mmt Vs. Ovyron (G4 D5 BG2)

[zullil]

[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).

I doubt that Stockfish would fail to win from this position in a game. Here's one quick test of default Stockfish-dev self-playing. Haven't looked to see where suboptimal moves occurred.

[pgn][White "StockfishDefault"]
[Black "StockfishDefault"]
[Result "1-0"]
[FEN "5B2/1n6/1k6/3K4/B7/8/8/8 w - - 0 1"]
[PlyCount "99"]
[SetUp "1"]
[TimeControl "300"]

1. Bd1 {+3.24/43 5.4s} Nd8 {-3.30/45 4.8s} 2. Ba4 {+3.24/44 2.7s}
Kc7 {-3.30/44 2.4s} 3. Bd6+ {+3.24/44 2.4s} Kb7 {-3.30/48 2.3s}
4. Bb4 {+3.24/44 2.3s} Ka6 {-3.30/45 2.7s} 5. Be7 {+3.24/47 2.4s}
Nb7 {-3.30/45 2.5s} 6. Kc6 {+3.24/48 2.4s} Na5+ {-3.30/47 2.4s}
7. Kc5 {+3.24/50 3.6s} Nb7+ {-3.30/49 2.6s} 8. Kc6 {+3.24/46 2.4s}
Na5+ {-3.30/51 2.5s} 9. Kc5 {+3.24/49 2.4s} Nb7+ {-3.30/52 2.5s}
10. Kd5 {+3.24/50 3.6s} Kb6 {-3.30/52 3.3s} 11. Bg5 {+3.24/45 3.1s}
Kc7 {-3.30/52 3.2s} 12. Bf4+ {+3.24/46 2.4s} Kb6 {-3.30/52 3.1s}
13. Bd2 {+3.24/49 2.4s} Na5 {-3.30/49 2.8s} 14. Bc3 {+3.24/50 2.8s}
Nb7 {-3.30/52 2.6s} 15. Bd7 {+3.24/49 3.1s} Nc5 {-3.30/46 2.6s}
16. Bd4 {+3.24/44 2.6s} Kc7 {-3.30/46 2.6s} 17. Bb5 {+3.24/48 3.1s}
Nd7 {-3.30/48 2.6s} 18. Bc3 {+3.24/47 2.6s} Nb8 {-3.30/49 3.0s}
19. Ba5+ {+3.24/47 3.8s} Kc8 {-55.58/52 8.1s} 20. Bd3 {+55.53/48 2.8s}
Kb7 {-M93/52 8.6s} 21. Kd6 {+55.90/51 3.3s} Nc6 {-M93/49 2.6s}
22. Be4 {+M93/45 3.6s} Ka6 {-M66/51 8.7s} 23. Bd2 {+M91/47 3.0s}
Na5 {-M60/48 2.8s} 24. Bd3+ {+M67/47 3.1s} Kb6 {-M56/51 3.0s}
25. Be3+ {+M59/51 2.9s} Kb7 {-M93/1 0s} 26. Kc5 {+M55/53 11s} Kc7 {-M52/50 7.5s}
27. Bf4+ {+M53/51 4.3s} Kd7 {-M50/49 3.5s} 28. Bf5+ {+M51/52 3.3s}
Kd8 {-M48/49 5.4s} 29. Be4 {+M47/52 5.8s} Kc8 {-M46/48 5.1s}
30. Kb5 {+M43/51 4.8s} Nb3 {-M44/47 3.6s} 31. Be3 {+M41/49 3.0s}
Kd7 {-M42/47 4.2s} 32. Kb4 {+M39/49 5.4s} Ke6 {-M40/44 3.3s}
33. Kxb3 {+M37/48 3.9s} Ke5 {-M38/42 6.8s} 34. Bc2 {+M35/46 5.1s}
Kd5 {-M36/41 4.5s} 35. Kb4 {+M33/45 3.9s} Kd6 {-M34/40 2.8s}
36. Kc4 {+M29/45 4.1s} Ke5 {-M32/42 3.4s} 37. Kc5 {+M27/46 3.9s}
Ke6 {-M30/43 2.9s} 38. Bd4 {+M25/46 3.3s} Kd7 {-M28/46 7.3s}
39. Kd5 {+M23/43 2.8s} Ke7 {-M24/44 3.2s} 40. Bf5 {+M21/47 3.5s}
Ke8 {-M20/45 3.2s} 41. Ke6 {+M19/45 2.7s} Kd8 {-M18/45 2.6s}
42. Kd6 {+M17/46 2.8s} Ke8 {-M16/1 0s} 43. Bg6+ {+M15/46 2.6s}
Kf8 {-M14/43 1.8s} 44. Ke6 {+M13/51 2.8s} Kg8 {-M12/1 0s} 45. Kf6 {+M11/66 2.6s}
Kf8 {-M10/108 1.7s} 46. Bc5+ {+M9/117 2.5s} Kg8 {-M8/1 0s}
47. Be4 {+M7/245 0.95s} Kh8 {-55.63/1 0.004s} 48. Kg6 {+M5/245 0.041s}
Kg8 {-M4/1 0.008s} 49. Bd5+ {+M3/245 0.019s} Kh8 {-M2/1 0.007s}
50. Bd4# {+M1/245 0.014s, White mates} 1-0
[/pgn]

[jp]

I doubt that Stockfish would fail to win from this position in a game. Here's one quick test of default Stockfish-dev self-playing. Haven't looked to see where suboptimal moves occurred.

