A historical game

[zullil]

This is already a perfect draw, interesting when SF will get it.

But you claimed the original position was a clear win for Black, if I recall.

SF captures the knight:

Code: Select all

info depth 40 seldepth 46 multipv 1 score cp 196 nodes 1425220512 nps 19064709 time 74757 pv c4d6 c7d6 d2c4 e8b8 c4a5 h8g8 a5c6 b8c8 b1c1 g7f6 g1g2 d7c7 b3a3 g8f7 a6a7 c8d7 c1a1 c7b7 f2f3 f7e8 a3a2 d7c8 a2a6 c8d7 e2c2 d7c8 c2c4 c8d7 a6a2 d7c8 c4e2 b7h7 a1c1 h7b7 e2d3 b7h7 a2a1 c8d7 d3e2 d7b7 e2f2 b7c8 a1a6 h7b7

[d]r3q2k/2pr2b1/Pp1N4/1PpPp1p1/4PpPp/1R5P/3NQP2/1R4K1 b - - 0 19

[Lyudmil Tsvetkov]

This is already a perfect draw, interesting when SF will get it.

But you claimed the original position was a clear win for Black, if I recall.

SF captures the knight:

Code: Select all

info depth 40 seldepth 46 multipv 1 score cp 196 nodes 1425220512 nps 19064709 time 74757 pv c4d6 c7d6 d2c4 e8b8 c4a5 h8g8 a5c6 b8c8 b1c1 g7f6 g1g2 d7c7 b3a3 g8f7 a6a7 c8d7 c1a1 c7b7 f2f3 f7e8 a3a2 d7c8 a2a6 c8d7 e2c2 d7c8 c2c4 c8d7 a6a2 d7c8 c4e2 b7h7 a1c1 h7b7 e2d3 b7h7 a2a1 c8d7 d3e2 d7b7 e2f2 b7c8 a1a6 h7b7

[d]r3q2k/2pr2b1/Pp1N4/1PpPp1p1/4PpPp/1R5P/3NQP2/1R4K1 b - - 0 19

I claimed that black has some 30cps advantage, which could be 10cps, as well as 50cps.

My move is Rd6.

[d]r3q2k/2p3b1/Pp1r4/1PpPp1p1/4PpPp/1R5P/3NQP2/1R4K1 w - - 0 20

Interesting, that in many lines black can sacrifice on d6 also the queen, and it is still a draw.

[peter]

[
You see, SF closes the game with g2-g4! It is afraid of black playing g5-g4 and getting advantage. That was my initial assessment of the position: a small, some 30cps black edge, if white plays g4, it is a draw, otherwise, black will play g4 itself and get winning chances.

Some time ago Lucas called my chess a 'cowardly approach', trying to close the position. And what is SF doing now? It closes the position itself. Is not this a cowardly approach?

Scuse me, Lyudmil, even risking a Rote Karte again, here I have to hop in for a very short notice once again yet:
As for my POV it's you closing the position with ..h4 after SF's g4, isn't it?
You could have gone on with your glorious Kingside attack with hxg4 instead too or simply play anything else, at least leaving it to SF again to close the position (on the Kingside only) or not to close it, couldn't you?
And my bet would have been in that case, Stockfish wouldn't have closed it, if you would have left him the choice.


And just one more little output of SF in MV mode of the position after g4 to illustrate that:

Code: Select all

r1n4k/2p1r1b1/Pp4q1/1PpPp1pp/2N1PpP1/2R4P/4QP1N/1R4K1 b - g3 0 1

Analysis by Stockfish 181114 64 POPCNT:

