Leela crushes any resistance in this position

[Nordlandia]

Tried a couple of games in 30m+30s format with contempt at 0 and 36 (+50% in contempt) both ended in draw. Strangely SF does not treat the bishop as a problem.

[MMarco]

This position reminds me some Capablanca games.

Yes!! Winter-Capablanca 1919: https://www.chessgames.com/perl/chessgame?gid=1055995

[Ovyron]

I'm pretty sure this is a draw.

If you were sure you could produce a defense against a single line thrown at you, just one. Since you're not sure you can produce this line, you can't be sure this is a draw.

[jp]

Tried a couple of games in 30m+30s format with contempt at 0 and 36 (+50% in contempt) both ended in draw. Strangely SF does not treat the bishop as a problem.

What do you mean by "does not treat the bishop as a problem"?

I'm pretty sure this is a draw.

If you were sure you could produce a defense against a single line thrown at you, just one. Since you're not sure you can produce this line, you can't be sure this is a draw.

Go on then. Throw the single line at me. (The drawn positions I gave as examples in this thread were from "single lines".) "Pretty sure" does not mean "sure". Of course, we cannot be 100% certain it's a draw (or a loss).

[MikeGL]

On my quick analysis with engine at the background, this seems to be a forced win as
long as the position is not oversimplified. Just exchange one rook and maintain 1 R + N + B. The game seems to be a forced win for black.

[pgn]
[Event "i7-5960X 4.5GHz / 2x 2080 Supe"]
[Site "i7-5960X 4.5GHz / 2x 2080 Supe"]
[Date "2020.07.03"]
[Round "?"]
[White "stockfish_20070107_x64_bmi2"]
[Black "lc0 v0.26.0"]
[Result "*"]
[TimeControl "1800+30"]
[SetUp "1"]
[FEN "r4rk1/pp6/1bn1R2p/6pb/5p2/1NP2N1P/PP3PPB/3R2K1 b - - 0 6"]
[Termination "unterminated"]
[PlyCount "35"]
[WhiteType "human"]
[BlackType "human"]

6. ... Rad8 {+0.51/19 31} 7. Rde1 {-0.35/38 91s} Bxf3 {+0.88/14 48} 8. gxf3
{-0.27/40 48} Kg7 {+0.93/14 31} 9. a4 {-0.40/47 165s} Rf6 {+1.25/13 73s}
10. h4 {-0.50/46 66s} Rd5 {+1.31/14 63s} (10. .. Rxe6 11. Rxe6 Rd1+ 12. Kg2
Rb1 13. hxg5 hxg5 14. Nd2 Rxb2 15. Ne4 Bd8 16. Nd6) 11. Kg2 {-0.97/41 212s}
Rxe6 {+1.49/15 70s} 12. Rxe6 {-0.26/41 68s} Bc7 {+1.52/14 40} 13. Nd4
{-0.17/55 33} Re5 {+1.89/16 77s} 14. Rxe5 {-0.39/56 178s} Bxe5 {+1.98/16
41} 15. Nxc6 {-0.39/71 33} bxc6 {+2.41/14 91s} 16. h5 {-0.39/76 73s} Kf6
{+3.62/12 181s} 17. Kf1 {-0.39/81 65s} c5 {+3.62/12 0.21s} 18. Ke2
{-0.39/88 57} c4 {+3.63/11 52} 19. Kf1 {-0.39/89 59} Ke6 {+3.86/10 148s}
20. Kg2 {-0.39/91 46} Bd6 {+3.86/10 75s} 21. Kh3 {-0.39/90 47} Bc5
{+3.76/11 71s} 22. Bg1 {-0.39/87 54} Kd6 {+3.58/12 77s} 23. Kg4 {-0.39/89
357s} Ke5 {+3.58/13 adjudication 52s, Draw by adjudication: user decision}
*

[/pgn]
/
/
Below position arised after 13.hxg5 line above.
Where black will have active R and a slight advantage. But needs deeper analysis.

[d]8/pp4k1/1bn1R3/6p1/P4p2/1NP2P2/1P3PKB/1r6 w - - 0 14

[pgn]
[Event "i7-5960X 4.5GHz / 2x 2080 Supe"]
[Site "i7-5960X 4.5GHz / 2x 2080 Supe"]
[Date "2020.07.03"]
[Round "?"]
[White "stockfish_20070107_x64_bmi2"]
[Black "lc0 v0.26.0"]
[Result "*"]
[TimeControl "1800+30"]
[SetUp "1"]
[FEN "r4rk1/pp6/1bn1R2p/6pb/5p2/1NP2N1P/PP3PPB/3R2K1 b - - 0 6"]
[Termination "unterminated"]
[PlyCount "20"]
[WhiteType "program"]
[BlackType "human"]

6. ... Rad8 {+0.51/19 31} 7. Rde1 {-0.35/38 91s} Bxf3 {+0.88/14 48} 8. gxf3
{-0.27/40 48} Kg7 {+0.93/14 31} 9. a4 {-0.40/47 165s} Rf6 {+1.25/13 73s}
10. h4 {-0.50/46 66s} Rxe6 11. Rxe6 Rd1+ 12. Kg2 Rb1 13. hxg5 hxg5 14. Nd2
Rxb2 15. Ne4 Bd8 16. Nd6 *
[/pgn]

[Ovyron]

Go on then. Throw the single line at me.

The line I'd send depends on the moves you'd make in a game. Would you be willing to play it? Because, if you're sure this is a draw, you'd just need to send a move that leads to a position that you're sure is also a draw. And then another. Do the thing you did that made you sure to decide your next move. Would you be willing to do that?

[jp]

On my quick analysis with engine at the background, this seems to be a forced win as long as the position is not oversimplified. Just exchange one rook and maintain 1 R + N + B. The game seems to be a forced win for black.

It's possible, but would need to be demonstrated.

At least avoiding exchanging both rooks might avoid all the fortress or TB draws that engine lines run into. The question is why the engines allow these exchanges.

I wonder what SF-NNUE thinks.