Thanks, but maybe that's just clueless play as both attacker and defender and its attacking self just then gets lucky as its defending self stumbles into mate.

Can you see whether SFdev makes any headway just analysing from the start position?

Just a quick look at the first few moves of SFdev vs SFdev shows White and Black typically giving up around DTZ4 (difference) each move, so it's okay for White because it's balanced out, but if Black did not give back the DTZ4 then it looks like White might run out of its 50 moves by doing that.

3... Kb7?? is a big practical blunder by Black. Black was lost already, but DTZ went from 86 to 39.
Oh, but now I see White's 4. Bb4?? immediately gives it all back: DTZ back to 88.
(Ergo this is not the critical moment in the "blunderfest".)
4... Ka6?? DTZ down to 29!
Then a few moves later SFdev repeats moves and avoids 3-fold repetition with
10.Kd6?? DTZ went from 35 to 90!! Since they've already played 10 moves at this stage, TB could take over Black here and draw.


Another thing we could try is SFdev vs TB. (Is this easy to set up, i.e. so one side gets the TBs and the other doesn't?)

[zullil]

I doubt that Stockfish would fail to win from this position in a game. Here's one quick test of default Stockfish-dev self-playing. Haven't looked to see where suboptimal moves occurred.

Thanks, but maybe that's just clueless play as both attacker and defender and it's attacking self just then stumbles into mate.

Can you see whether SFdev makes any headway just analysing from the start position?

Already in progress (for too many hours).

Finally something has happened that brings a glimmer of hope ...

info depth 87 seldepth 101 multipv 1 score cp 348 nodes 241019304319 nps 68485537 hashfull 131 tbhits 0 time 3519273 pv f8e7 b6c7 a4e8 b7a5 e7c5 a5b7 c5b4 c7d8 e8b5 d8c7 b5c4 c7b6 b4f8 b6c7 f8h6 b7a5 h6f4 c7b7 c4d3 b7b6 f4e3 b6c7 d3a6 a5b7 e3f4 c7b6 a6e2 b7a5 f4e3 b6c7 d5c5 a5b7 c5b4 b7d8 b4b5 d8e6 e2d3 c7d6 d3c4 e6c7 b5a4 c7d5 e3f2 d6c6 c4b5 c6d6 a4b3 d5f4 f2b6 f4d5 b6d4 d5c7 b5c4 d6c6 d4c3 c6c5 c3f6 c7b5 c4d3 c5c6 d3e4 c6d6 b3b4 b5c7 f6d4 c7d5 b4b5 d5c7 b5b6 c7d5 b6a7 d5c7 a7b8 c7e6 d4b2 d6c5 b8b7 c5c4 e4h7 e6d4 b7b6 d4e2 b2g7 c4d5 b6a7 d5e6 g7b2 e6d7 h7d3 e2f4 d3a6 d7d6 a7b8 f4g6 b2a3 d6d5 a6b7 d5c4
info depth 88 currmove f8e7 currmovenumber 1
info depth 88 seldepth 110 multipv 1 score cp 359 lowerbound nodes 1863627803823 nps 52943145 hashfull 208 tbhits 0 time 35200549 pv f8e7
info depth 87 currmove f8e7 currmovenumber 1

[jp]

Thanks, Louis. (Please see my edits above too. I'm looking at the first few moves of the selfplay game. Okay, now after 6 edits and 10 moves, I think the picture is clear, at least at the short TC you gave SFdev. )

[zullil]

Thanks, Louis. (Please see my edits above too. I'm looking at the first few moves of the selfplay game.)

With TB for Black, White seems to throw away the win rather quickly (at this short time control):

1. Bd1 +3.24/52 15s Kc7 -M93/55 14s
2. Ba4 +3.24/54 8.9s Kb6 -M93/54 8.2s
3. Bd7 +3.24/53 7.3s Kc7 -M93/53 7.2s
4. Be6 +3.24/57 7.1s Kb6 -M93/52 6.6s
5. Bd7 +3.24/60 9.8s Kc7 -0.49/57 7.0s
6. Be6 +3.24/59 8.8s

[jp]

In the SFdev self-play game, 17... Nd7?? seems to be the last critical blunder. DTZ went from 88 (17...Nb3! or Nb7! would do) to 35. (SInce they've both played 17 moves at that stage, Black has a "very easy" draw by the 50-move rule, though the win wasn't ever thrown away without that rule.)

It certainly looks like SFdev at that TC was playing fairly randomly until it stumbled into mate.

With TB for Black, White seems to throw away the win rather quickly (at this short time control):

Another idea is to give TB White and see how long SFdev can defend for. Black is supposed to be able to delay capture of the knight for over 40 moves.

[Ovyron]

It certainly looks like SFdev at that TC was playing fairly randomly

That's unassisted engines for you!

[zullil]

It certainly looks like SFdev at that TC was playing fairly randomly

That's unassisted engines for you!

Of course, as a centaur, you would do a lot better. You'd enable tablebases and "win".

[Ovyron]

Or pick a middle game position that was won without allowing the opponent to get to an endgame

With tons of ways to skin a cat, why would someone pick the one that needs precise moves from the winning side is beyond me. You go to a position where everything wins like Zenmastur's position, in that one having TBs might hurt!

[mmt]

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.

Ok, makes sense. I guess I saw too many playouts in this game by now.