1. +-  (1.74): 29...h4 30.d6 cxd6 31.Rd1 Rea7 32.Nf3 Qf6 33.Rb3 Rb8 34.Rbd3 Bf8 35.Qa2 Be7 36.Nfd2 Bd8 37.Nb1 Bc7 38.Nc3 Qe6 39.Nd5 Bd8 40.Qe2 Kg7 41.Qb2 Rba8 42.Qa2 Qf7 43.Qd2 Qe6 44.Qc3 Rb8 45.f3 Rba8 46.R1d2 Rf7 47.Qb3 Kh7 48.Rd1 
2. +-  (2.44): 29...Qe8 30.gxh5 Bf6 31.Qg4 Nd6 32.Nxd6 cxd6 33.Nf3 Rh7 34.Nd2 Bd8 35.Nc4 Bc7 36.Kh2 Qxh5 37.Rg1 Qxg4 38.Rxg4 Rh5 39.h4 Rxh4+ 40.Rxh4+ gxh4 41.Rh3 Kg7 42.Rxh4 Rb8 43.Kg2 Ra8 44.Kf3 Rc8 45.Rh1 Re8 46.Ke2 Kg6 47.Kd3 Ra8 48.Rh4 Kg7 49.Rh3 Kg6 50.Rh1 Kg7 51.Ke2 Kg6 52.Rh4 
3. +-  (2.44): 29...Bf6 30.gxh5 Qe8 31.Qg4 Nd6 32.Nxd6 cxd6 33.Nf3 Rh7 34.Nd2 Bd8 35.Nc4 Bc7 36.Kh2 Qxh5 37.Rg1 Qxg4 38.Rxg4 Rh5 39.h4 Rxh4+ 40.Rxh4+ gxh4 41.Rh3 Kg7 42.Rxh4 Rb8 43.Kg2 Ra8 44.Kf3 Rc8 45.Rh1 Re8 46.Ke2 Kg6 47.Kd3 Ra8 48.Rh4 Kg7 49.Rh3 Kg6 50.Rh1 Kg7 51.Ke2 Kg6 52.Rh4 Kg7 53.Kd3 
4. +-  (2.45): 29...hxg4 30.hxg4 Bf6 31.Rh3+ Rh7 32.Rxh7+ Qxh7 33.Nf3 Na7 34.Kg2 Qe7 35.Qd3 Kg7 36.d6 Qd7 37.Nfxe5 Bxe5 38.Nxe5 Qxd6 39.Qxd6 cxd6 40.Nc4 Rb8 41.Nxd6 Kf6 42.Nc4 Kf7 43.Rh1 Nxb5 44.Rh7+ Kg8 45.Rh6 Nc7 46.Rg6+ Kf7 47.Ne5+ Ke7 48.a7 Ra8 49.Rxb6 Ne6 50.Rb8 Nc7 

[zullil]

My move is Rd6.

[d]r3q2k/2p3b1/Pp1r4/1PpPp1p1/4PpPp/1R5P/3NQP2/1R4K1 w - - 0 20

Interesting, that in many lines black can sacrifice on d6 also the queen, and it is still a draw.

Stockfish plays Nc4.

Code: Select all

info depth 40 seldepth 45 multipv 1 score cp 187 nodes 369597577 nps 15797468 time 23396 pv d2c4 h8g8 g1f1 g8f8 b3d3 a8a7 f2f3 f8e7 e2a2 e7d7 a2b2 d7c8 f1g2 c8b8 b1a1 b8c8 d3d1 c8b8 d1d2 b8c8 d2d3 g7f8 d3d1 f8g7 d1b1 c8d8 b1c1 d8c8 a1a3 c8b8 c1h1 b8c8 h1a1 c8b8 a1c1 b8c8 c1d1 c8d8 d1d2 d8c8

[d]r3q2k/2p3b1/Pp1r4/1PpPp1p1/2N1PpPp/1R5P/4QP2/1R4K1 b - - 1 20
The evaluation of 187 doesn't change as a function of depth, signaling a draw. Now the question is, can Black force this line of play to occur? At several points we followed sequences of moves from a PV, as if all the moves in the sequence were best possible. Perhaps SF would have deviated, although its inability to recognize a blocked position as a draw might cripple it in this example.

So we'll call the original position drawn, unless White can do better than the line just investigated.

[Lyudmil Tsvetkov]

My move is Rd6.

[d]r3q2k/2p3b1/Pp1r4/1PpPp1p1/4PpPp/1R5P/3NQP2/1R4K1 w - - 0 20

Interesting, that in many lines black can sacrifice on d6 also the queen, and it is still a draw.

Stockfish plays Nc4.

