Ethereal 12 (3400) loses to God! (Most Amazing Game I've Seen)

[jp]

Single-threaded Ethereal 12.04 finds Nf5 and sticks to it in 31ms at depth 11 here.

Thanks. That's interesting. So Ethereal 12.04 find it and sticks to it at depth 11, Komodo 13.2 at depth 13, SF10 at depth 15 (which it possibly reaches at least as quickly as the other engines reach lower depths).

I wonder why Ethereal did not play it in this game. After the 18th-move blunder, the pgn says "{2.31/22 15}", which I'm guessing is eval 2.31, depth 22.

There could be a relevant difference between 12.00 and 12.04, though I'm doubtful.

Yes, agree with both. If it's a bug that's not reproducible, it won't be easy to know what happened.
(I wonder what the "15" means. Time?)

[MikeGL]

I have been studying and playing chess seriously for 20 years, and in all this time, I have NEVER seen a game like this.

I have been very impressed with Ethereal 12, it is very, very strong and now with a more refined positional style to boot.

But the way Lc0, even on my very weak hardware (laptop with built-in Nividia card) beat it from a position where Ethereal thought it was +4 is the most amazing thing I've EVER seen.

Witness this amazing game.

[pgn][Event "Elites Sharp, Blitz 3min+2sec-1"]
[Site "?"]
[Date "2020.04.03"]
[Round "2.7"]
[White "Ethereal 12.00 (PEXT)"]
[Black "Lc0 v0.23.2+git.c8d9095, 58462."]
[Result "0-1"]
[ECO "A55"]
[Annotator "1.34;1.10"]
[PlyCount "154"]
[EventDate "2020.04.02"]
[EventType "tourn"]
[SourceTitle "Fritz Engine Tournament"]
[Source "Doe"]

