KomodoDragon brilliancy Ruy Lopez - Marshall! A marshall reborn?

[MikeGL]

This game is nice because I once thought Marshall was already analyzed to death and
the consensus was that this openning just leads to draw. This is the reason why it is only played by black to hold a draw as seen in top GM's games.

But not this time where SF 12 as white was crushed by black in a Marshall openning.
[pgn]
[Event "CPU engine"]
[Site "?"]
[Date "2020.11.11"]
[Round "8"]
[White "Stockfish 12 200902"]
[Black "KomodoDragon 1.0 avx2"]
[Result "0-1"]
[ECO "C89o"]
[EventDate "2020.??.??"]
[PlyCount "122"]
[TimeControl "180+1"]

1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.O-O Be7 6.Re1 b5 7.Bb3 O-O 8.c3 d5
9.exd5 Nxd5 10.Nxe5 Nxe5 11.Rxe5 c6 12.Re1 Bd6 13.d4 Qh4 14.g3 Qh3 15.Re4
g5 16.Qf1 Qh5 17.Nd2 Nf4 18.gxf4 gxf4 19.Kh1 Kh8 20.Bd1 Qh6 21.Bf3 Rg8 22.
Qd1 Bh3 23.Re1 Qh4 24.Ne4 Bg2+ 25.Bxg2 f3 26.Ng3 Rxg3 27.fxg3 Bxg3 28.h3
fxg2+ 29.Kxg2 Bxe1 30.Qf3 Rg8+ 31.Kf1 Rg6 32.Be3 Rf6 33.Qxf6+ Qxf6+ 34.
Kxe1 Qf3 35.Kd2 f5 36.Re1 f4 37.Bxf4 Qxf4+ 38.Re3 Qh2+ 39.Kc1 Qg1+ 40.Kd2
Qb1 41.h4 Qxb2+ 42.Kd3 Qb1+ 43.Kd2 Qxa2+ 44.Kc1 a5 45.h5 Kg7 46.h6+ Kxh6
47.Rh3+ Kg6 48.Rg3+ Kf5 49.Kd1 a4 50.Rh3 Kg4 51.Re3 a3 52.d5 Qxd5+ 53.Kc2
a2 54.Re1 Qf5+ 55.Kb2 Qf2+ 56.Ka1 Qxe1+ 57.Kxa2 Qxc3 58.Kb1 b4 59.Ka2 b3+
60.Ka3 c5 61.Ka4 Qb4# 0-1
[/pgn]

credit to Stefan Pohl for the game.

[MikeGL]

ok I just checked in my database of Human-Human GM games, 17...Nf4! move is not found, maybe this is a Novelty by engines.
Even SF 12 cannot consider this move (17...Nf4!!) . I wonder if KomodoDragon is NNUE or pure NN similar to Lc0, because this is a very different move which is a bit difficult for A/B engines.
[d]r1b2rk1/5p1p/p1pb4/1p1n2pq/3PR3/1BP3P1/PP1N1P1P/R1B2QK1 b - - 3 17
bm Nf4

[Nay Lin Tun]

What time control and how many cores did you use?
Is it bullet or blitz game?

[lkaufman]

ok I just checked in my database of Human-Human GM games, 17...Nf4! move is not found, maybe this is a Novelty by engines.
Even SF 12 cannot consider this move (17...Nf4!!) . I wonder if KomodoDragon is NNUE or pure NN similar to Lc0, because this is a very different move which is a bit difficult for A/B engines.
[d]r1b2rk1/5p1p/p1pb4/1p1n2pq/3PR3/1BP3P1/PP1N1P1P/R1B2QK1 b - - 3 17
bm Nf4

KomodoDragon is NNUE, not pure NN. If it were pure NN it would need a GPU to be so strong.

[MikeGL]

What time control and how many cores did you use?
Is it bullet or blitz game?

I checked it says: [TimeControl "180+1"]

Can't see the number of cores used. I even can't find the exact PGN file now, because I just download a dozen pgn files and
then unzips all of them then throws/imports them into my native database format.

[MikeGL]

ok I just checked in my database of Human-Human GM games, 17...Nf4! move is not found, maybe this is a Novelty by engines.
Even SF 12 cannot consider this move (17...Nf4!!) . I wonder if KomodoDragon is NNUE or pure NN similar to Lc0, because this is a very different move which is a bit difficult for A/B engines.
[d]r1b2rk1/5p1p/p1pb4/1p1n2pq/3PR3/1BP3P1/PP1N1P1P/R1B2QK1 b - - 3 17
bm Nf4

KomodoDragon is NNUE, not pure NN. If it were pure NN it would need a GPU to be so strong.

Thanks for the info.
This is what I love with online chess forums. Engine author, GM Kaufman himself answers the question within minutes.

[Nay Lin Tun]

Prabably 21. Bf3? was the blunder as suggested my droidfish running 10-15 seconds per position.

[MikeGL]

Prabably 21. Bf3? was the blunder as suggested my droidfish running 10-15 seconds per position.

Maybe there are other alternatives but I think 21.Bf3 move doens't deserve a "?" tag because it is also a strong move.

