22. Qa4 f5 23. Qb3 f4 24. Bd2 Rxe2 25. Rxe2 Qxe2 26. Qxd5+ Re6 27. h4

click for larger view

22. g3 f5 23. c4 Nb4 24. a3 Nd3 25. Rd1 f4 26. Rxd3 Qxc4 27. Bd4 cxd4 28. Rxe5 dxe5 29. Qd7 Rf8 30. Rf3

click for larger view

A couple of other general points -- many kibitzers (over the past few years) have pointed out various 'mistakes', 'blunders' and even 'losing moves'. It's rarely so simple, though I can definitively refute one suggestion: whoever said that black was winning a pawn after 19.Bf4 Rxe2 20.Rxe2 Rxe2 21.Qxe2 Qxf5?? may have overlooked 22.Qe8#, when a fine positional grip is ruined by an awkward checkmate. (Although, in fact, 19.Bf4 seems to lead to rough equality -- but it's not a nice move to have to make, and if it's white's best then he's already gone badly wrong).

I agree that the key move is white's 19th, and that 19.Qb7 turns out badly. Is it necessarily a 'mistake', though? Let alone a blunder?

I previously let an engine (probably Fritz 11) grind away at that position, and it had to reach roughly 24-ply before it saw that there was anything seriously wrong. A human player OTB, even a future world champion, would be even less able to calculate all the way to a decisive advantage for black.

19.Qb7 (or the alternative Qc6, which can transpose) is precisely the sort of gamble that many strong GMs would make at the board. It seems to offer winning chances, with risks attached. White will pick up queenside pawns, gain tempo with the attack on the Nb6 (albeit an illusory tempo as the knight is so strong on d5/f4) and bring about a position where he will have a won ending if he can withstand black's kingside counter-attack.

Some players would fall at the first hurdle (me, I suspect, included) by underestimating 20...Qe6 -- if I was white, trying to decide on my 19th move, I'd probably have focused on 20...Nd5. But the queen semi-zwischenzug is very strong, increasing the power on the e-file and opening the way for the f-pawn, among other things. I'd expect a GM (or an engine) to anticipate it at move 19.

More subtly, I'd have expected black to push the f-pawn quickly, with ...f5-f4 aimed at the blockading bishop and ultimately the king. Cleverly, Capa found a more direct attack (with no risk of exposing the black king on the rank or diagonal).

It's true there are a few points where both players miss the best move. In black's case it doesn't seem to matter much: Capa missed a couple of chances to finish the game quickly, but I don't think he ever seriously let Alekhine back into it.

If 19.Qb7 is a blunder, then white's entire plan - starting a few moves earlier - is bad. Yet turning black's advantage into a win takes a lot of top-level play. It's not just Capa's technique. I'm sure Alekhine saw the outline of white's attack and gambled on surviving it.

Even after the pseudo-sac of a rook on e3, black has to play well to win. At the board, between normal players, white could still hope for chances in the position around move 30: heavy pieces on an open board, mate and perpetual threats on both sides.

As far as I can see, 23...Nxg2 *does* win -- but 23...Qg4 is stronger, and 23...Qc4+ (with ...Nxg2 to follow) is stronger still. And white's 30.Kb3 really *was* a blunder -- 30.Kc1 might have struggled on, but the move played should lose at once to 30...Ra8. Capa's 30...Qc6? returns the favour, and he was lucky to retain a winning advantage.

I think back in this era, move 30 was the end of the first time control. Capablanca was the fastest player ever, but even with ample time in the clock, it's noticeable that every one, including the very best, makes more inaccuracies near time control probably mostly for psychological reasons. The other explanation is:

<notyetagm: No, 23 ... Nxg2! is a winning move. Dr. Tarrasch pointed out that 23 ... Qg4! was a quicker win, that's all. So Capablanca has two winning moves to choose from at move 23 and he played the one that he had probably seen like five moves earlier.>

Capablanca had probably already seen that 30...Qc6 would bring him the win in any variation; and rather than look around wasting time for more winning lines, he simply automatically banged out his last several moves in his usual rapid manner. Apparently he was particularly prone to this habit once he saw a clear winning line.

I may be missing something in the OE. This position, Opening Explorer, has only 2 games with 8. Nf5. In any case, if it is a common move, it probably shouldn't be. Scanning the engine opening books, it looks like 8. Re1 is the sanest choice. I'm not sure I needed a book for that since Re1 looks fine by inspection whereas Nf5 looks goofy to me.

