is this a fortress?

[zenpawn]

Some weekends ago, I played a game as black where my opponent had a chance to defend my attack by trading a queen and pawn for two rooks to reach this position:

[d]3k4/2p3pp/5p2/8/2N3P1/PP3RP1/2q3PK/8 b - - 0 34

During the game, I felt this would be a fortress since even if the rook can be pushed off the f3 square, it also has e3. SF 10 gives an eval of approximately -2 at a depth of 45. Playing out its lines, it eventually starts sacrificing the a- and g-pawns instead of, say, just shuffling the king, and black's advantage thus increases.

[MikeGL]

Some weekends ago, I played a game as black where my opponent had a chance to defend my attack by trading a queen and pawn for two rooks to reach this position:

[d]3k4/2p3pp/5p2/8/2N3P1/PP3RP1/2q3PK/8 b - - 0 34

During the game, I felt this would be a fortress since even if the rook can be pushed off the f3 square, it also has e3. SF 10 gives an eval of approximately -2 at a depth of 45. Playing out its lines, it eventually starts sacrificing the a- and g-pawns instead of, say, just shuffling the king, and black's advantage thus increases.

Maybe contempt issues of the engine. Otherwise it won't sac its pawns if it has a bit of 'respect' against black player.

[hgm]

It seems easy for white to hold a draw here. But why do you even think black is better? Probably white should play for a win here. But black is lucky that the white KIng is susceptile to perpetual check.

[Ovyron]

Yes, there's a certain popular white opening where black actually has the edge, but no engine will show the right score, because it depends on a critical position where white has an useless queen but engines don't realize it and like white better.

Figures engines greatly overestimate the value of the queen in lots of positions, unfortunately they're so rare that investigating them (play a bunch of Leela v Stockfish games, wait for an instance where Stockfish keeps the queen and Leela doesn't, and Leela wins, and see if Leela's eval is more accurate) wouldn't be as productive as elo-chasing for their purposes.

[zenpawn]

It seems easy for white to hold a draw here. But why do you even think black is better? Probably white should play for a win here. But black is lucky that the white KIng is susceptile to perpetual check.

I agree with you. During the game, I was worried my opponent (white) would defend in this way and hold the draw easily via fortress. For some reason, SF 10 thinks black is up +2, but its lines are unconvincing to say the least.

[MikeB]

Some weekends ago, I played a game as black where my opponent had a chance to defend my attack by trading a queen and pawn for two rooks to reach this position:

[d]3k4/2p3pp/5p2/8/2N3P1/PP3RP1/2q3PK/8 b - - 0 34

During the game, I felt this would be a fortress since even if the rook can be pushed off the f3 square, it also has e3. SF 10 gives an eval of approximately -2 at a depth of 45. Playing out its lines, it eventually starts sacrificing the a- and g-pawns instead of, say, just shuffling the king, and black's advantage thus increases.

maybe not , Honey-FD ( Honey with fortress detection code written by Joe Ellis)

