overestimated the attack started with 24.g4. This kingside aggression only created problems for White. If Alekseev played the accurate 28...Bb7!, he would get reasonable chances to succeed. However, he went for total exchanges, and soon found himself suffering in a rook ending without a pawn. Yet, he managed to defend it, and the game ended in a draw.
Round 3 Review
Karjakin-Aronian
Everyone plays as good as the opponent allows to. In this game Levon failed to impress because Sergey was really impressive. White made a new move in a modern line of the Ruy Lopez -15.b3! Perhaps Black had to change his plan, keeping the king on f8. In this case he would have a chance to create some counterplay with his bishops: 15...Bc5 16.Bb2 Bg4!, which would compensate for the weakness of the e5-pawn. However, Aronian played in a more standard way, placing the king to e7, which led to big complications, and eventually lost a pawn. Karjakin converted his advantage in a refined fashion. One cannot pass by the excellent 24.Nb6!, which transposed the game into a rook ending. Soon Black missed his last chance to resist: instead of 27...Rb5 he had to counterattack on the second rank by 27...Rd2!, and then Rb8-e8-e2. In the game White managed to promote his pawn and defended against the perpetual. Grischuk-Kasimdzhanov I found many unexpected resources in the analysis of this complicated game. Kasimdzhanov employed a very risky but interesting plan in a well-known variation of the Queen's Gambit Accepted. I thought that after 14.Nxb6 Black had to take with the c7-pawn, keeping the king's shelter intact. After 14...cxb6 White cannot retreat the knight from e2 because of the d4-pawn weakness, while 15.Ba4 is well met by 15...g5! with strong counterplay for Black. However, Rustam selected 14...axb6, allowing the opponent to establish deadly-looking pressure against the c6-knight. The ideas behind this move were clarified by the extremely tricky maneuver 17...Qc8! - Black sacrificed a pawn for the initiative. The subsequent moves did not look forced. For instance, instead of 19...fxe5 Black could play 19...Nc6! Later Kasimdzhanov did not find the most accurate 25...Qa6!, which would allow continuing to play on equal terms. After 25...Qb7? Grischuk destroyed Black's defense by the powerful 26.a4! In the time trouble that followed, the assessment varied from "White wins" to "Black can hold". On the 39th move White could win by 39.Qe4! with the idea 39...Qg5+ 40.Kh1! Qxd2 41.Qa8#. However, Alexander played 39.Ra2, and Black overstepped the time limit, although he could put up some resistance by 39...Rd8! Now 40.Qe4? Qg5+! loses for White. Of course, even after 39...Rd8 White still has an edge, but the outcome of the game could well be different.
Akopian-Leko The Hungarian grandmaster outskilled the Armenian in opening preparation. In the Meran Variation of the Slav Defense he surprised the opponent with a refined novelty 17...Qd5! Black created strong pressure. In my opinion White had to consider sacrificing an exchange in order to seize the initiative. For instance, it is interesting to look at 21.Bd2! Nb3 22.Rb4 Bxg2! (a fine resource) 23.Bxg2 Nxa1 24.Bc3. Of course, such moves are easier to suggest that actually play on the board. Akopian decided to give up a pawn, and after a few accurate opponent's moves (23...Ba6!, 26...Qc2!) had to accept a difficult endgame without a pawn, which he failed to hold. Leko's technique in the final phase of the game was perfect - a textbook example of converting the advantage!
Alekseev-Eljanov A tragic story! At the early phase of the game it looked like the Russian is doomed to fall, however, it was the Ukrainian who perished in the end. In a fashionable line of the Queen's Gambit Eljanov employed an excellent novelty 12...Qe8! (earlier Black only played 12...b5, allowing the attacking 13.a4!). Alekseev reacted poorly. He couldn't play 13.Be2 due to 13...Ne4!, but he had to make a preparatory move for the development - 13.Qc2! The move he selected - 13.f3?! - left the White's king in the center and allowed Black advancing his queenside pawns. Black's advantage became absolutely clear by the 20th move. I think the most accurate solution was 20...Bd7! In this case Black could prepare the с4-с3 under the perfect circumstances, as White's most natural defense 21.Nb1 Rfc8! 22.Nc3 (otherwise c4-c3) 22...Nxc3 23.Qxc3 Bxg4! loses a pawn. Pavel wasted a tempo, allowing White to make a useful move - 20...Re8 21.h3, and then started playing on the queenside, which did not make the desired impact. Evgeny defended well, and his bluff 26.h4! shook the opponent's confidence. Pavel started making mistakes. His 26...Bb5? was a grave error, changing the evaluation dramatically from "almost equal" to "a huge advantage to White". The advantage was duly converted into a victory.
Bacrot-Svidler The first impression of a non-conflict game is wrong. There was real tension in it. In the Chigorin Variation of the Ruy Lopez the players covered the entire board with solid pawn chains. White had a small loophole on the kingside, but in order to use it he had to dare sacrificing material. The best opportunity occurred, in my opinion, on the 30th move. I think White could play 30.Nf5!? gxf5 31.gxf5, and then slowly increase pressure by g2-g4, doubling the rooks and maybe even rushing the kingside pawns h3-h4 and g4-g5. Surprisingly, he could afford playing in such style without a piece and almost at no risk! Black only could wait behind bars... After the move played in the game - 30.Bh6 - Svidler closed all the doors by 30...g5!, and a draw became inevitable.
Ivanchuk-Gelfand This game was very interesting, and it seems both players conducted it almost perfectly. In the Slav Defense White got the bishop pair, and Black obtained active piece play against the enemy king. I think Ivanchuk incorrectly allowed the Black's queen to h3. Instead of the rushy 18.e4 I suggest the prophylactic 18.Kg2! - the king can protect himself! After that White has all the chances to prevail in the center. After Black organized the scary setup "Qh3 + Ng4", White's advantage evaporated. Gelfand masterfully increased tension (21...Be3!, 25...f5!), and then forced the move repetition by daring cavalry ride 28...Nc1!
Mamedyarov-Kamsky In the Gruenfeld Defense Black made an interesting novelty, but we did not get a chance to see the depth of Gata's preparation. The critical line is probably 14.dxc6 followed by trading the queens, however, Shakhriyar reacted in a very tricky fashion. He traded queens in a different way, but failed to get any real advantage. Soon White's advanced pawns became targets of Black's counterattack. It was followed by exchange of logical and easy-to-calculate (at the grandmaster level at least) tactical blows that resulted in an equal rook ending, which ended peacefully.
Kamsky-Svidler
Gelfand-Bacrot
Eljanov-Ivanchuk
Leko-Karjakin
Mamedyarov-Akopian
Aronian-Grischuk
Kasimdzhanov-Alekseev
Karjakin-Aronian
I found many unexpected resources in the analysis of this complicated game. Kasimdzhanov employed a very risky but interesting plan in a well-known variation of the Queen's Gambit Accepted. I thought that after 14.Nxb6 Black had to take with the c7-pawn, keeping the king's shelter intact. After 14...cxb6 White cannot retreat the knight from e2 because of the d4-pawn weakness, while 15.Ba4 is well met by 15...g5! with strong counterplay for Black. 
Akopian-Leko
Alekseev-Eljanov
Bacrot-Svidler
