Deep puzzles: what the analysis missed

Positions from real Lichess games where the server analysis (1,000,000 nodes a move) approved the move played and a 20,000,000-node search calls it a mistake or blunder against a unique best move. Companion to the write-up on brianhliou.com.

32 chapters

1. Oleksandr Bortnyk: 20...b5? (blitz, 2026-07-11)8 moves

   Oleksandr Bortnyk to move, Black. Lichess's analysis gave the move played no mark (-0.63 before, -0.41 after). A 20-million-node search calls it a blunder, 21 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: -0.63 to -0.41, no label. At 20M nodes: blunder, 19.7 win% given up, 21 points worse than Bxe4.

2. Oleksandr Bortnyk: 54...Qf6? (blitz, 2026-08-28)8 moves

   Oleksandr Bortnyk to move, Black. Lichess's analysis gave the move played no mark (-1.58 before, -1.03 after). A 20-million-node search calls it a blunder, 20 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: -1.58 to -1.03, no label. At 20M nodes: blunder, 17.5 win% given up, 20 points worse than Qa7+.

3. Oleksandr Bortnyk: 58...Qa8? (blitz, 2026-08-28)8 moves

   Oleksandr Bortnyk to move, Black. Lichess's analysis gave the move played no mark (-1.87 before, -1.28 after). A 20-million-node search calls it a blunder, 20 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: -1.87 to -1.28, no label. At 20M nodes: blunder, 20.4 win% given up, 20 points worse than Qa7+.

4. Sergei Zhigalko: 20...Qd7? (blitz, 2026-07-12)8 moves

   Sergei Zhigalko to move, Black. Lichess's analysis gave the move played no mark (-1.38 before, -1.30 after). A 20-million-node search calls it a mistake, 14 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: -1.38 to -1.30, no label. At 20M nodes: mistake, 14.6 win% given up, 14 points worse than Rxd2.

5. Anish Giri: 15.Bb5? (blitz, 2026-05-09)3 moves

   Anish Giri to move, White. Lichess's analysis gave the move played no mark (+0.45 before, +0.19 after). A 20-million-node search calls it a mistake, 12 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: +0.45 to +0.19, no label. At 20M nodes: mistake, 11.4 win% given up, 12 points worse than Bxf6.

6. Dmitry Andreikin: 37...exf5? (blitz, 2026-09-10)2 moves

   Dmitry Andreikin to move, Black. Lichess's analysis gave the move played no mark (-0.50 before, +0.00 after). A 20-million-node search calls it a mistake, 12 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: -0.50 to +0.00, no label. At 20M nodes: inaccuracy, 5.6 win% given up, 12 points worse than Nxe5.

7. Sergei Zhigalko: 34...Qa4? (blitz, 2026-07-12)3 moves

   Sergei Zhigalko to move, Black. Lichess's analysis gave the move played no mark (-1.30 before, -0.98 after). A 20-million-node search calls it a mistake, 12 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: -1.30 to -0.98, no label. At 20M nodes: blunder, 15.6 win% given up, 12 points worse than Nh3+.

8. Magnus Carlsen: 35...Ngh7? (blitz, 2022-02-28)8 moves

   Magnus Carlsen to move, Black. Lichess's analysis gave the move played no mark (-0.47 before, -0.07 after). A 20-million-node search calls it a mistake, 11 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: -0.47 to -0.07, no label. At 20M nodes: inaccuracy, 9.7 win% given up, 11 points worse than Qxh4.

9. Sergei Zhigalko: 52.Rh6+? (blitz, 2026-07-14)8 moves

   Sergei Zhigalko to move, White. Lichess's analysis gave the move played no mark (+1.71 before, +1.19 after). A 20-million-node search calls it a mistake, 10 win-chance points against a best move that is unique. Find it before opening the sideline.

   The game move. Lichess: +1.71 to +1.19, no label. At 20M nodes: inaccuracy, 7.2 win% given up, 10 points worse than Rxb7.

10. Sergei Zhigalko: 33...e4? (blitz, 2026-07-14)8 moves

    Sergei Zhigalko to move, Black. Lichess's analysis gave the move played no mark (+0.84 before, +1.04 after). A 20-million-node search calls it a blunder, 23 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +0.84 to +1.04, no label. At 20M nodes: blunder, 22.9 win% given up, 23 points worse than Kd7.

