From time to time, I see positions where a rook gets trapped in a corner by a pawn and a bishop, for example bRa1 + wPa2 + wBb1. I wondered about a rook getting trapped in other squares of the board with creative patterns. Then, I composed a position that SF 18 solves with a little effort, given the material unbalance:
[d]qk6/1p1ppp1B/pP1p4/P3p2P/5p2/5P2/1NPr4/6K1 w - - 0 1
[pgn][Event ""]
[Site ""]
[Round ""]
[Date "2026.08.27"]
[SetUp "1"]
[FEN "qk6/1p1ppp1B/pP1p4/P3p2P/5p2/5P2/1NPr4/6K1 w - - 0 1"]
[Result "1-0"]
1.Bd3 e4 (1...Kc8 2.h6 d5 (2...Kd8 3.h7 e6 4.h8=Q+ Ke7 5.Qxa8 1-0) 3.h7 d6 4.h8=Q+ Kd7 5.Qxa8 1-0) 2.h6 (2.Bxe4? (2.fxe4?? f3 3.h6 Rg2+ 4.Kh1 Rg4 5.h7 Rh4+ 6.Kg1 Rxh7 0-1) 2...Re2 3.h6 Re1+ 4.Kg2 Re2+ 1/2-1/2) 2...exd3 3.h7 ({(3.c4 also wins)}) 3...Rxc2 4.h8=Q+ Rc8 5.Qg7+ ({(Many moves win. It should be a matter of technique)}) 1-0[/pgn]
The sidelines might be more interesting than the mainline. The case is that the rook gets trapped on d2, with other surrounding squares forbidden to it: c2 and e2 by the bishop; d1 by the knight; d3 by both knight and pawn; and f2, g2 and h2 by the king. The white passed pawn can advance fearlessly because the black king and the black queen are also trapped.
------------
The position admits variations, mainly toying with the group of central black pawns. For example, with bPf6 instead of bPf7 (qk6/1p1pp2B/pP1p1p2/P3p2P/5p2/5P2/1NPr4/6K1 w - - 0 1), if the black king marches Kb8 → Kc8 → Kd8 → Ke8, trying to escape through f7, then Bg6+, Kf8/Kd8; h8=Q#/h8=R#. In the same position, if the black king marches Kb8 → Kc8 → Kd8 and moves e6, trying to escape through e7 as in one of the sidelines before, then h8=Q+, Ke7; Qg7+, Ke8; Bg6+, Kd8; Kf8#. So, avoid the tempting Qxa8 in that case and just go for the checkmate with Qg7+.
There is a jam of black pawns through the middle files... why? To make the position work and try to fool the engine, but there is more. Just imagine the original position without bPe5 and bPf4. Please look at the nice trick that saves black's day:
[pgn][Event ""]
[Site ""]
[Round ""]
[Date "2026.08.27"]
[SetUp "1"]
[FEN "qk6/1p1ppp1B/pP1p4/P6P/8/5P2/1NPr4/6K1 w - - 0 1"]
[Result "0-1"]
1.Bd3 Kc8 ({(Also 1...d5 with the same idea)}) 2.h6 Qb8 ({(Also 2...d5 with the same idea)}) 3.h7 Rh2! 4.Kxh2 d5+ 5.Kg2 Qe5 0-1[/pgn]
This trick works because the b8-h2 diagonal is free (would be too slow to free in the original position) and also there is not bPf6 blocking the e5-h8 diagonal that avoids a safe white promotion on h8.
As you surely have guessed, the doubled pawns on d-file are to make an escape through d7 too slow because there must be two pawn moves instead of one to free d7.
------------
Last but not least, I tried SF 18 with single core and Multi-PV disabled, to be deterministic (though I left Syzygy EGTB on) in a variety of positions. I found one that took depth 46 to be solved under these conditions!
Code: Select all
FEN: qk6/1p1pp2B/pP1p4/P6P/5p2/5P2/1NPr4/6K1 w - - 0 1
Stockfish-18-windows-x86-64-sse41-popcnt:
Found 145 WDL and 145 DTZ tablebase files (up to 5-man).
Available processors: 0-3
Using 1 thread
NNUE evaluation using nn-c288c895ea92.nnue (125MiB, (102384, 1024, 15, 32, 1))
NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))
Network replica 1: Shared memory.
[...]
40/13 00:01 426.887 569.182 0,00 Bh7e4 Rd2e2 h5h6 Re2e1+ Kg1g2 Re1e2+
41/13 00:01 490.109 553.795 0,00 Bh7e4 Rd2e2 h5h6 Re2e1+ Kg1g2 Re1e2+
42/13 00:01 523.583 554.056 0,00 Bh7e4 Rd2e2 h5h6 Re2e1+ Kg1g2 Re1e2+ Kg2h3 Re2e1
43/13 00:01 594.120 559.962 0,00 Bh7e4 Rd2e2 h5h6 Re2e1+ Kg1g2 Re1e2+
44/13 00:01 727.000 542.132 0,00 Bh7e4 Rd2e2 h5h6 Re2e1+ Kg1g2 Re1e2+ Kg2h3 Re2e1 Kh3h2 Re1e2+
45/13 00:01 753.504 546.017 0,00 Bh7e4 Rd2e2 h5h6 Re2e1+ Kg1g2 Re1e2+ Kg2h3 Re2e1
46/37 00:04 2.232.085 599.861 +6,39 Bh7d3 Qa8a7 h5h6 Kb8a8 b6xa7 Ka8xa7 h6h7 e7e5 h7h8Q e5e4 Bd3xe4 d6d5 Be4d3 d5d4 Qh8e5 b7b6 a5xb6+ Ka7b7Ajedrecista.