Krush-Yip, Part Three

The thrilling conclusion to one of the best games of the year.

This post concludes our analysis of Krush-Yip from the U.S. Women’s Championship. Here’s a link to the whole game if you want to play through it on a digital board.

Once you’ve mastered the basics, have become fluent in the language of chess, honed your calculation skills, a lot of what happens during a game is about making decisions. It’s not easy to beat an opponent of similar strength; after all, they know about the same amount as you do. And so the difference maker is often which player makes better decisions, the one who is better able to find the balance between taking risks and maintaining an defensible position in case things don’t work out. A lot of these decisions are made with imperfect information, when the evaluation of the position is beyond the horizon of human calculation.

The greatest players in chess history—Fischer, Kasparov, Carlsen, to name three—have a keen intuitive sense of just how far they can push for a win without going over the edge. Their consistent feel for what’s possible is what separates them from their contemporaries and rivals. That intuitive sense is one of the most impressive features of Carissa Yip’s play in her fourth round game against Irina Krush. In this installment she makes a very difficult decision to play on without knowing exactly what will happen and without leaving herself in an objectively losing position¹.

Here’s where we left off last time:

It’s a state of mutual paralysis. White would like to swing her rooks around to the kingside and attack from the flank, but any move off the b-file allows black to untangle her pieces, starting with Ne5. Black would like to advance her h-pawn, but white’s likely to round it up if it goes to h2 too soon.

The next few moves demonstrate this, as the players try to feel out the new situation: 40 … Qh4 41 Bd3 Qg3 42 Bf1 (the only reasonable move, as 42 Be2? loses to 42 … h2) 42 … Qh4 (42 … h2 43 Qg2 Qh4 44 Be2 Ne5 45 Rh1 is one of many variations in which black loses the h-pawn but wins c4 and d5; the computer thinks that it’s still just 0.00) 43 Bd3 (an undercover draw offer, perhaps, but one that Yip declines) 43 … Ne5!

This is a hell of a gutsy decision by Yip. Her tournament position suggests that she should be content to draw, particularly with the black pieces against one of her top rivals. But instead she rolls the dice, trusting that she can white’s attack on the b-file isn’t decisive and that Krush will be forced to make some difficult decisions over the next few moves.

There’s only one way to get better at this kind of intuitive, long-range decision making, and it’s to eschew draws whenever possible. There’s no way to know if you’re taking the right kinds of risks if you don’t fight on when you can, and this sustained desire to win is how you coax mistakes out of even very experienced and strong opponents (such as Irina Krush).

The first choice facing Krush isn’t particularly interesting. She can play 44 Be2, which allows 44 … h2 and necessitates 45 Rxb6, or she can just get it over with and take on b6, sacrificing the bishop in the process: 44 Rxb6! Nxd3:

Now there’s a real choice, whether to play Rb7 or Qb5. Krush played the latter, so we’ll examine 45 Rb7+ first. Taking the rook loses routinely: 45 … Bxb7? 46 Qxb7+ Kd8 47 Qb8+ Ke7 48 Rb7+ Kf6 49 Qd8+, skewering the king to the queen.

We're on the tightrope now: black has to play 45 … Kd8 and white’s only good response is 46 Qb5. What should black do next? 46 … Bxb7? 47 Qxb7 Qxc4 is tempting, as black’s up a piece, the Ra6 is protected, and Qxc3+ is coming. But that’s all too slow, as white’s attack comes crashing through. 48 Qb8+ Ke7 49 Rb7+ (I spent some time on 49 Qh8, which is also strong, but not mate in five strong) 49 … Kf6 50 Qf8+! Ke5 51 Re7+ Kxd5 52 Qf3+ with mate to follow.

How about some counterplay with the h-pawn? The immediate 46 … h2? loses to 47 Qxa6! h1=Q (47 … Bxb7 48 Rxb7 is no better, all white’s attack needs is two heavy pieces) 48 Qxd6+ Ke8 49 Qg6+ Kd8 50 Qg8#. There’s another way to make this work: black should first include 46 … Qh8!? 47 Kb3 and only now, with the king obstructing the b-file, 47 … h2. After 48 Qxa6 Bxb7 49 Qxd6+! Kc8 50 Rh1 the position is fiendishly complicated, but it seems to be smoother sailing for white than black.

In this case the fourth try is the charm: 46 … Ne5! is objectively the best move to keep the balance, although only just barely:

Let’s go through each of white’s winning attempts:

• 47 Qxa6 Nxc4+ 48 Ka2 Qe4! (only move) 49 Qa8 (threatening Qxc8+ and with mate to follow) 49 … Qc2+ 50 R1b2 Qxa4+ with a perpetual on a4 and d1.