Code: Select all

info depth 40 seldepth 45 multipv 1 score cp 187 nodes 369597577 nps 15797468 time 23396 pv d2c4 h8g8 g1f1 g8f8 b3d3 a8a7 f2f3 f8e7 e2a2 e7d7 a2b2 d7c8 f1g2 c8b8 b1a1 b8c8 d3d1 c8b8 d1d2 b8c8 d2d3 g7f8 d3d1 f8g7 d1b1 c8d8 b1c1 d8c8 a1a3 c8b8 c1h1 b8c8 h1a1 c8b8 a1c1 b8c8 c1d1 c8d8 d1d2 d8c8

[d]r3q2k/2p3b1/Pp1r4/1PpPp1p1/2N1PpPp/1R5P/4QP2/1R4K1 b - - 1 20
The evaluation of 187 doesn't change as a function of depth, signaling a draw. Now the question is, can Black force this line of play to occur? At several points we followed sequences of moves from a PV, as if all the moves in the sequence were best possible. Perhaps SF would have deviated, although its inability to recognize a blocked position as a draw might cripple it in this example.

So we'll call the original position drawn, unless White can do better than the line just investigated.

My move is Ra7.

[d]4q2k/r1p3b1/Pp1r4/1PpPp1p1/2N1PpPp/1R5P/4QP2/1R4K1 w - - 0 21

Now everything is blocked.

I do not mind against suggesting improvements for white, but I severely doubt there are such.

Without a blocked position, SF eval immediately after f4 was played was some 80cps white edge, a wrong one, and by a large margin, so it is not simply a question of not recognizing blocked positions.

The pointy black chain plays an enormous role, even though black has very bad bishop, and even though white pressures strongly on the queen side. What fails SF is piece blocking eval, as well as pointy chain score.

[Lyudmil Tsvetkov]

[
You see, SF closes the game with g2-g4! It is afraid of black playing g5-g4 and getting advantage. That was my initial assessment of the position: a small, some 30cps black edge, if white plays g4, it is a draw, otherwise, black will play g4 itself and get winning chances.

Some time ago Lucas called my chess a 'cowardly approach', trying to close the position. And what is SF doing now? It closes the position itself. Is not this a cowardly approach?

Scuse me, Lyudmil, even risking a Rote Karte again, here I have to hop in for a very short notice once again yet:
As for my POV it's you closing the position with ..h4 after SF's g4, isn't it?
You could have gone on with your glorious Kingside attack with hxg4 instead too or simply play anything else, at least leaving it to SF again to close the position (on the Kingside only) or not to close it, couldn't you?
And my bet would have been in that case, Stockfish wouldn't have closed it, if you would have left him the choice.


And just one more little output of SF in MV mode of the position after g4 to illustrate that:

Code: Select all

r1n4k/2p1r1b1/Pp4q1/1PpPp1pp/2N1PpP1/2R4P/4QP1N/1R4K1 b - g3 0 1

Analysis by Stockfish 181114 64 POPCNT:

1. +-  (1.74): 29...h4 30.d6 cxd6 31.Rd1 Rea7 32.Nf3 Qf6 33.Rb3 Rb8 34.Rbd3 Bf8 35.Qa2 Be7 36.Nfd2 Bd8 37.Nb1 Bc7 38.Nc3 Qe6 39.Nd5 Bd8 40.Qe2 Kg7 41.Qb2 Rba8 42.Qa2 Qf7 43.Qd2 Qe6 44.Qc3 Rb8 45.f3 Rba8 46.R1d2 Rf7 47.Qb3 Kh7 48.Rd1 
2. +-  (2.44): 29...Qe8 30.gxh5 Bf6 31.Qg4 Nd6 32.Nxd6 cxd6 33.Nf3 Rh7 34.Nd2 Bd8 35.Nc4 Bc7 36.Kh2 Qxh5 37.Rg1 Qxg4 38.Rxg4 Rh5 39.h4 Rxh4+ 40.Rxh4+ gxh4 41.Rh3 Kg7 42.Rxh4 Rb8 43.Kg2 Ra8 44.Kf3 Rc8 45.Rh1 Re8 46.Ke2 Kg6 47.Kd3 Ra8 48.Rh4 Kg7 49.Rh3 Kg6 50.Rh1 Kg7 51.Ke2 Kg6 52.Rh4 
3. +-  (2.44): 29...Bf6 30.gxh5 Qe8 31.Qg4 Nd6 32.Nxd6 cxd6 33.Nf3 Rh7 34.Nd2 Bd8 35.Nc4 Bc7 36.Kh2 Qxh5 37.Rg1 Qxg4 38.Rxg4 Rh5 39.h4 Rxh4+ 40.Rxh4+ gxh4 41.Rh3 Kg7 42.Rxh4 Rb8 43.Kg2 Ra8 44.Kf3 Rc8 45.Rh1 Re8 46.Ke2 Kg6 47.Kd3 Ra8 48.Rh4 Kg7 49.Rh3 Kg6 50.Rh1 Kg7 51.Ke2 Kg6 52.Rh4 Kg7 53.Kd3 
4. +-  (2.45): 29...hxg4 30.hxg4 Bf6 31.Rh3+ Rh7 32.Rxh7+ Qxh7 33.Nf3 Na7 34.Kg2 Qe7 35.Qd3 Kg7 36.d6 Qd7 37.Nfxe5 Bxe5 38.Nxe5 Qxd6 39.Qxd6 cxd6 40.Nc4 Rb8 41.Nxd6 Kf6 42.Nc4 Kf7 43.Rh1 Nxb5 44.Rh7+ Kg8 45.Rh6 Nc7 46.Rg6+ Kf7 47.Ne5+ Ke7 48.a7 Ra8 49.Rxb6 Ne6 50.Rb8 Nc7 