11 Cores each with pondering on, 8192M hash , game 5 min with 5 sec inc.
Lc0 scoring - 70% is under two pawns, 100% is two queens advantage. nice queen sac at the end to secure the win
[pgn][Event "Mac Pro x5690 3.46 Ghz "] [Site "Mac-Pro.local"] [Date "2019.10.07"] [Round "-"] [White "Honey 100719"] [Black "Honey 100719"] [Result "0-1"] [TimeControl "300+5"] [FEN "3k4/2p3pp/5p2/8/2N3P1/PP3RP1/2q3PK/8 b - - 0 1"] [SetUp "1"] 1... Kc8 {+70.18/29 +280} 2. a4 {+30.34/31 +283} Qd1 {+67.86/32 +266} 3. Nb2 {+32.65/32 +278} Qd4 {+68.46/33 +271} 4. Nd3 {+32.22/31 +272} Qxg4 {+71.40/33 +276} 5. Nf2 {+32.43/28 +271} Qh5+ {+67.35/38 +235} 6. Kg1 {+32.61/34 +262} f5 {+73.60/33 +229} 7. Nd3 {+23.56/33 +219} Qh6 {+78.59/36 +216} 8. Nf4 {+24.96/33 +192} g5 {+79.57/32 +202} 9. Ne2 {+20.17/33 +190} Qd6 {+79.75/31 +206} 10. Kf1 {+23.95/31 +170} Qd5 {+81.62/33 +202} 11. Rc3 {+23.51/29 +173} Qd6 {+80.97/34 +183} 12. Ke1 {+15.90/33 +163} Kb7 {+82.04/35 +177} 13. Rc4 {+23.06/29 +156} c5 {+82.14/32 +176} 14. Kf2 {+15.40/32 +130} Kb6 {+84.53/33 +181} 15. Rc3 {+13.98/39 +101} h5 {+86.39/33 +172} 16. Re3 {+12.67/32 +90} g4 {+88.08/29 +171} 17. Ke1 {+10.35/35 +43} Qd5 {+91.85/33 +158} 18. Kf2 {+8.52/31 +36} Ka5 {+92.45/27 +162} 19. Rc3 {+9.23/29 +34} Kb4 {+92.69/29 +167} 20. Re3 {+6.95/28 +24} Qd2 {+93.48/28 +168} 21. Re5 {+9.05/25 +27} f4 {+93.75/26 +169} 22. Re4+ {+6.18/27 +11} Kxb3 {+95.48/26 +161} 23. Rxf4 {+4.88/27 +16} c4 {+96.32/25 +160} 24. Re4 {+2.97/26 +17} Qa5 {+96.95/26 +161} 25. Ke3 {+3.17/25 +7} c3 {+99.01/25 +158} 26. Nd4+ {+1.61/25 +5} Kb2 {+99.98/28 +120} 27. Kd3 {+0.01/31 +10} Qa6+ {+100.00/24 +118} 28. Nb5 {+0.00/29 +11} c2 {+100.00/32 +112} 29. Rb4+ {+0.00/32 +16} Kc1 {+100.00/32 +107} 30. Rc4 {+0.00/29 +21} Kd1 {+100.00/31 +103} 31. Nc3+ {+0.83/29 +21} Ke1 {+100.00/34 +107} 32. Na2 {+0.00/37 +15} Kf2 {+100.00/35 +112} 33. Kc3 {+0.00/41 +13} Kxg3 {+100.00/39 +113} 34. Kb3 {+0.00/23 +5} Qxc4+ {+100.00/37 +94} 35. Kxc4 {+0.00/39 +5} h4 {+100.00/39 +97} 36. Nc1 {+0.00/36 +6} h3 {+100.00/42 +92} 37. gxh3 {+0.00/39 +11} gxh3 {+1000.22/40 +89} 38. Kb3 {+0.00/33 +13} h2 {+1000.20/39 +22} 39. Kxc2 {-1000.25/17 +18} h1=Q {+1000.19/40 +27} 40. Nb3 {-1000.19/21 +23} Kf4 {+1000.18/32 +32} 41. Kc3 {-1000.18/20 +28} Ke3 {+1000.17/36 +37} 42. Kc4 {-1000.16/31 +33} Qe4+ {+1000.16/35 +19} 43. Kb5 {-1000.15/27 +38} Qb1 {+1000.14/24 +23} 44. Kb4 {-1000.13/29 +42} Kd3 {+1000.13/29 +28} 45. a5 {-1000.12/28 +47} Qb2 {+1000.12/27 +33} 46. a6 {-1000.11/26 +51} Qc3+ {+1000.11/27 +38} 47. Kb5 {-1000.10/25 +56} Qxb3+ {+1000.10/27 +42} 48. Kc6 {-1000.09/26 +61} Kd4 {+1000.09/23 +47} 49. a7 {-1000.07/26 +66} Qd5+ {+1000.07/28 +52} 50. Kc7 {-1000.06/32 +71} Kc5 {+1000.06/35 +57} 51. a8=Q {-1000.05/54 +76} Qxa8 {+1000.05/48 +62} 52. Kd7 {-1000.04/1 +81} Qe4 {+1000.04/24 +67} 53. Kc7 {-1000.03/245 +86} Qh7+ {+1000.03/245 +72} 54. Kd8 {-1000.02/245 +91} Kd6 {+1000.02/245 +77} 55. Ke8 {-1000.01/245 +96} Qe7# {+1000.01/245 +82} {Black mates} 0-1[/pgn]

Note - once mate was found , I checked "fast play" to finish the game rapidly. Thats why you see big clocks gains at the end.

[hgm]

Seems to me white abandons the fortress without any reason, and blunders away a Pawn in the process.

Not sure what this 'fortress code' does, but it doesn't seem to be much good.

[MikeB]

Seems to me white abandons the fortress without any reason, and blunders away a Pawn in the process.

Not sure what this 'fortress code' does, but it doesn't seem to be much good.

That may be true - but I don't see how the "fortress" here holds. What am I missing?