• 47 Rb8 (again threatening to capture on c8 followed by a mating attack down the b-file) 47 … Nxc4+ 48 Ka2 Rb6! 49 Rxb6 Qe4! This is all quite logical if you realize that you need to deflect the Rb8 to win a tempo to create counterplay against the white king, but the problem is that it’s not so easy to realize!

• 47 Rg1 Bxb7! 48 Rg8+ Kc7 49 Rg7+ Kd8! and there’s nothing more than a perpetual for white.

I think there’s basically no way that Yip saw all of this when she decided to play on, but the computer variations support her intuition that she could survive the worst white could throw at her king. That said, the winning chances (that is, the chances of the other player making bad mistake) would have been with Krush had she had played 45 Rb7+.

If this was the only variation that Yip had to calculate when considering 43 … Ne5, it would be impressive enough, but she also had to reckon with the move Krush actually played, 45 Qb5, because there’s only one acceptable counter, the lovely and unexpected 45 … Nb4! The knight miraculously holds the queenside together, defending c6 and a6 and cutting off white’s Rb1. And finally Yip provoked the mistake that she had been hoping for.

Krush played 46 Rxa6?, although it’s not immediately clear why this move is wrong². I imagine that her idea was 46 … Bxa6 47 Qxa5+ Kd7 48 cxb4 Qg3+ and now 49 Rb3, which looks very promising from a distance. White’s up an exchange and both black pieces are under attack:

Conventional remedies such as 49 … cxb4+? allow white to consolidate her material advantage. No, we have to play for mate: 49 … Qe1! is best and in fact it’s actually winning³. Here’s a plausible variation: 50 Rxh3 Qc1+ 51 Ka2 Bxc4+ 52 Rb3 Qc2+ 53 Ka1 Qd1+! (53 … Qxb3 54 Qa7+ and there’s no good way to escape from the checks) 54 Rb1 Qd4+ 55 Rb2 Qg1+! 56 Rb1 Qg3!

What is black doing, you might ask? Just getting the queen to the best square, where it protects the d6-pawn, before threatening mate on a3. Computers are incredible at finding these geometric maneuvers; I’m not sure if Yip would have found this idea or not.

With d6 defended, the attempted perpetual 57 Qa7+ Ke8 58 Qb8+ Kf7 59 Qc7+ Kg6 fizzles out. The computer suggests 57 Rd1 as the best defense, but black’s attack rages on: 57 … Qa3+ 58 Kb1 Ba2+ 59 Kc2 Bb3+ 60 Kd2 Qb2+ 61 Ke3 Qe5+! 62 Kf2 and only now is it time to take the white rook.

I’m not sure why Krush decided against 49 Rb3: it gave her the best hope of holding out, and it’s a shame that she didn’t force Yip to find this long and difficult variation. Instead, after 49 Kb2? it came crashing down in a hurry: 49 … Qe5+ 50 Kc2 Qe2+ 51 Kc3 Qxc4+ 52 Kb2 Qd4+ 53 Kc1 Qc3+ 54 Kd1 h2, and white resigned:

This game is a testament to Yip’s fighting spirit and her extremely high quality of play over the first two-thirds of the tournament. The last three rounds saw her come crashing back to earth, a reminder that what Fischer did in 63-64⁴ is basically impossible, a DiMaggio-esque streak that goes against everything we know about the difficulty of winning a bunch of games in a row against strong opponents.

Yip’s eight game run was also important for raising my interest in the women’s championship. Given how busy my life is during the school year, I tend to follow the open section of most major events such as the U.S. Championship or the Olympiad. I know more about the players there and have more interest in the storylines—but this year, I jumped straight to the women’s games, way more interested in the “Yip goes for 11/11” storyline than whether Awonder Liang could mount a comeback against Caruana⁵.

Now that the analysis is done, it’s time to weigh in:

Thanks for reading, next up in the chess world is the championship match between Ding Liren and Gukesh.

1. Not to be confused with putting herself in a position that could easily be lost; that’s the risk-taking part.

2. It’s better to wrap things up with 46 Rb8! Qg3 47 Rxc8+ and white will give perpetual check.

3. It’s worth noting that other ways of threatening Qa1#, such as 49 … Qe5 and 49 … Qg7 both fail to 50 bxc5. Only 49 … Qe1 pins the b-pawn to the white queen on a5.

4. And then again, with interest, in the candidates cycle that led to the 1972 World Championship.

5. Or the unfortunate Christopher Yoo storyline.