g4 blocks the g5 pawn, so SF is the first to start blocking the position, I only followed suit.

Unfortunately, both g4 for white and h4 for black seem obligatory, as otherwise either black or white gets an advantage.

To tell you the truth, I feel so calm now with my moves, when I do not look at SF output. Whenever I look at SF output and follow its advice, disaster strikes as in the previous instances with this position. One should play more on his own. In any case, if someone does not prove otherwise, both SF and Komodo evals after 15...f4 seem to be quite off target, close to the full pawn white edge margin.

[peter]

Now the question is, can Black force this line of play to occur? At several points we followed sequences of moves from a PV, as if all the moves in the sequence were best possible.

That's a good remark, Louis, I even wanted to say something like this by myself, but on the other hand didn't want to bother Lyudmil even more than I already did.

Instead I watched the moves of the game with my own SF and here and there had the impression SF would have choosen other moves instead of the ones of its own PV from some positions before.
But that obviously was part of the rules Lyudmil and you had from the start, for my POV it's not quite the way to play an engine on its full strength.
But if you two had wanted to do so, you shouldn't have shown the full output of SF at every move at all.
To see the variants an engine computes as PV helps a lot of course to make own ones accordingly, even if it's not possible to choose as many moves of them as already played as much you want. At least this relativates the compuing- time drastically.

With this said I guess I have fully deserved my final Rote Karte.

[yanquis1972]

i think its a very valid point, as i find the pv line is not worth a great deal unless the moves are forced -- especially after 2 or so moves (4 half moves).

second, lyudmil: you did very clearly state that 14...f4 was a winning move. ive certainly yet to see any evidence that black is even better at all.

[peter]

ive certainly yet to see any evidence that black is even better at all.

...after just having saved the draw by a tiny edge just before it would have costed one more full point again.

[Lyudmil Tsvetkov]

Now the question is, can Black force this line of play to occur? At several points we followed sequences of moves from a PV, as if all the moves in the sequence were best possible.

That's a good remark, Louis, I even wanted to say something like this by myself, but on the other hand didn't want to bother Lyudmil even more than I already did.

Instead I watched the moves of the game with my own SF and here and there had the impression SF would have choosen other moves instead of the ones of its own PV from some positions before.
But that obviously was part of the rules Lyudmil and you had from the start, for my POV it's not quite the way to play an engine on its full strength.
But if you two had wanted to do so, you shouldn't have shown the full output of SF at every move at all.
To see the variants an engine computes as PV helps a lot of course to make own ones accordingly, even if it's not possible to choose as many moves of them as already played as much you want. At least this relativates the compuing- time drastically.

With this said I guess I have fully deserved my final Rote Karte.

Gelbe Karte this time.

I think the reason for the result is the approach taken. If I ad played an open game, I would have been smashed under equal conditions. With closed positions, however, I have my chances.

I think a relatively strong human player with the right strategical approach could easily win or draw a match against SF, no matter how many cores it uses, in case the TC is at least 1 hour for the game. You just need to choose closed positions. In this case, the play is just on 2 results: either a draw, if the engine does not make mistakes, or the human wins, if the engine makes a serious positional mistake. On the other hand, in open game the engine will outplay, outsearch and outcalculate you, no doubt about that.