Page 9 of 51

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 6:30 am
by Ovyron
Alayan wrote: Sat Jan 25, 2020 7:38 pm So, white needs to deviate from this line somewhere. ChessDBCN hasn't found something convincing yet, its score after 7. ... Ne7 has moved to -1.30 after I fed it the refutations to its preferred line.
I managed to get the score up to -0.90 by feeding it my refutations to its black's main attacks.

Looks like you can just pick a side and refute its lines and make its score whatever you want.

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 7:43 am
by jp
Ovyron wrote: Thu Jan 30, 2020 6:30 am
Alayan wrote: Sat Jan 25, 2020 7:38 pm after I fed it the refutations to its preferred line.
feeding it my refutations to its black's main attacks.

Looks like you can just pick a side and refute its lines and make its score whatever you want.
How did you two feed it refutations? Do you just get it to look at the last position of your refuting line, or do you give it all positions in the line?

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 2:53 pm
by Ovyron
jp wrote: Thu Jan 30, 2020 7:43 am How did you two feed it refutations?
Just play the line into it. Either its line will end and you'll get a button to request analysis, and you click it (it says to wait 5 seconds, but it's more than that; but eventually it'll do it), or there will be a position with several moves analyzed, but yours is missing (it appears with some ?? Score) so you make it and push the button and wait (but nothing happens if you just wait, you have to go back and forth between moves until a move has been added.)

If your line is good, ChessDBCN will score it as best because it can't see a way for its side to score better. If not, it'll have refuted your line (but wait! did it really find a move at low depth that destroys your entire line? you must be doing something wrong... but here low depth Stockfish isn't strong enough to find the winning moves against the Grob...)

In practice, this is public information, so it'd be stupid for someone to just upload their strongest lines on there, because if their opponent sees a line that it was missing, you could even not beat them anymore. And that's why I only uploaded moves for the white side, but never the moves I intend to make against mmt.

P.S. - And it's very unstable. When I posted this, it was correct, now it's back to -1.30.

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 4:22 pm
by Alayan
I can confirm you had it pushed down over -1.00. I pushed it back to -1.30 by feeding it counter-analysis.

But it likes Rb5 after Na4 in the h5 line, while my homefish in the end preferred Rb8 (Rb5 was the top move for a few depths). It could be refuted back and forth for a while I think.

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 5:02 pm
by Ovyron
Alayan wrote: Thu Jan 30, 2020 4:22 pm I pushed it back to -1.30 by feeding it counter-analysis.
Ahh. Yeah, I could still drag it up with counter-analysis.

I guess this is fun, it's like playing an indirect correspondence game. But it's clear accurate scores on the DB require "babysitting", and I think its biggest flaw is that in some positions it'll never look for alternatives (those where only a single move appears scored and the rest are ??) I got it back to -1.01 with a single request in such a position where CDBCN could have done it itself if it looked for alternatives on that position, so it's scary to think how stronger its moves could be with a better algorithm (which would make it babysit itself).

I was impressed by the way it expanded some of the lines by itself, but now it's clear it was you :)

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 7:28 pm
by Uri Blass
Ovyron wrote: Thu Jan 30, 2020 2:53 pm
jp wrote: Thu Jan 30, 2020 7:43 am How did you two feed it refutations?
Just play the line into it. Either its line will end and you'll get a button to request analysis, and you click it (it says to wait 5 seconds, but it's more than that; but eventually it'll do it), or there will be a position with several moves analyzed, but yours is missing (it appears with some ?? Score) so you make it and push the button and wait (but nothing happens if you just wait, you have to go back and forth between moves until a move has been added.)

If your line is good, ChessDBCN will score it as best because it can't see a way for its side to score better. If not, it'll have refuted your line (but wait! did it really find a move at low depth that destroys your entire line? you must be doing something wrong... but here low depth Stockfish isn't strong enough to find the winning moves against the Grob...)

In practice, this is public information, so it'd be stupid for someone to just upload their strongest lines on there, because if their opponent sees a line that it was missing, you could even not beat them anymore. And that's why I only uploaded moves for the white side, but never the moves I intend to make against mmt.

P.S. - And it's very unstable. When I posted this, it was correct, now it's back to -1.30.
You assume the target of people is only to beat their opponent.
This assumption may be wrong.

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 9:39 pm
by Ovyron
Uri Blass wrote: Thu Jan 30, 2020 7:28 pm You assume the target of people is only to beat their opponent.
This assumption may be wrong.
If an opponent exists then the target of people is only to beat them. If their target isn't to beat them, then it's not an opponent, it's some kind of "analysis buddy."

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 10:34 pm
by Jouni
Nice, that now also 2. Bg2 is examined! Any ideas how long it takes before 1.g4 is proven to be loss or draw? 10 years? 100 years?

