The tiebreaks at the World Cup are getting shorter. There were Armageddon games in the first two rounds, while rounds 3 and 4 made it through the 5" minute games. (And with an 80-90 minute "ten minute" game in round 4 it took longer than a normal series making it to the Armageddon game.) Today, the tiebreaks finished as quickly as possible; to wit, after the initial pair of 25-minute games.
Dmitry Andreikin was the first one through, and a very convincing winner over Peter Svidler. In the first game Andreikin played one of his typical low-theory lines, in this case a Tromp-turned-Torre Attack, and it was a twofold success. Andreikin did obtain a small advantage, and Svidler was forced to solve problems over the board rather than relying on prep or anything like it. Svidler did manage to equalize at one point and perhaps got a bit too bold. 20...f5 seems to me a very risky move to make in a rapid game, as it offers White various opportunities to open the game up when Black will be short of time. A safer way was 20...Ne7, aiming to trade all the rooks on the c-file and go for the quick handshake. That inaccuracy was compounded by 23...Nxa2 - 23...Rxc1 and only then taking on a2 seems more accurate, as Svidler's version allowed 24.Rxc8 Rxc8 and now, as advertised, 25.e4. After 28.Re6 Black was still objectively okay but as a practical matter it was starting to look dangerous. 28...Kg8 was an error, and after 29.Qg3! Black has some annoying threats to deal with like 30.Nh5 and 30.Rxh6. Svidler's reply, 29...Nd5 was natural and logical...and an absolute blunder. After the sneaky shot 30.Qb3! there was nothing for Black to do but resign, as the knight is lost - at least it is unless Black wants to lose the queen, e.g. 30...Nxf4 31.Re7+.
The rematch didn't go any better for Svidler, except insofar as he received a charity draw offer in the end. For a few moments Svidler looked as if he might get to enjoy a relatively safe extra pawn in another Advance Caro-Kann with 3...c5, but the critical moment came on move 18. Had Svidler played 18.a3 he would have enjoyed some advantage. He chose 18.Nf3 instead, and while this is a move White wants to play it's too soon. After 18...axb4 19.cxb4 Ra3! followed soon by ...Qa6, ...Rd3 and ...Qa3 Black managed to infiltrate and regain the material (with positional interest) without allowing White any real attacking chances on the kingside. Had Andreikin needed to win he probably would have played differently on move 34; instead, it was enough to force a draw with 34...Rg4 35.Rf2 Qe1+ 36.Rf1 Qe2 37.Rf2 Qe1+. The draw was agreed and the match was over, earning Andreikin the chance to play his friend and teammate Evgeny Tomashevsky for a shot at the finals and an automatic berth into the next Candidates' event.
In the other quarter-final Vachier-Lagrave was a nominal underdog against Fabiano Caruana, but the former's play in both the classical and rapid disciplines in the event rendered the rating difference immaterial. In the first game Caruana had White and was the one pressing, at least in theory, but Vachier-Lagrave defended so accurately that Caruana never came close to a genuine edge. In game 2, however, the Frenchman called the tune from early on. Caruana played the Dutch, which isn't normally part of his opening repertoire, and while both players occasionally seemed a little unsure of themselves Vachier-Lagrave's play came across as more purposeful and coherent. There was never any question about who stood better, only whether White's advantage would grow into something major. Practically speaking, the decisive error may have come at move 33 when Black played 33...Bf6. The upshot was that after 34.b3 Black's knight could no longer safely retreat to d6, and after 34...Na5 the misplaced knight and Black's weak dark squares were a serious problem. The last chance, objectively speaking, came on move 48. Caruana, who was very short of time and way behind on the clock, needed to take the knight. Vachier-Lagrave would soon regain the piece and maintain an advantage, but it was a small chance for Caruana. He declined the sac with 48...Qg7, and the rest was a rout. In the final position the d-pawn can't be captured without allowing the h-pawn to queen, and if 66...Kf8 67.Kd4 Kg7 68.Kc5 Kh6 69.Kxb5 Kxh5 70.Kb6 followed by 71.Kc7 leads to the knight's elimination and the d-pawn's promotion. That means that Vachier-Lagrave will face Vladimir Kramnik in the other semi-final.
Who will win these matches? At this point it's crazy to pick against any of these guys. Kramnik is surely the strongest player of the four, but he's also the oldest and possibly the most tired. He has less motivation, as he has already qualified for the Candidates', and if it comes down to tiebreaks one might wonder if he's (that much) better than Vachier-Lagrave in rapid and blitz. In the other semi Andreikin has been an absolute assassin in the rapid tiebreaks, going 3.5/4 against Sergey Karjakin and Svidler combined - and it could have been 4-0. On the other hand, his play in the classical games hasn't been as out of this world, while Tomashevsky has been playing like a super-hero, rising to the occasion every time. I am going to go with Kramnik and Tomashevsky, with a codicil: if they don't win in the classical stage I think they will lose in the tiebreaks.