The short story of a long castling

[Lyudmil Tsvetkov]

Just very brief remarks on the 25th game of the TCEC final.

[d]r3k2r/2q1ppb1/1pp3pp/p1p1Pb2/P7/1P1P1N1P/1BP1QPP1/R3K2R b KQkq - 0 1

Here Komodo castles with black long 15...0-0-0.

The opening is perfectly equal, and black could have retained a balanced game after the much more natural 0-0.

For example: 15...0-0 16. 0-0 Qd7 17. Nd2 Be6 18. f4 f5, with perfect equality.

[d]r4rk1/3qp1b1/1pp1b1pp/p1p1Pp2/P4P2/1P1P3P/1BPNQ1P1/R4RK1 w - f6 0 5

White can not break on the king side, neither playing d4 is dangerous, as the black king is already safe. On 19. Nc4, black plays Qc7, while ef6 ef6 also leads to a perfect draw.

On the other hand, 15...0-0-0 is a major mistake. The king is left in bad safety, and this will sooner or later tell. And indeed, SF won after 33.Rd1, followed by d4 and the black position fell apart precisely because of the bad placement of the black king, hiding in a more open place behind uncomfortable doubled pawns.

Anyone having a clue why Komodo castled long in that game?

[kbhearn]

in your specific line 17. Nh4 prevents your idea of a black f5 getting in. Changing blacks 16th move leads to your planned setup being prevented in other ways. In general it appears as though black has a very passive position and that in many lines white eventually plays g4+f5 after black o-o with significant pressure. It's hard to say this is better than the game continuation without serious analysis.

[Dicaste]

Just very brief remarks on the 25th game of the TCEC final.

[d]r3k2r/2q1ppb1/1pp3pp/p1p1Pb2/P7/1P1P1N1P/1BP1QPP1/R3K2R b KQkq - 0 1

Here Komodo castles with black long 15...0-0-0.

The opening is perfectly equal, and black could have retained a balanced game after the much more natural 0-0.

For example: 15...0-0 16. 0-0 Qd7 17. Nd2 Be6 18. f4 f5, with perfect equality.

[d]r4rk1/3qp1b1/1pp1b1pp/p1p1Pp2/P4P2/1P1P3P/1BPNQ1P1/R4RK1 w - f6 0 5

White can not break on the king side, neither playing d4 is dangerous, as the black king is already safe. On 19. Nc4, black plays Qc7, while ef6 ef6 also leads to a perfect draw.

On the other hand, 15...0-0-0 is a major mistake. The king is left in bad safety, and this will sooner or later tell. And indeed, SF won after 33.Rd1, followed by d4 and the black position fell apart precisely because of the bad placement of the black king, hiding in a more open place behind uncomfortable doubled pawns.

Anyone having a clue why Komodo castled long in that game?

Artista's continuation

Code: Select all

1... O-O 2. O-O Qd7 3. Nh4 Be6 4. f4 Kh7 5. Qf2 f5 6. Qg3 Qe8 7. Rad1 Qf7 8.
Nf3 Rad8 *

[Lyudmil Tsvetkov]

in your specific line 17. Nh4 prevents your idea of a black f5 getting in. Changing blacks 16th move leads to your planned setup being prevented in other ways. In general it appears as though black has a very passive position and that in many lines white eventually plays g4+f5 after black o-o with significant pressure. It's hard to say this is better than the game continuation without serious analysis.

I think mine was a serious analysis.

Why should I change my 16th move, when I think it is the strongest?
One of the reasons the queen is on d7, is precisely to protect Bf5.

In certain lines, g6-g5 for black ensures equality. I did not specifically check it, as I did not consider Nh4 at all, but I think after 17.Nh4, 17...g5 should fully equalise.

The major distinction between the game continuation and the suggested one is that already the black king is completely safe, so that moves that were earlier possible, now are not for white.
At the same time, after white castles short, certain pawn moves involving the white pawn cover are already difficult to play. For example, g4 is dubious in all lines.

Besides that, if I follow your logic, one should say that there was no reason for the black loss, as, besides castling long, Komodo seemingly did not make any other obvious mistakes.

Or, at least I did not observe them.
But I would be glad if you point them out to me.

[Lyudmil Tsvetkov]

Concerning the return 26th game, also a very interesting one, I think the main SF failure was again an outwardly unobservable positional mistake.

12...f6 loses the game straight, as after this all tactics, including fe5 and Nh4, lead to a positionally very difficult, I think altogether lost endgame, with an isolated pawn on the e file and bad backward b6 pawn.

Instead, 12...Ba6 (the natural reply to Komodo's c3 move, weakening the d3 square) should have given an easy draw.

For example: 12...Ba6 13. Rad1 f6 (now this is fine, as black has specific tactical threats; all black must take care of is that it does not capture fe5, staying with an isolated e pawn, but instead wait for white to capture ef6 and recapture ef6 itself) 14. 0-0 Qd5 15. Rfe1 Rad8 16. c4 Qd7, and black is fine.

[d]3r1rk1/p2qp1bp/bpp2pp1/2p1Pn2/2P1NB2/3P1N1P/PP1Q1PP1/3RR1K1 w - - 0 8

In the above line, black has tactically induced white's c4, after which the d4 square becomes available to the knight. There is perfect equality on the board.

