IRLchess

Three years ago, when a report on the Bunratty Masters 2006 was posted here, I asked how on earth White could have recovered from the diagrammed position, where Black has just played 31… Qh3 (How?, February 28, 2017).

Van Voorthuijsen – Kane, Bunratty Masters 2006
32. ?

The ICU games archive gave “…” at the end, indicating that some moves were omitted, and the result was 1-0. Since Komodo 9.3 evaluates the position above at around -28, this must be close to an all time record for recovery from a lost position, and I wondered how it could have happened.

I was delighted to find on Wednesday that Peewee van Voorthuijsen had commented on the original post, with the continuation of the game, for which many thanks.

The time limit was a flat all moves in 90 minutes, and both players must have been short of time, Black catastrophically so. A few moves earlier, White had been better (see second diagram), and now with 27. Rd1 would have a clear advantage, e.g., 27… Qe5!? (27… Rxd1? 28. Rxd1 and White is winning) 28. Bxe4 Rxd1+ 29. Rxd1 Bxe4 30. Rd4 Bf3 31. Qc7 Qxc7 32. Nxc7 Rc8 33. Rc4 Bd5 34. Rc5 Bxa2, with a safe pawn plus. Instead, presumably with the clock playing a role, he went astray via 27. Bxe4? Bxe4 28. Nc3??. Here 28. Nf4!? is the move, with interesting play. After the move played, Black was winning, and after 28… Bf3 29. Qc5 Qe6 30. Kf1 Rbd8 31. Ke1 Qh3, the first diagrammed position was reached.

The simplest win was 32. Qb5 Qxh2 33. Qf1 Bg2, and this is where Komodo 9.3’s lopsided evaluation comes from. Once Black missed this via 32… Qg2?, the evaluation dropped to -7. After several missed wins, including a mate in 2, he seems to have lost on time.

[Click to replay the full game.]

All in all, it’s probably a good thing that the old no-increment days are no more. But there’s no doubt it led to some dramatic finishes.

A revised report has been uploaded.

The celebrated ending of two knights versus a pawn is the hardest of the ‘basic’ endings, and was exhaustively analyzed long before the tablebase era. Interesting though it is, its importance is mostly theoretical, as the ending is very rare in practice. (It’s sometimes given as an example of what the practical player should not study.)

But “very rare” and “never” are two different things, and the ending cropped up yesterday in the Bunratty Challengers.

White – Mirza, Bunratty Challengers 2020
48. … ?

The diagrammed position shows the game at the critical point, a couple of moves before the ending in question appeared. The basic idea is well known: White needs to keep one black pawn on the board to avoid stalemate, and will usually need to blockade it behind the ‘Troitsky line’ e4-f5-g6-h4. Even then, the required winning procedure often runs foul of the 50-move rule.

The usual tablebases (e.g., Nalimov) show that from the diagrammed position Black loses with best play in 97 moves, but this does not account for the 50-move rule.

But the Syzygy tablebases give the information we need. From the diagram, Black has four moves that hold the draw via the 50-move rule: 48… Kc4, 48… Kc6, 48… g5+, and 48… h6. Instead, Diana Mirza played 48… h5?. The ending is so obscure that the ‘?’ is harsh, but this changes the valuation. White could now win via 49. Ng5! (only move), blockading the g-pawn behind the Troitsky line. It’s still a close-run thing: after, say, 49… h4 50. Ne3 h3 51. Kg3 Kc5 52. Kxh3, White mates in 49 moves with best play, with the next pawn move resetting the counter in 44 moves.

As it was, White erred with 49. Kg5?, when it’s a draw even without the 50-move rule. Using = to mean drawn with best play without the 50-move rule, and [=] to mean drawn only because of the rule, the play continued 49… Kc4? [=] (49… h4!, only move, =) 50. Ndf6? (50. Ne3+ [=]; 50. Nf4 [=]) 50… h4=. After 51. Kxh4 White continued, optimistically enough, for another 32 moves.

[Click to replay the full game.]

This is not the first Irish game to reach this ending: Keogh – Bjørgvik, World Senior 65+ Championship 2018 is the only other game in the ICU archive that did so. At all relevant stages that game was won for the two knights if not for the 50-move rule, but drawn once the rule is taken into account.