11. Jerry (ChessNetwork): 28.c4? (blitz, 2026-08-19)6 moves

    Jerry (ChessNetwork) to move, White. Lichess's analysis gave the move played no mark (+0.00 before, +0.00 after). A 20-million-node search calls it a blunder, 23 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +0.00 to +0.00, no label. At 20M nodes: inaccuracy, 8.4 win% given up, 23 points worse than h4.

12. Ediz Gürel: 31.Nd4? (blitz, 2026-02-04)2 moves

    Ediz Gürel to move, White. Lichess's analysis gave the move played no mark (+1.42 before, +1.06 after). A 20-million-node search calls it a blunder, 22 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +1.42 to +1.06, no label. At 20M nodes: blunder, 22.6 win% given up, 22 points worse than d6.

13. Ediz Gürel: 55.Qe8+? (blitz, 2026-07-04)8 moves

    Ediz Gürel to move, White. Lichess's analysis gave the move played no mark (+1.35 before, +1.14 after). A 20-million-node search calls it a blunder, 21 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +1.35 to +1.14, no label. At 20M nodes: blunder, 17.0 win% given up, 21 points worse than Ne5.

14. Ediz Gürel: 57.Na3? (blitz, 2026-07-04)8 moves

    Ediz Gürel to move, White. Lichess's analysis gave the move played no mark (+1.12 before, +1.14 after). A 20-million-node search calls it a blunder, 21 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +1.12 to +1.14, no label. At 20M nodes: blunder, 20.1 win% given up, 21 points worse than Ne5.

15. Dmitry Andreikin: 47...Ka4? (blitz, 2026-08-28)8 moves

    Dmitry Andreikin to move, Black. Lichess's analysis gave the move played no mark (-0.79 before, -0.75 after). A 20-million-node search calls it a blunder, 18 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.79 to -0.75, no label. At 20M nodes: mistake, 11.6 win% given up, 18 points worse than c5.

16. Dmitry Andreikin: 69...Kg3? (blitz, 2026-08-28)8 moves

    Dmitry Andreikin to move, Black. Lichess's analysis gave the move played no mark (+1.32 before, +1.67 after). A 20-million-node search calls it a blunder, 17 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +1.32 to +1.67, no label. At 20M nodes: blunder, 16.0 win% given up, 17 points worse than g5.

17. Magnus Carlsen: 15...Ne5? (blitz, 2023-07-01)2 moves

    Magnus Carlsen to move, Black. Lichess's analysis gave the move played no mark (-0.94 before, -0.43 after). A 20-million-node search calls it a blunder, 16 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.94 to -0.43, no label. At 20M nodes: mistake, 13.1 win% given up, 16 points worse than f5.

18. Alireza Firouzja: 57.Bc4? (blitz, 2024-05-18)8 moves

    Alireza Firouzja to move, White. Lichess's analysis gave the move played no mark (-0.78 before, -0.85 after). A 20-million-node search calls it a blunder, 16 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.78 to -0.85, no label. At 20M nodes: mistake, 13.9 win% given up, 16 points worse than Be4.

19. Dmitry Andreikin: 47...Qf1+? (blitz, 2026-08-28)2 moves

    Dmitry Andreikin to move, Black. Lichess's analysis gave the move played no mark (-0.71 before, -0.72 after). A 20-million-node search calls it a mistake, 15 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.71 to -0.72, no label. At 20M nodes: mistake, 11.8 win% given up, 15 points worse than Qg3.

20. Alireza Firouzja: 54.Be4? (blitz, 2024-05-18)8 moves

    Alireza Firouzja to move, White. Lichess's analysis gave the move played no mark (-1.30 before, -1.20 after). A 20-million-node search calls it a mistake, 13 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -1.30 to -1.20, no label. At 20M nodes: mistake, 10.0 win% given up, 13 points worse than Bd3.

21. Alireza Firouzja: 28...Qe6? (blitz, 2024-05-18)8 moves

    Alireza Firouzja to move, Black. Lichess's analysis gave the move played no mark (-0.87 before, -0.68 after). A 20-million-node search calls it a mistake, 13 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.87 to -0.68, no label. At 20M nodes: inaccuracy, 9.9 win% given up, 13 points worse than Qf7.