Re: 1.g4 opening is losing?

Posted: Thu Jan 30, 2020 11:03 pm
by Ovyron
Depends on your definition of "proven." We will never have proof (strangely, it's just because we'd not have enough space to store it.)

If you mean like having an algorithm that is able to draw against any attack (if it draws) or win against any defense (if it loses) then we're probably already there with people like Zenmastur or Harvey Williamson being able to produce a winning line against any move by white on the fly (and that's why you wouldn't see them playing the white side, not even them have been able to find a line that draws against their own analysis.) If it's a draw then 1.g4 is not special and the algorithm that is able to draw it would also be able to draw any other drawn position, so your question is like asking how far away are we from proving that chess is a draw with perfect play (...unless it turns out there's in fact a line that wins for white...)

For comparison, everyone should know that 1.g4 is not a forced win for white, but that also can't be proven! :mrgreen:

Re: 1.g4 opening is losing?

Posted: Wed Feb 05, 2020 6:34 pm
by fastgm
Here is a game, Stockfish 11 vs Stockfish 11, default settings, every move till depth 50, 1.g4 d5

[pgn][Event "Match - 1.g4 d5 - Stockfish 11 depth 50"]
[Site "FastGM"]
[Date "2020.01.22"]
[Round ""]
[White "Stockfish 11"]
[Black "Stockfish 11"]
[Result "0-1"]
[ECO "A00"]
[Opening "Grob's attack"]
[PlyCount "174"]
[TimeControl "inf"]