[zenpawn]

OK, I'm coming around to this not being a fortress. I played a couple games against SF 8 from this position, attempting to use f3 and/or e3 for the rook and keeping the queenside steady with a4 and b3 pawns and the knight on c4. Unfortunately, the queen was able to force some zugzwangs to win the g4 pawn and eventually put a black pawn there, taking away the f3 square for the rook. I thought e3 would work just as well with the knight covering it, but more zugzwangs followed such that the white king could no longer shuffle.

I then pitted SF10 as white against SF7 as black in a 5+1 game:

[pgn][Event "Computer Chess Game"] [Site "ROOK"] [Date "2019.10.07"] [Round "-"] [White "Stockfish 10 64"] [Black "Stockfish 7 64"] [Result "0-1"] [TimeControl "300+1"] [FEN "3k4/2p3pp/5p2/8/2N3P1/PP3RP1/2q3PK/8 b - - 0 1"] [SetUp "1"] 1... Kc8 {+1.47/26 5} 2. a4 {-1.56/28 14} Qd1 {+1.45/30 0.1} 3. g5 {-2.13/29 13} fxg5 {+1.45/32} 4. Rf8+ {-2.22/27 4} Kb7 {+1.45/27} 5. Na5+ {-2.26/29 7} Ka6 {+1.61/23 8} 6. Ra8+ {-1.96/25 4} Kb6 {+1.61/1 0.1} 7. b4 {-1.82/27 2.1} Qh5+ {+1.51/25 5} 8. Kg1 {-1.86/1} c5 {+1.48/25 2.2} 9. Rb8+ {-1.94/27 5} Kc7 {+1.47/26 1.9} 10. Rb5 {-2.67/29 16} cxb4 {+1.44/25 6} 11. Rxb4 {-2.79/28 0.1} Qd1+ {+1.44/24 5} 12. Kh2 {-2.81/28 1.1} Qd6 {+1.50/24 4} 13. Rb7+ {-2.99/31 19} Kc8 {+1.90/26} 14. Rb3 {-2.92/28 7} Qh6+ {+2.13/24 9} 15. Kg1 {-3.14/28 0.1} Qf6 {+2.08/25 2.6} 16. Nc4 {-2.97/30 12} Qd4+ {+2.31/19 7} 17. Ne3 {-3.18/31 0.1} Qxa4 {+2.36/20 2.8} 18. Rd3 {-3.18/32 4} g4 {+2.45/24 7} 19. Kf2 {-3.18/36 0.1} Qa2+ {+2.45/28 6} 20. Ke1 {-3.18/39 0.1} h5 {+2.45/29 2.6} 21. Rd2 {-3.18/34 4} Qf7 {+2.45/31 6} 22. Ke2 {-3.18/34 4} Kc7 {+2.45/32 6} 23. Ke1 {-3.18/36 7} Kc6 {+2.45/33 0.1} 24. Ke2 {-3.18/37 4} Qf6 {+2.45/28 6} 25. Rd5 {-3.77/31 21} Qb2+ {+2.45/31 0.1} 26. Ke1 {-3.77/30 6} Qc3+ {+2.67/30 38} 27. Kf2 {-3.88/33 5} g5 {+2.95/22 6} 28. Ke2 {-3.53/25 3} h4 {+2.89/23} 29. gxh4 {-3.88/31 10} gxh4 {+3.25/24 2.0} 30. Rd3 {-3.88/33 5} Qe5 {+3.37/27 14} 31. Rd5 {-4.69/33 18} Qb2+ {+3.41/29} 32. Rd2 {-5.04/31 9} Qb5+ {+3.41/30 2.1} 33. Ke1 {-5.04/32 3} Qe5 {+3.41/30 1.7} 34. Rc2+ {-5.04/32 2.8} Kb5 {+4.32/25 7} 35. Re2 {-5.05/32 4} Qd4 {+4.31/28 4} 36. g3 {-5.05/37 0.1} h3 {+4.31/29 3} 37. Nf1 {-5.05/37 0.8} Qa1+ {+4.31/31 5} 38. Kf2 {-5.05/43 6} Qa7+ {+4.31/33 0.1} 39. Ke1 {-5.05/42 6} Kc4 {+4.47/32 14} 40. Rd2 {-7.34/37 39} Kc3 {+8.55/30 40} 41. Rh2 {-7.34/25 7} Kd3 {+12.70/24 19} 42. Rd2+ {-15.69/25 7} Ke4 {+11.81/22 5} 43. Rf2 {-7.01/24 2.7} Qa5+ {+14.83/23 4} 44. Ke2 {-12.58/23 5} Kd4 {+16.29/21 0.1} 45. Rf8 {-49.10/24 21} Qa2+ {+43.81/30} 46. Ke1 {-8.58/24 5} h2 {+49.03/30 11} 47. Rd8+ {-60.23/21 2.7} Kc5 {+49.14/29 0.1} 48. Nxh2 {-61.10/26 8} Qxh2 {+53.01/32 1.9} 49. Rc8+ {-73.50/25 10} Kd4 {+56.81/34} 50. Rd8+ {-99.36/27 8} Ke3 {+61.41/32 8} 51. Re8+ {-99.54/33 0.1} Kf3 {+61.52/27 1.1} 52. Rf8+ {-99.60/30 0.5} Kxg3 {+82.98/32 7} 53. Rd8 {-99.66/34} Qh4 {+87.14/26 5} 54. Rd7 {-99.68/29 1.6} Kf3+ {+99.65/25 3} 55. Kd1 {-99.74/27 0.8} g3 {+99.71/25} 56. Rd3+ {-99.76/34 2.0} Ke4 {+99.77/32 1.1} 57. Ra3 {-99.80/36 0.1} Qe7 {+99.79/31 1.1} 58. Ra2 {-99.80/37 1.0} Qb4 {+99.81/35 2.0} 59. Kc1 {-99.82/41 0.7} Qc4+ {+99.83/36 0.1} 60. Kb1 {-99.84/48 0.8} Qxa2+ {+99.85/44 3} 61. Kxa2 {-99.86/79 0.1} g2 {+99.87/46 0.8} 62. Ka3 {-99.88/83 0.5} g1=Q {+99.89/48 3} 63. Kb4 {-99.90/100 0.1} Kd4 {+99.91/56 0.7} 64. Kb5 {-99.92/118 1.0} Qg6 {+99.93/127 0.1} 65. Ka4 {-99.94/127 0.7} Qb6 {+99.95/127 0.1} 66. Ka3 {-99.48/1 0.2} Kc3 {+99.97/127 0.1} 67. Ka2 {-99.98/127 0.4} Qb2# {+99.99/127} {Xboard adjudication: Checkmate} 0-1 [/pgn]