I think that is what positional chess is all about: you are fighting for squares and do not pay much attention to tactics. If your positional play is relevant, the tactical should also be, but the opposite is not true.

Now, why the game continuation is simply lost?

[d]r1bq1rk1/p3p2p/1pp3p1/2p1b3/4N2N/2PP3P/PP1Q1PP1/R4RK1 w - - 0 16

How many engines readily recognize this is already lost for black?

SF shows just some 20cps white advantage, so it obviously does not think so. But the reality is otherwise.

Nominally, eval should say equality: you have the pair of bishops, some 50cps, for an isolated e pawn, 20cps, plus a doubled c pawns, another 20cps.

However, white has a range of other, subtler advantages. Ne4 is outposted; if white manages to transfer its other knight to c4, the knight there also becomes outposted, as it is impossible for black to reasonably play b6-b5; the b6 black pawn after white's a4 will become backward and vulnerable.

Besides, with white knights on e4 and c4, the knights are both on light squares and could be attacked only by black's light-coloured bishop. If the bishop captures one of them, black is left with a bad dark-coloured bishop for a good white knight, as all important pawns in the position, e7/e5, c5, central black pawns, are blocked on squares the colour of the black bishop.

When considering penalty for pawns on squares the colour of the own bishop, it must be said that the penalty for a central blocked pawn might be worth the penalty for 3 or 4 other, less important non-central blocked pawns, and a whole cohort of pawns that are not central and blocked.

So that the white knights are completely invulnerable, and the bad black bishop is the dark-coloured one.

Playing Ba6 in this particular line also seems impossible, as the e6 square is left without control.

I even do not consider other tactical black retorts, as Bf5, Bf6, early e5, etc., as positionally black simply remains very bad.

So that, objectively, after 12...f6 black already loses the game.

[d]r2q1rk1/6bp/1pp3p1/p1p1pb2/P1N1N3/2PP3P/1P2QPP1/R3R1K1 w - e6 0 22

Now you see, what happens couple of moves later on: e5 isolated, c pawns doubled, b6 backward, Ne4 and Nc4 outposted. SF gives just 40cps white advantage even here, but objectively white already leads by at least 150cps.

So that consequances of 12...f6 were absolutely irreparable.

I called the initial thread describing the 25th game The short story of a long castling, but maybe this one was A long story of a short castling. (with a further mistake added)

Anyone that agrees on at least one of the things I said here?

[Lyudmil Tsvetkov]

Sorry, I forgot to post the main diagram after white's 12th move - c3.

[d]r1bq1rk1/p3ppbp/1pp3p1/2p1Pn2/4NB2/2PP1N1P/PP1Q1PP1/R3K2R b KQ - 0 12

My claim is that here 12...f6 loses straight for black, while 12...Ba6 gives equality.

[Lyudmil Tsvetkov]

Actually, not 12...f6, but fe5 is the major mistake.

Ba6 could still have been played on the very next move with excellent drawing chances, as lines transpose.

[cdani]

Just to reinforce the generic argument, if you see for example the code of Stockfish for the endgame rook + pawn against rook, you see some specialized code. Why it's necessary? Because sometimes it will not be able to find good lines in a reasonable amount of time.

So if some specialized code it's necessary for and endgame of 5 pieces, you can bet what you want that it will be necessary a lot of code to improve in many different middlegame positions. And the next engine that will do it, will win clearly the basically tactical beasts like stockfish.

[Lyudmil Tsvetkov]

Just to reinforce the generic argument, if you see for example the code of Stockfish for the endgame rook + pawn against rook, you see some specialized code. Why it's necessary? Because sometimes it will not be able to find good lines in a reasonable amount of time.

So if some specialized code it's necessary for and endgame of 5 pieces, you can bet what you want that it will be necessary a lot of code to improve in many different middlegame positions. And the next engine that will do it, will win clearly the basically tactical beasts like stockfish.

Thanks Daniel, for supporting the eval approach.

I think you are the only one here...

[kbhearn]

Well you asked why the engine wouldn't go kingside, and looking at engine lines quite simply why they'd avoid it as a sample line would be something along the lines of :

15 ... o-o 16. o-o Be6 17. Nd2 Qd7 18. Rae1 where black will never be able to play f5, white has more space, more activity, and all the chances. f4+g4 seems to be quite playable in most lines simply because black has no activity, opening lines on the kingside means his king is opened up too and white is better placed to take advantage of it. Is it more likely to survive than with the king on the queenside? maybe... But it is an entirely different game to analyse and may not be sufficient.

Regarding your line: 16 ... Qd7?! 17. Nh4 g5? leads to difficulties after 18. Nxf5 Qxf5 19. f4 gxf4 20. Rae1 at which point in multipv several attempts to flail around are presented with evals ranging from +1.3 to +1.5. Essentially black desperately tries to activate, successfully manages that, but then white infiltrates the queenside and it all starts to slowly fall apart as b6 and c6 are weak and impossible to protect while maintaining enough activity to suppress white's kingside attack...

As to a proposal for an alternate mistake to blame, i'm not sure i'd be able to posit one. Just that your suggestion of o-o instead is hardly a cure leading to a simple draw - just a different inferior position that may or may not be holdable.