{Intel(R) Core(TM) i7-8565U CPU @ 1.80GHz 1992 MHz W=19.9 plies; 2,081kN/s;
239,128 TBAs B=9.5 plies; 4kN/s; 609 TBAs} 1. d4 Nf6 2. c4 d6 3. Nc3 Nbd7 4. e4
e5 5. Nf3 c6 6. Be2 exd4 7. Nxd4 {1.34/16 8} g6 {1.10/7 5 (Ne5)} 8. Bf4 {
1.16/20 22} Nc5 {1.17/10 6} 9. f3 {1.54/21 17} Nh5 {1.25/12 7} 10. Be3 {
1.50/20 6} Bg7 {1.20/12 1} 11. g3 {1.30/19 9 (g4)} a5 {0.97/9 12 (0-0)} 12. Qd2
{1.47/20 11} a4 {1.02/10 2} 13. O-O-O {1.35/22 14} Qe7 {0.99/10 7 (Qa5)} 14. g4
{1.84/20 9} Nf6 {1.05/12 3} 15. Bf4 {1.66/20 7 (Nc2)} Bxg4 {1.55/10 13} 16. Qe3
{1.87/22 7 (Nc2)} O-O {1.45/9 5} 17. fxg4 {1.80/22 6} a3 {1.11/11 4 (Nfxe4)}
18. b3 {2.31/22 15} Nfxe4 {0.97/13 4 (Rfe8)} 19. Bf3 {2.10/22 9 (Nxe4)} Nxc3 {
0.80/14 6} 20. Qxe7 {2.22/23 5} Nxa2+ {1.03/15 3} 21. Kb1 {2.08/23 4 (Kd2)} Nb4
{0.08/11 7} 22. Qe3 {2.00/23 13} a2+ {0.36/11 12 (Ne6)} 23. Kb2 {2.31/24 7} Ne6
{0.18/19 3} 24. Bxd6 {2.26/21 3 (Bh6)} c5 {-0.13/10 5 (a1Q+)} 25. Bxc5 {
2.80/23 12 (Bxb7)} Nxc5 {-0.24/11 6} 26. Kc3 {2.70/22 3} Nba6 {0.17/14 5 (Nc6)}
27. Bd5 {2.65/20 11 (b4)} Nc7 {-0.38/13 6} 28. Kc2 {2.04/21 12} Nxd5 {-0.27/13
4} 29. cxd5 {2.10/21 7} Na6 {-0.38/11 11 (Rfe8)} 30. Kb2 {2.28/19 6} Nc7 {
-0.60/10 4 (Nb4)} 31. Rd2 {2.02/19 8 (Ka1)} Nxd5 {-0.69/9 8 (Rfd8)} 32. Qg3 {
1.06/20 6 (Qe4)} Nb4 {-0.96/9 7 (Rfd8)} 33. Ra1 {1.09/21 2 (Rhd1)} Rad8 {
-0.74/9 6 (Rfd8)} 34. Qc3 {2.32/20 3 (Rxa2)} Nc6 {-1.34/14 5} 35. Rad1 {
1.55/20 4 (Rxa2)} Nxd4 {-0.39/10 10 (Ra8)} 36. Rxd4 {5.00/18 2} Rc8 {-0.28/9 5}
37. Qb4 {5.13/19 1} Ra8 {-0.21/9 2} 38. Ka1 {4.78/19 2} Rfe8 {-0.13/8 9 (Rfd8)}
39. Qc3 {4.84/16 2 (g5)} Rec8 {-0.07/7 9 (h6)} 40. Qd2 {4.44/19 5 (Qb2)} Rd8 {
-0.03/6 7 (Re8)} 41. Qb2 {5.00/19 4 (Qf4)} Rdc8 {0.11/7 5 (Re8)} 42. g5 {
5.51/19 3} Rf8 {0.28/7 5} 43. Rf1 {5.62/18 2 (h4)} Rfe8 {0.17/6 4 (h6)} 44. Qc3
{4.80/18 2 (b4)} Rac8 {0.05/8 5} 45. Qd2 {4.80/20 2} Rc2 {0.24/10 3 (Re2)} 46.
Qf4 {5.60/19 3} Kh8 {0.25/9 4} 47. h4 {5.43/19 2} Rce2 {0.14/9 6} 48. Rfd1 {
5.70/20 2} Kg8 {0.09/8 3} 49. Qd6 {5.80/20 2 (h5)} Be5 {0.02/7 9 (Ra8)} 50. Qd5
{5.82/20 3 (Qc5)} Rc8 {-0.32/8 4 (Rf2)} 51. Qd7 {5.59/19 2 (b4)} Rcc2 {-0.36/9
4 (Ra8)} 52. Rf1 {4.25/19 2 (Qd5)} Rc7 {-0.42/10 3} 53. Qd8+ {1.98/18 2 (Qd5)}
Kg7 {-0.56/8 0} 54. Rfd1 {0.97/20 3 (Qd5)} Rcc2 {-0.71/9 6 (Rc6)} 55. Qd5 {
0.67/22 2} h5 {-0.62/10 3 (Kh8)} 56. gxh6+ {3.52/17 2} Kxh6 {-0.54/10 2} 57.
Rh1 {3.98/19 2 (Qd8)} b6 {-0.77/8 10 (Rb2)} 58. Rhd1 {4.24/19 3 (h5)} b5 {
-0.83/7 3 (Rb2)} 59. Rh1 {4.29/18 2 (h5)} Rc8 {-0.91/7 3} 60. Rhd1 {3.53/17 2
(Rf1)} b4 {-0.83/8 3 (Rcc2)} 61. Rf1 {0.01/23 2 (h5)} Rcc2 {-0.51/8 3 (Rc7)}
62. h5 {0.01/25 3} g5 {-0.51/9 1} 63. Rh1 {0.01/25 2 (Qd8)} Rc3 {-0.82/7 3
(Rc7)} 64. Qd8 {0.00/24 2 (Rf1)} Rf3 {-1.01/7 3 (Kh7)} 65. Rg1 {0.00/21 3
(Rhd1)} g4 {-1.68/7 1 (Kh7)} 66. Qb6+ {-5.19/19 2 (Qf8+)} Kh7 {-1.86/8 2} 67.
Qc5 {-5.73/18 2} g3 {-2.35/8 1} 68. Qxe5 {-6.09/17 2 (Qc4)} Rxe5 {-5.88/8 2}
69. Rg4 {-6.48/18 3} f5 {-6.23/10 3 (Ree3)} 70. R4xg3 {-5.35/17 2} Rxg3 {
-6.69/11 1} 71. Rxg3 {-5.52/19 2} Kh6 {-7.40/10 2} 72. Rg1 {-9.61/18 2 (Rh3)}
Kxh5 {-9.51/7 3 (Ra5)} 73. Rf1 {-10.65/19 3} Rb5 {-10.47/7 2 (Ra5)} 74. Kxa2 {
-5.60/17 2 (Kb2)} Kg4 {-11.40/7 3} 75. Rg1+ {-8.64/18 2 (Kb2)} Kf3 {-12.53/6 3}
76. Rg8 {-10.47/20 2 (Rg5)} f4 {-16.14/6 2} 77. Rf8 {-12.64/18 3 (Rg1)} Kg4 {
-16.18/6 3 (Kg3)} 0-1[/pgn]