[MikeB]

OK, I'm coming around to this not being a fortress. I played a couple games against SF 8 from this position, attempting to use f3 and/or e3 for the rook and keeping the queenside steady with a4 and b3 pawns and the knight on c4. Unfortunately, the queen was able to force some zugzwangs to win the g4 pawn and eventually put a black pawn there, taking away the f3 square for the rook. I thought e3 would work just as well with the knight covering it, but more zugzwangs followed such that the white king could no longer shuffle.

I then pitted SF10 as white against SF7 as black in a 5+1 game:

[pgn][Event "Computer Chess Game"] [Site "ROOK"] [Date "2019.10.07"] [Round "-"] [White "Stockfish 10 64"] [Black "Stockfish 7 64"] [Result "0-1"] [TimeControl "300+1"] [FEN "3k4/2p3pp/5p2/8/2N3P1/PP3RP1/2q3PK/8 b - - 0 1"] [SetUp "1"] 1... Kc8 {+1.47/26 5} 2. a4 {-1.56/28 14} Qd1 {+1.45/30 0.1} 3. g5 {-2.13/29 13} fxg5 {+1.45/32} 4. Rf8+ {-2.22/27 4} Kb7 {+1.45/27} 5. Na5+ {-2.26/29 7} Ka6 {+1.61/23 8} 6. Ra8+ {-1.96/25 4} Kb6 {+1.61/1 0.1} 7. b4 {-1.82/27 2.1} Qh5+ {+1.51/25 5} 8. Kg1 {-1.86/1} c5 {+1.48/25 2.2} 9. Rb8+ {-1.94/27 5} Kc7 {+1.47/26 1.9} 10. Rb5 {-2.67/29 16} cxb4 {+1.44/25 6} 11. Rxb4 {-2.79/28 0.1} Qd1+ {+1.44/24 5} 12. Kh2 {-2.81/28 1.1} Qd6 {+1.50/24 4} 13. Rb7+ {-2.99/31 19} Kc8 {+1.90/26} 14. Rb3 {-2.92/28 7} Qh6+ {+2.13/24 9} 15. Kg1 {-3.14/28 0.1} Qf6 {+2.08/25 2.6} 16. Nc4 {-2.97/30 12} Qd4+ {+2.31/19 7} 17. Ne3 {-3.18/31 0.1} Qxa4 {+2.36/20 2.8} 18. Rd3 {-3.18/32 4} g4 {+2.45/24 7} 19. Kf2 {-3.18/36 0.1} Qa2+ {+2.45/28 6} 20. Ke1 {-3.18/39 0.1} h5 {+2.45/29 2.6} 21. Rd2 {-3.18/34 4} Qf7 {+2.45/31 6} 22. Ke2 {-3.18/34 4} Kc7 {+2.45/32 6} 23. Ke1 {-3.18/36 7} Kc6 {+2.45/33 0.1} 24. Ke2 {-3.18/37 4} Qf6 {+2.45/28 6} 25. Rd5 {-3.77/31 21} Qb2+ {+2.45/31 0.1} 26. Ke1 {-3.77/30 6} Qc3+ {+2.67/30 38} 27. Kf2 {-3.88/33 5} g5 {+2.95/22 