22. Oleksandr Bortnyk: 31...Nd5? (blitz, 2026-08-28)8 moves

    Oleksandr Bortnyk to move, Black. Lichess's analysis gave the move played no mark (-1.05 before, -0.85 after). A 20-million-node search calls it a mistake, 13 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -1.05 to -0.85, no label. At 20M nodes: blunder, 17.7 win% given up, 13 points worse than Qb6.

23. Sergei Zhigalko: 12...h6? (blitz, 2026-08-15)8 moves

    Sergei Zhigalko to move, Black. Lichess's analysis gave the move played no mark (-1.54 before, -1.07 after). A 20-million-node search calls it a mistake, 13 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -1.54 to -1.07, no label. At 20M nodes: mistake, 13.0 win% given up, 13 points worse than Nd4.

24. Ediz Gürel: 41...Rd5? (blitz, 2026-03-14)8 moves

    Ediz Gürel to move, Black. Lichess's analysis gave the move played no mark (-0.67 before, -0.25 after). A 20-million-node search calls it a mistake, 13 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.67 to -0.25, no label. At 20M nodes: mistake, 10.9 win% given up, 13 points worse than Rd4.

25. Alireza Firouzja: 37.Bd4? (blitz, 2024-05-18)8 moves

    Alireza Firouzja to move, White. Lichess's analysis gave the move played no mark (-2.20 before, -2.23 after). A 20-million-node search calls it a mistake, 12 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -2.20 to -2.23, no label. At 20M nodes: inaccuracy, 8.5 win% given up, 12 points worse than Bb4.

26. Dmitry Andreikin: 31...Kf7? (blitz, 2026-08-24)8 moves

    Dmitry Andreikin to move, Black. Lichess's analysis gave the move played no mark (+1.29 before, +1.07 after). A 20-million-node search calls it a mistake, 12 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +1.29 to +1.07, no label. At 20M nodes: mistake, 13.9 win% given up, 12 points worse than Rc8.

27. Magnus Carlsen: 35...Rh1+? (blitz, 2023-07-01)8 moves

    Magnus Carlsen to move, Black. Lichess's analysis gave the move played no mark (+0.42 before, +0.88 after). A 20-million-node search calls it a mistake, 11 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +0.42 to +0.88, no label. At 20M nodes: mistake, 11.5 win% given up, 11 points worse than Rh2.

28. Sergei Zhigalko: 19.Bxc6? (blitz, 2026-08-15)8 moves

    Sergei Zhigalko to move, White. Lichess's analysis gave the move played no mark (+1.45 before, +1.62 after). A 20-million-node search calls it a mistake, 11 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +1.45 to +1.62, no label. At 20M nodes: mistake, 14.9 win% given up, 11 points worse than Bf4.

29. Sergei Zhigalko: 23.g3? (blitz, 2026-07-11)8 moves

    Sergei Zhigalko to move, White. Lichess's analysis gave the move played no mark (-1.26 before, -1.69 after). A 20-million-node search calls it a mistake, 11 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -1.26 to -1.69, no label. At 20M nodes: inaccuracy, 7.9 win% given up, 11 points worse than Be2.

30. Oleksandr Bortnyk: 14.Be4? (blitz, 2026-08-28)8 moves

    Oleksandr Bortnyk to move, White. Lichess's analysis gave the move played no mark (+2.27 before, +1.95 after). A 20-million-node search calls it a mistake, 11 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: +2.27 to +1.95, no label. At 20M nodes: inaccuracy, 5.7 win% given up, 11 points worse than Ng5.

31. Sergei Zhigalko: 36...Bb1? (blitz, 2026-08-15)8 moves

    Sergei Zhigalko to move, Black. Lichess's analysis gave the move played no mark (-0.95 before, -0.67 after). A 20-million-node search calls it a mistake, 11 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.95 to -0.67, no label. At 20M nodes: inaccuracy, 6.7 win% given up, 11 points worse than Bc2.

32. Sergei Zhigalko: 15...Bg6? (blitz, 2026-07-14)8 moves

    Sergei Zhigalko to move, Black. Lichess's analysis gave the move played no mark (-0.89 before, -0.62 after). A 20-million-node search calls it a mistake, 10 win-chance points against a best move that is unique. Find it before opening the sideline.

    The game move. Lichess: -0.89 to -0.62, no label. At 20M nodes: mistake, 10.3 win% given up, 10 points worse than Nc2.