To do this to a weak engine is one thing...but to Ethereal?

With a BISHOP vs a queen?

What the hell is going on here? Amazing.

Very nice. Thanks for sharing.

[yurikvelo]

Can you explain in plain language why it is a blunder? I can't and I'm a chess coach.

Otherwise drawn/won game where after 1 move it become inevitably lost/drawn and there was at least 1 better move, which keep draw/win.

Code: Select all

=0,00	64.Rhd1 Rce3 65.Rf1 Re4 66.Qc6+ Kxh5 67.Rh1+ Kg4
=0,00	64.Rf1 Rcc2 65.Qd8 Rf2 66.Rxf2 Rxf2 67.Qh8+ Bxh8
-11,20	64.Qxe5 Rxe5 65.Kxa2 Re2+ 66.Ka1 Rxb3 67.Rf1 Re6
-12,66	64.Rg1 Rf3 65.Qc5 Rf4 66.Qxe5 Rxe5 67.Rd6+ f6
-46,08	64.Qd8 Rf3 65.Qd5 Rf4 66.Rhd1 Kh7 67.Qxe5 Rxe5		<----- Ethereal blunder
-47,96	64.Qd6+ Bxd6 65.Rxd6+ Re6 66.Rd8 Ra6 67.Rg1 Rxb3
-63,84	64.Rd2 Rc5+ 65.Kxa2 Rxd5 66.Rxe2 Ra5+ 67.Kb1 Ra1+
-148,92	64.Qe4 Rc1+ 65.Rxc1 Rxe4 66.Kxa2 Rxd4 67.Rc6+ Kh7
-M20	64.Rb1 Rd3 65.Qxe5 axb1R+ 66.Kxb1 Rxb3+ 67.Ka1 Rxe5

[Dann Corbit]

Can you explain in plain language why it is a blunder? I can't and I'm a chess coach.

Otherwise drawn/won game where after 1 move it become inevitably lost/drawn and there was at least 1 better move, which keep draw/win.

Code: Select all

=0,00	64.Rhd1 Rce3 65.Rf1 Re4 66.Qc6+ Kxh5 67.Rh1+ Kg4
=0,00	64.Rf1 Rcc2 65.Qd8 Rf2 66.Rxf2 Rxf2 67.Qh8+ Bxh8
-11,20	64.Qxe5 Rxe5 65.Kxa2 Re2+ 66.Ka1 Rxb3 67.Rf1 Re6
-12,66	64.Rg1 Rf3 65.Qc5 Rf4 66.Qxe5 Rxe5 67.Rd6+ f6
-46,08	64.Qd8 Rf3 65.Qd5 Rf4 66.Rhd1 Kh7 67.Qxe5 Rxe5		<----- Ethereal blunder
-47,96	64.Qd6+ Bxd6 65.Rxd6+ Re6 66.Rd8 Ra6 67.Rg1 Rxb3
-63,84	64.Rd2 Rc5+ 65.Kxa2 Rxd5 66.Rxe2 Ra5+ 67.Kb1 Ra1+
-148,92	64.Qe4 Rc1+ 65.Rxc1 Rxe4 66.Kxa2 Rxd4 67.Rc6+ Kh7
-M20	64.Rb1 Rd3 65.Qxe5 axb1R+ 66.Kxb1 Rxb3+ 67.Ka1 Rxe5

True, but finding the right move involves a very deep search with tablebase files. I suppose that technically speaking it is a blunder. But I guess that Fischer, Alekhine, and Capablanca might have made the same move after a good long think.