6} 28. Ke2 {-3.53/25 3} h4 {+2.89/23} 29. gxh4 {-3.88/31 10} gxh4 {+3.25/24 2.0} 30. Rd3 {-3.88/33 5} Qe5 {+3.37/27 14} 31. Rd5 {-4.69/33 18} Qb2+ {+3.41/29} 32. Rd2 {-5.04/31 9} Qb5+ {+3.41/30 2.1} 33. Ke1 {-5.04/32 3} Qe5 {+3.41/30 1.7} 34. Rc2+ {-5.04/32 2.8} Kb5 {+4.32/25 7} 35. Re2 {-5.05/32 4} Qd4 {+4.31/28 4} 36. g3 {-5.05/37 0.1} h3 {+4.31/29 3} 37. Nf1 {-5.05/37 0.8} Qa1+ {+4.31/31 5} 38. Kf2 {-5.05/43 6} Qa7+ {+4.31/33 0.1} 39. Ke1 {-5.05/42 6} Kc4 {+4.47/32 14} 40. Rd2 {-7.34/37 39} Kc3 {+8.55/30 40} 41. Rh2 {-7.34/25 7} Kd3 {+12.70/24 19} 42. Rd2+ {-15.69/25 7} Ke4 {+11.81/22 5} 43. Rf2 {-7.01/24 2.7} Qa5+ {+14.83/23 4} 44. Ke2 {-12.58/23 5} Kd4 {+16.29/21 0.1} 45. Rf8 {-49.10/24 21} Qa2+ {+43.81/30} 46. Ke1 {-8.58/24 5} h2 {+49.03/30 11} 47. Rd8+ {-60.23/21 2.7} Kc5 {+49.14/29 0.1} 48. Nxh2 {-61.10/26 8} Qxh2 {+53.01/32 1.9} 49. Rc8+ {-73.50/25 10} Kd4 {+56.81/34} 50. Rd8+ {-99.36/27 8} Ke3 {+61.41/32 8} 51. Re8+ {-99.54/33 0.1} Kf3 {+61.52/27 1.1} 52. Rf8+ {-99.60/30 0.5} Kxg3 {+82.98/32 7} 53. Rd8 {-99.66/34} Qh4 {+87.14/26 5} 54. Rd7 {-99.68/29 1.6} Kf3+ {+99.65/25 3} 55. Kd1 {-99.74/27 0.8} g3 {+99.71/25} 56. Rd3+ {-99.76/34 2.0} Ke4 {+99.77/32 1.1} 57. Ra3 {-99.80/36 0.1} Qe7 {+99.79/31 1.1} 58. Ra2 {-99.80/37 1.0} Qb4 {+99.81/35 2.0} 59. Kc1 {-99.82/41 0.7} Qc4+ {+99.83/36 0.1} 60. Kb1 {-99.84/48 0.8} Qxa2+ {+99.85/44 3} 61. Kxa2 {-99.86/79 0.1} g2 {+99.87/46 0.8} 62. Ka3 {-99.88/83 0.5} g1=Q {+99.89/48 3} 63. Kb4 {-99.90/100 0.1} Kd4 {+99.91/56 0.7} 64. Kb5 {-99.92/118 1.0} Qg6 {+99.93/127 0.1} 65. Ka4 {-99.94/127 0.7} Qb6 {+99.95/127 0.1} 66. Ka3 {-99.48/1 0.2} Kc3 {+99.97/127 0.1} 67. Ka2 {-99.98/127 0.4} Qb2# {+99.99/127} {Xboard adjudication: Checkmate} 0-1 [/pgn]

I don't see the fortress, it seems to me that back can force a pawn off the board, either a or b pawn or the g4 pawn and once a white pawn comes off , it's over.