Since almost 4 years have passed since I posted, I decided to run Rybka 3 sliding back from 21...Nd5 where black is clearly in some difficulty. I probably used Fritz 8 for the original post.

19. Qb7 rated about half a pawn in black's favor. If white plays 19. Bf4, the rating is even. A difference of half a pawn for one move is enough for me to call it a mistake. Here's the Rybka line to maintain equality. It's a simple line which doesn't really need an engine to discover.

19. Bf4 Rxe2 20. Rxe2 Rxe2 21. Qxe2 Kf8 22. g4 h5 23. h3 hxg4 24. hxg4

click for larger view

There is still a lot of life in the position and one could guess that Capa would manage to win against a munchkin Alekhine.

16. Bd2 is slightly better on the Rybbie scale to Be3. Since the pin on the Be3 caused so much trouble for white, I would say Bd2 is a lot better. But I wouldn't call Be3 a mistake.

14. Qf3 Ne5 15. Qxb7 Qxf5 is a better then 14. Bxc6. No reason to give up the two bishops here.

13. Bf4 may be better than the pointless Nxf6. It's "nice" to win the minor exchange but not when the knight is a better piece than the bishop.

10...Ne5 or Re8 is probably better than Nd7. Rybbie goes from = to after Nd7.

Fritz/Chessbase seems to decide that it's [C66] -- Ruy Lopez, Steinits Defence -- whatever the move-order.

Here's the rest of Shabanov vs Kuzmin: an interesting little game.

8.Nf5 0-0
9.Nxe7+ Qxe7
10.Nd5 Nxd5
11.exd5 Ne5
12.Bxd7 Qxd7
13.b3 Qf5
14.Bb2 Rfe8
15.Qd2 Ng6
16.Rae1 Nf4
17.f3 Rxe1
18.Rxe1 Nxd5
19.Re4 h5
20.Qd3 1-0

On my own, I'd have thought that the delayed fianchetto on b2 was an excellent idea -- clear most of the centre away, then open up on the dark squares. And I'd have thought black took a risky pawn with 18...Nxd5. But Fritz tells it differently -- it was equal, apparently, until the blunder 19...h5?? which lost at once. Instead, 19...Nf6 is OK for black.

I walked through the OE using this game and the Scotch and it produced the same set of games after move 8. Clearly there are OE magic tricks that I'm missing.

Agree. Just to be clear, when I say "annotation by result", I don't mean using computers -- I mean all those books that assume, because Player A won and Player B lost, everything A did was right and everything B did was wrong.>

Heh. Oops. After I posted I realized I had no idea what you meant but let it go in case I got lucky.

30... c4+ 31. Kb4 Qc6 32. b3 d5 33. Rf2 f6 34. Rg2 g6 35. Rd2 Ra8 36. Qxa8+ Qxa8 37. bxc4 dxc4 38. Kc5 f5 39. Rd4 Qxa2 40. Rd8+ Kg7 41. Rd4 Qxh2 42. Rd7+ Kh6 43. Kxc4 -132

click for larger view

30... Ra8 31. Rxd6 g5 32. Qxa8+ Qxa8 33. Rd2 f5 34. c4 Qa5 35. Kc2 Qxa2 36. Kc3 Qa4 37. b3 Qa1+ 38. Kc2 Qe5 39. h3 Kg7 40. Rd3 Kf6 41. Rd2 h6 42. Rd3 h5 43. Rd2 -17

click for larger view

23...♕c4+!! 24.♔g1 ♘xg2! 25.♔xg2 ♖g5+! (the point of that insertion of the check: now 26.♔f1 is illegal) 26.♗xg5 ♕g4+ 27.♔f1 ♕h3+ 28.♔g1 ♖xe1#

So White is forced to interpose a ♖ on e2, so Black wins the exchange more cleanly. Even worse, Black is no hurry to take the pinned ♖, and after winning the exchange e2, the replacement ♖ will also be pinned. After 23...♕c4+ 24.♖de2, Blacks avoids any back-rank mates with 24... g6, and White really has no good moves to prevent the following:

23... Qc4+ 24. Kg1 Nxg2 25. Kxg2 Rg5+

click for larger view

PS

It's also worth noticing what happens when White does not take the Knight.

23... Qc4+ 24. Kg1 Nxg2 25. Ree2 Nh4 26. f4 R5e7

click for larger view