[Milos]

Can you explain in plain language why it is a blunder? I can't and I'm a chess coach.

Otherwise drawn/won game where after 1 move it become inevitably lost/drawn and there was at least 1 better move, which keep draw/win.

Code: Select all

=0,00	64.Rhd1 Rce3 65.Rf1 Re4 66.Qc6+ Kxh5 67.Rh1+ Kg4
=0,00	64.Rf1 Rcc2 65.Qd8 Rf2 66.Rxf2 Rxf2 67.Qh8+ Bxh8
-11,20	64.Qxe5 Rxe5 65.Kxa2 Re2+ 66.Ka1 Rxb3 67.Rf1 Re6
-12,66	64.Rg1 Rf3 65.Qc5 Rf4 66.Qxe5 Rxe5 67.Rd6+ f6
-46,08	64.Qd8 Rf3 65.Qd5 Rf4 66.Rhd1 Kh7 67.Qxe5 Rxe5		<----- Ethereal blunder
-47,96	64.Qd6+ Bxd6 65.Rxd6+ Re6 66.Rd8 Ra6 67.Rg1 Rxb3
-63,84	64.Rd2 Rc5+ 65.Kxa2 Rxd5 66.Rxe2 Ra5+ 67.Kb1 Ra1+
-148,92	64.Qe4 Rc1+ 65.Rxc1 Rxe4 66.Kxa2 Rxd4 67.Rc6+ Kh7
-M20	64.Rb1 Rd3 65.Qxe5 axb1R+ 66.Kxb1 Rxb3+ 67.Ka1 Rxe5

True, but finding the right move involves a very deep search with tablebase files. I suppose that technically speaking it is a blunder. But I guess that Fischer, Alekhine, and Capablanca might have made the same move after a good long think.

Lol, it's not a blunder it's a bug. Most of engines see it in milliseconds.
And Fisher, Alekhine or Capablanca are total patzers compared to the two engines from this game. It's like saying that some club player would miss something that took some effort for Carlsen to see.

[chrisw]

Can you explain in plain language why it is a blunder? I can't and I'm a chess coach.

Otherwise drawn/won game where after 1 move it become inevitably lost/drawn and there was at least 1 better move, which keep draw/win.

Code: Select all

=0,00	64.Rhd1 Rce3 65.Rf1 Re4 66.Qc6+ Kxh5 67.Rh1+ Kg4
=0,00	64.Rf1 Rcc2 65.Qd8 Rf2 66.Rxf2 Rxf2 67.Qh8+ Bxh8
-11,20	64.Qxe5 Rxe5 65.Kxa2 Re2+ 66.Ka1 Rxb3 67.Rf1 Re6
-12,66	64.Rg1 Rf3 65.Qc5 Rf4 66.Qxe5 Rxe5 67.Rd6+ f6
-46,08	64.Qd8 Rf3 65.Qd5 Rf4 66.Rhd1 Kh7 67.Qxe5 Rxe5		<----- Ethereal blunder
-47,96	64.Qd6+ Bxd6 65.Rxd6+ Re6 66.Rd8 Ra6 67.Rg1 Rxb3
-63,84	64.Rd2 Rc5+ 65.Kxa2 Rxd5 66.Rxe2 Ra5+ 67.Kb1 Ra1+
-148,92	64.Qe4 Rc1+ 65.Rxc1 Rxe4 66.Kxa2 Rxd4 67.Rc6+ Kh7
-M20	64.Rb1 Rd3 65.Qxe5 axb1R+ 66.Kxb1 Rxb3+ 67.Ka1 Rxe5

White could maybe draw, but not win. Isn’t the main feature if this game that Leela/AZ-MCTS technology saw the drawn theme way way back, played for it, and held the state that was the draw. The AB program, essentially evaluating each node as if it had no past, was just a stand alone node, had no idea of flow. AB program had no idea it was stuck and had nothing it could do. That’s the level of the difference, simply orders of magnitude of more “understanding” of the GAME, as a result of understanding the flow of the game.