Code: Select all

29/45  +0.21  Qe1 Bf5 Bf3 Rg8 Re5 Bh3 Nf1 Qg7 Ng3 fxg3 hxg3 Bxe5 Qxe5 Rae8 Qxg7+ Rxg7 Be3 Rg6 a4 Bg4 Bg2 Bf5 axb5 axb5 Ra7 Kg7 d5 cxd5 Bd4+ Kg8 Bxd5 Be6 
29/12   0.00  Bf3 Rg8 Qd1 Bh3 Re1 f5 c4 Qg5 Rg1 Qh4 Re1
29/51  -1.07  Qe2 Rg8 Re8 Bg4 Rxg8+ Rxg8 Qxg4 Rxg4 Bxg4 Qh4 Bf3 Qxf2 Bxc6 Qe1+ Kg2 Qe2+ Kh1 b4 Bf3 Qd3 Kg2 bxc3 bxc3 Qxc3 Rb1 a5 Kf1 Bb4 a3 Bxa3 Bxa3 

[mwyoung]

This game is nice because I once thought Marshall was already analyzed to death and
the consensus was that this openning just leads to draw. This is the reason why it is only played by black to hold a draw as seen in top GM's games.

But not this time where SF 12 as white was crushed by black in a Marshall openning.
[pgn]
[Event "CPU engine"]
[Site "?"]
[Date "2020.11.11"]
[Round "8"]
[White "Stockfish 12 200902"]
[Black "KomodoDragon 1.0 avx2"]
[Result "0-1"]
[ECO "C89o"]
[EventDate "2020.??.??"]
[PlyCount "122"]
[TimeControl "180+1"]

1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.O-O Be7 6.Re1 b5 7.Bb3 O-O 8.c3 d5
9.exd5 Nxd5 10.Nxe5 Nxe5 11.Rxe5 c6 12.Re1 Bd6 13.d4 Qh4 14.g3 Qh3 15.Re4
g5 16.Qf1 Qh5 17.Nd2 Nf4 18.gxf4 gxf4 19.Kh1 Kh8 20.Bd1 Qh6 21.Bf3 Rg8 22.
Qd1 Bh3 23.Re1 Qh4 24.Ne4 Bg2+ 25.Bxg2 f3 26.Ng3 Rxg3 27.fxg3 Bxg3 28.h3
fxg2+ 29.Kxg2 Bxe1 30.Qf3 Rg8+ 31.Kf1 Rg6 32.Be3 Rf6 33.Qxf6+ Qxf6+ 34.
Kxe1 Qf3 35.Kd2 f5 36.Re1 f4 37.Bxf4 Qxf4+ 38.Re3 Qh2+ 39.Kc1 Qg1+ 40.Kd2
Qb1 41.h4 Qxb2+ 42.Kd3 Qb1+ 43.Kd2 Qxa2+ 44.Kc1 a5 45.h5 Kg7 46.h6+ Kxh6
47.Rh3+ Kg6 48.Rg3+ Kf5 49.Kd1 a4 50.Rh3 Kg4 51.Re3 a3 52.d5 Qxd5+ 53.Kc2
a2 54.Re1 Qf5+ 55.Kb2 Qf2+ 56.Ka1 Qxe1+ 57.Kxa2 Qxc3 58.Kb1 b4 59.Ka2 b3+
60.Ka3 c5 61.Ka4 Qb4# 0-1
[/pgn]

credit to Stefan Pohl for the game.

Centipawn analysis is saying B.S. to this game.
There is no reason for the mistake in the game. It should not have been played by a Sober Stockfish 12!

The losing move was just a pure blunder for a engine, maybe a BUG.

[Nay Lin Tun]

Prabably 21. Bf3? was the blunder as suggested my droidfish running 10-15 seconds per position.

Maybe there are other alternatives but I think 21.Bf3 move doens't deserve a "?" tag because it is also a strong move.

Code: Select all

29/45  +0.21  Qe1 Bf5 Bf3 Rg8 Re5 Bh3 Nf1 Qg7 Ng3 fxg3 hxg3 Bxe5 Qxe5 Rae8 Qxg7+ Rxg7 Be3 Rg6 a4 Bg4 Bg2 Bf5 axb5 axb5 Ra7 Kg7 d5 cxd5 Bd4+ Kg8 Bxd5 Be6 
29/12   0.00  Bf3 Rg8 Qd1 Bh3 Re1 f5 c4 Qg5 Rg1 Qh4 Re1
29/51  -1.07  Qe2 Rg8 Re8 Bg4 Rxg8+ Rxg8 Qxg4 Rxg4 Bxg4 Qh4 Bf3 Qxf2 Bxc6 Qe1+ Kg2 Qe2+ Kh1 b4 Bf3 Qd3 Kg2 bxc3 bxc3 Qxc3 Rb1 a5 Kf1 Bb4 a3 Bxa3 Bxa3 

The error is most likely within the series of 5 moves from move 21. to 26. If you send this pgn to Bryan-Stockfish, he may do backward forward analysis and microscope the blunder in more detail.