1. g4 {book} d5 {book} 2. g5 {-0.93/50 26812s} e5 {+1.68/50 23962s}
3. Bg2 {-1.00/50 12146s} Nc6 {+1.72/50 6973s} 4. d4 {-1.01/50 6076s}
exd4 {+1.60/50 2872s} 5. Nf3 {-1.09/50 7516s} Bc5 {+1.48/50 6526s}
6. Nbd2 {-0.93/50 6235s} Bb6 {+1.53/50 8610s} 7. Nb3 {-1.13/50 8374s}
Nge7 {+1.71/50 5848s} 8. a4 {-1.16/50 5853s} Na5 {+1.43/50 4947s}
9. Nfxd4 {-1.18/50 6581s} c5 {+1.34/50 5300s} 10. Nf3 {-1.20/50 6389s}
Nxb3 {+1.56/50 9386s} 11. cxb3 {-1.39/50 6856s} O-O {+1.68/50 9839s}
12. h4 {-1.46/50 6754s} Nc6 {+1.71/50 9540s} 13. h5 {-1.58/50 5159s}
Bg4 {+1.94/50 2612s} 14. Rh4 {-1.26/50 5800s} Qd7 {+1.85/50 4209s}
15. Nh2 {-1.37/50 7797s} Be6 {+2.12/50 5459s} 16. Bf4 {-1.60/50 11925s}
Rae8 {+2.01/50 10981s} 17. Kf1 {-1.61/50 4894s} Bd8 {+1.94/50 5155s}
18. Nf3 {-1.75/50 2569s} Bf5 {+2.28/50 16987s} 19. Kg1 {-1.79/50 8266s}
f6 {+2.08/50 4760s} 20. g6 {-1.96/50 5603s} hxg6 {+2.47/50 3769s}
21. hxg6 {-1.96/50 2189s} Bxg6 {+2.46/50 3628s} 22. Ne1 {-2.06/50 1266s}
Be4 {+2.84/50 2861s} 23. Bd6 {-1.92/50 665s} f5 {+3.04/50 2598s}
24. Bxf8 {-2.93/50 2629s} Bxh4 {+3.40/50 3118s} 25. Bxc5 {-3.21/50 2185s}
d4 {+3.40/50 2095s} 26. a5 {-3.23/50 1572s} Re6 {+3.79/50 2342s}
27. b4 {-3.43/50 595s} Rg6 {+3.96/50 1778s} 28. f3 {-3.08/50 2018s}
Bxe1 {+4.02/50 2011s} 29. fxe4 {-3.55/50 856s} fxe4 {+4.40/50 710s}
30. Qxe1 {-3.83/50 529s} Qg4 {+4.67/50 555s} 31. Qf1 {-3.89/50 277s}
Rf6 {+4.78/50 423s} 32. Qc1 {-4.35/50 1117s} Qg3 {+4.82/50 879s}
33. Ra3 {-3.88/50 1005s} e3 {+4.90/50 675s} 34. Rxe3 {-4.44/50 384s}
dxe3 {+4.84/50 382s} 35. Qxe3 {-4.62/50 292s} Qe1+ {+5.14/50 2626s}
36. Kh2 {-4.62/50 245s} Rh6+ {+5.57/50 3483s} 37. Bh3 {-4.62/50 371s}
Qh4 {+5.58/50 2195s} 38. Qb3+ {-4.73/50 572s} Kh8 {+5.58/50 266s}
39. Qf3 {-4.73/50 37s} Nd4 {+5.69/50 543s} 40. Bxd4 {-4.73/50 78s}
Qxd4 {+5.70/50 603s} 41. Qf8+ {-4.73/50 127s} Kh7 {+5.82/50 478s}
42. Qf5+ {-5.38/50 1909s} g6 {+5.99/50 1457s} 43. Qf7+ {-6.07/50 3975s}
Qg7 {+6.46/50 2030s} 44. Qf4 {-6.30/50 1708s} Qe7 {+6.58/50 365s}
45. Kg3 {-7.08/50 4676s} Rh5 {+6.58/50 85s} 46. Bg4 {-7.41/50 5499s}
Rb5 {+6.58/50 94s} 47. Qd2 {-7.61/50 3531s} Rxb4 {+7.36/50 2288s}
48. Bf3 {-7.70/50 3718s} Rb5 {+7.36/50 94s} 49. Kf2 {-8.51/50 14269s}
Qc5+ {+8.39/50 4964s} 50. Kf1 {-8.51/50 272s} Rxa5 {+9.64/50 13489s}
51. Qd7+ {-8.74/50 6257s} Kh6 {+10.79/50 17088s} 52. Qh3+ {-10.49/50 37164s}
Kg5 {+11.83/50 19151s} 53. Kg2 {-10.62/50 12291s} Qd4 {+15.51/50 70063s}
54. b3 {-18.75/50 279436s} Re5 {+19.25/50 53109s} 55. Qh7 {-19.81/50 84162s}
b5 {+50.69/50 84708s} 56. Qh1 {-47.55/50 94552s} a5 {+50.69/50 1999s}
57. Kg3 {-58.46/50 10153s} Re3 {+50.69/50 350s} 58. Kg2 {-58.78/50 775s}
a4 {+52.60/50 834s} 59. bxa4 {-147.92/50 4199s} bxa4 {+59.50/50 1035s}
60. Qb1 {-147.92/50 2966s} Re5 {+147.92/50 2239s} 61. Qc1+ {-M62/50 497s}
Kf6 {+M61/50 2010s} 62. Kf1 {-M58/50 308s} Kg7 {+M59/50 137s}
63. Qc7+ {-M54/50 130s} Kh6 {+M55/50 113s} 64. Qb8 {-M52/50 231s}
Qa1+ {+M53/50 176s} 65. Kf2 {-M50/50 20s} Rf5 {+M51/50 74s}
66. Qh2+ {-M48/50 34s} Kg7 {+M49/50 52s} 67. Qc7+ {-M46/50 79s}
Rf7 {+M47/50 50s} 68. Qc4 {-M44/50 59s} a3 {+M45/50 87s} 69. Kg2 {-M42/50 60s}
Qe5 {+M43/50 27s} 70. Qa2 {-M40/50 45s} Qg5+ {+M41/50 4.7s}
71. Kf2 {-M36/50 41s} Qc5+ {+M39/50 14s} 72. Kg2 {-M36/50 24s}
Rf5 {+M37/50 7.8s} 73. Qa1+ {-M34/50 9.5s} Kh7 {+M35/50 2.7s}
74. Qh1+ {-M32/50 5.6s} Kg8 {+M27/50 23s} 75. Kh3 {-M30/50 6.6s}
Qe3 {+M25/50 2.4s} 76. Qd1 {-M24/50 7.8s} Rh5+ {+M23/50 0.77s}
77. Kg2 {-M22/50 0.36s} Rg5+ {+M21/50 0.42s} 78. Kh3 {-M20/50 2.0s}
Qf4 {+M19/50 1.7s} 79. Qd8+ {-M18/50 0.22s} Kg7 {+M17/50 0.20s}
80. Qe7+ {-M16/50 0.22s} Kh6 {+M15/50 0.18s} 81. Qxg5+ {-M14/50 0.14s}
Qxg5 {+M13/50 0.26s} 82. Bd5 {-M12/50 0.24s} Qxd5 {+M11/50 0.26s}
83. e4 {-M10/50 0.41s} Qxe4 {+M9/50 0.48s} 84. Kg3 {-M8/50 0.12s}
a2 {+M7/50 0.12s} 85. Kf2 {-M6/50 0.010s} a1=Q {+M5/50 0.009s}
86. Kg3 {-M4/50 0.003s} Qg1+ {+M3/50 0.003s} 87. Kh3 {-M2/50 0.001s}
Qgg2# {+M1/50 0.003s, Black mates} 0-1[/pgn]