mmt Vs. Ovyron (G4 D5 BG2)

[SheikhYerbouti]

Stockfish 13 as in 100 elo better than current SF may never happen. Because SF will typically underprune in some areas, overprune in others where for instance it can advantage a position in one area of a game but adversely affect it in a similar area. Enter patches that also appear to conflict, judging by many regression tests, and you've got a program that generalizes too much. Contrast that with theoretically optimal neural net. It would appear the only thing holding Leela and deep learning programs back is unfamiliarity with certain positions, crappy openings and more abstract positions in the late-mid game and endgame.

[jp]

[d]5B2/1n6/1k6/3K4/B7/8/8/8 w - - 0 1

Can someone, anyone, with a fast machine please run SFdev (no tablebases of course!) on this and report when, if ever, SF finds the win? There are only 5 pieces, so I don't think it'll take too long on a fast machine to reach depths of 90 or 100+. Finding the win means it works out how to capture the knight, so it'll show that in the PV and presumably the score will shoot off.


This position is taken from another thread. There are over a dozen moves here that won't spoil the win, but that doesn't make it easy for engines. SF10 fails (depth 91, eval +3.65).

[mmt]

1. g4 d5 2. Bg2 Bxg4 3. c4 c6 4. Qb3 e6 5. Qxb7 Nd7 6. Nc3 Ne7 7. cxd5 exd5 8. d4 Rb8 9. Qa6 Rb6 10. Qd3 Ng6 11. h3 Be6 12. Nf3 Bd6 13. h4 h5 14. b3 Nf6 15. Bg5 O-O 16. e3 Re8 17. Kf1 Bg4 18. Ne1 Bb4 19. Na4 Rb8 20. Nc2 Be7 21. f3 Be6 22. Nc5 Bc8 23. Kf2 Nd7 24. Ne6 Qa5 25. Bxe7 Rxe7 26. b4 Qb6 27. Ng5 Ba6 28. Qa3 Rbe8 29. Bf1 Bxf1 30. Raxf1 Qc7 31. Qd3 a5 32. a3 axb4 33. axb4 Qd6 34. Rfg1 Nb6 35. Qf5 Nf8 36. Re1

[d]4rnk1/4rpp1/1npq4/3p1QNp/1P1P3P/4PP2/2N2K2/4R2R b - - 1 1
LC0 at -2.86. SF at -3.80. Resign once LC0 get below -3.5 and SF below -4.5 I think?

[Ovyron]

Trust me, I'm a centaur!"

Except I don't care if people trust me or believe what I say. Yesterday I had Italian sausage pizza and watched the Interview With The Vampire movie. I can make that claim without evidence or proof and if people don't believe it it's their problem, I know those things are true. With chess it's not different.

[mmt]

If Nc4 then Nh3

[Ovyron]

(I promise to accept mmt's conditional)

[Ovyron]

LC0 at -2.86. SF at -3.80. Resign once LC0 get below -3.5 and SF below -4.5 I think?

I resigned my game against Harvey when my potential draw lines hit 0, even the move before I had hope (instead of resigning I sent a long conditional to at least choose my losing position, but that was just for fun as at that point I knew I was lost). I have no idea how many potential draw lines remaining you have.

1. g4 d5 2. Bg2 Bxg4 3. c4 c6 4. Qb3 e6 5. Qxb7 Nd7 6. Nc3 Ne7 7. cxd5 exd5 8. d4 Rb8 9. Qa6 Rb6 10. Qd3 Ng6 11. h3 Be6 12. Nf3 Bd6 13. h4 h5 14. b3 Nf6 15. Bg5 O-O 16. e3 Re8 17. Kf1 Bg4 18. Ne1 Bb4 19. Na4 Rb8 20. Nc2 Be7 21. f3 Be6 22. Nc5 Bc8 23. Kf2 Nd7 24. Ne6 Qa5 25. Bxe7 Rxe7 26. b4 Qb6 27. Ng5 Ba6 28. Qa3 Rbe8 29. Bf1 Bxf1 30. Raxf1 Qc7 31. Qd3 a5 32. a3 axb4 33. axb4 Qd6 34. Rfg1 Nb6 35. Qf5 Nf8 36. Re1 Nc4 37. Nh3 Ra7

[d]4rnk1/r4pp1/2pq4/3p1Q1p/1PnP3P/4PP1N/2N2K2/4R2R w - -

From the human perspective: Is white about to be up on material?!

[jp]

LC0 at -2.86. SF at -3.80. Resign once LC0 get below -3.5 and SF below -4.5 I think?

No, please don't resign too early. It should be obvious to a computerless human. Right now, Komodo doesn't even see a big problem.

[Ovyron]

I don't have a problem playing till mate, the slowest mate I could find to make a statement

Plot twist: 37...Ra7?? throws the game away and now it's a trivial draw

[jp]

4rnk1/r4pp1/2pq4/3p1Q1p/1PnP3P/4PP1N/2N2K2/4R2R w - -

Komodo 13.2, depth 28 [29]: -1.70 (Re2) [-1.72].