[jp]

That’s the level of the difference, simply orders of magnitude of more “understanding” of the GAME, as a result of understanding the flow of the game.

If you insist on the flow of the whole game, then obviously we have to say that Leela was totally busted by move 18. Then, Ethereal blundered terribly, possibly because of a bug, failing to find the move that it otherwise finds (at least in v12.04) at depth 11 in 31ms on a single thread.

Even after blundering terribly, Ethereal still had a probable win, certainly not a loss, but managed to stuff up even then.

It seems you have a stronger case if you focus e.g. just on the position at move 42 or whatever, not the whole game, which is full of blunders.

[Milos]

White could maybe draw, but not win. Isn’t the main feature if this game that Leela/AZ-MCTS technology saw the drawn theme way way back, played for it, and held the state that was the draw. The AB program, essentially evaluating each node as if it had no past, was just a stand alone node, had no idea of flow. AB program had no idea it was stuck and had nothing it could do. That’s the level of the difference, simply orders of magnitude of more “understanding” of the GAME, as a result of understanding the flow of the game.

Oh man, how do you not get tired of constantly repeating that nonsense???
White missed/blundered/bugged a win at least 10 times in this game.
The only one who has no idea about anything seems to be you. Your knowledge of chess seems to be frozen in time when you wrote your last engine 20+ years ago. Or 1500 Elo ago...

[jp]

Can you explain in plain language why it is a blunder? I can't and I'm a chess coach.

Otherwise drawn/won game where after 1 move it become inevitably lost/drawn and there was at least 1 better move, which keep draw/win.

True, but finding the right move involves a very deep search with tablebase files. I suppose that technically speaking it is a blunder. But I guess that Fischer, Alekhine, and Capablanca might have made the same move after a good long think.

But the blunder or bug on move 18 involves no deep search (only depth 11), no tablebases, is both technically and non-technically a blunder and can be explained in plain language to a child.

Without this blunder on move 18, there would be no blunder on move 64 to talk about.

[Dann Corbit]

Can you explain in plain language why it is a blunder? I can't and I'm a chess coach.

Otherwise drawn/won game where after 1 move it become inevitably lost/drawn and there was at least 1 better move, which keep draw/win.

True, but finding the right move involves a very deep search with tablebase files. I suppose that technically speaking it is a blunder. But I guess that Fischer, Alekhine, and Capablanca might have made the same move after a good long think.

But the blunder or bug on move 18 involves no deep search (only depth 11), no tablebases, is both technically and non-technically a blunder and can be explained in plain language to a child.

Without this blunder on move 18, there would be no blunder on move 64 to talk about.

After this "Blunder" black is still in trouble after a 44 ply search with millions of tablebase hits
[d]r4rk1/1p2qpbp/2pp1np1/2n5/2PNPBP1/pPN1Q3/P3B2P/2KR3R b - - acd 44; bm Nfxe4; c3 "Nfxe4"; ce -221; pm Nfxe4; pv Nfxe4 Bf3 Nxc3 Qxe7 Nxa2+ Kd2 Rfe8 Qxd6 Bxd4 Qxd4 Nxb3+ Kc2 Nxd4+ Rxd4 Nb4+ Kb3 c5 Rd6 a2 Ra1 Rab8 Bd2 b5 Rd7 bxc4+ Kxc4 Rb6 Kxc5 Reb8 Bxb4 Rb5+ Kd6 R8b6+ Kc7 Rxb4 Rxa2 Rf6 Ra3 Rc4+ Kd8 Rcf4 Bd1 Kg7 h3 Re6 Bf3 Re5 Rc3 Kh6 Rcd3 Kg7 Ra3 Rg5 Bd1 Re5 Kc7 Rc5+ Kd6 Rg5 Rd3 Ra5 Rb7 Rf6+ Kc7 Rf4 Kb6 Re5 Rbd7 Rb4+ Kc7 Rf4 Kc6 Rf6+ Kb7 Rf4 Kc7 Ra5;