Can find right move but didn't

[tissatussa]

I just compiled your v3.1.0 dev and let it play a game in CuteChess GUI against another strong engine. It made a draw but could have won the game .. it failed to find the right move in this position :

After the game i investigated the concerning position in SCID and it turns out Belette CAN find the right move (Qh1) immediately :

Why did this happen ? I encounter such 'behaviour' with other engines too .. is it a bug in the Hash management ? My gut feeling says so ..

---

[Event "engine vs engine"]
[Site "Holland @ https://lichess.org/jsws8TaD "]
[Date "2024.09.08"]
[Round "?"]
[White "Belette v3.1.0 dev"]
[Black "4KU v5.1"]
[Result "1/2-1/2"]
[ECO "D20"]
[GameDuration "00:12:44"]
[Opening "Queen's gambit accepted"]
[PlyCount "84"]
[TimeControl "300+3"]

1. d4 {+0.21/31 14s} d5 {-0.17/23 10s} 2. c4 {+0.18/33 13s} dxc4 {-0.40/25 12s}
3. e3 {+0.25/32 15s} c5 {-0.26/25 12s} 4. Bxc4 {+0.35/34 12s} Nf6 {-0.46/24 14s}
5. Nf3 {+0.26/33 13s} Qc7 {-0.52/24 22s} 6. d5 {+0.37/29 12s} a6 {-0.47/26 16s}
7. a4 {+0.43/34 13s} Bf5 {-0.69/25 15s} 8. O-O {+0.52/32 11s}
Nbd7 {-0.57/25 11s} 9. Nc3 {+0.54/34 13s} O-O-O {-0.66/25 27s}
10. Qe2 {+0.52/36 10s} Bg4 {-0.62/29 9.2s} 11. h3 {+0.63/37 16s}
Bxf3 {-0.65/29 21s} 12. Qxf3 {+0.65/34 13s} Ne5 {-0.65/26 6.0s}
13. Qe2 {+0.59/36 11s} g5 {-0.68/26 6.6s} 14. Ba2 {+0.86/31 8.8s}
g4 {-0.89/24 6.5s} 15. hxg4 {+0.99/34 8.7s} Nexg4 {-0.69/27 16s}
16. f4 {+1.34/32 8.6s} Rg8 {-0.55/26 5.8s} 17. e4 {+0.98/32 22s}
Nh5 {-0.84/24 7.7s} 18. Rf3 {+1.27/32 8.8s} Qb6 {-0.82/24 6.0s}
19. a5 {+1.07/34 13s} Qh6 {-0.79/23 4.5s} 20. g3 {+0.95/32 7.0s}
Qg6 {-0.99/22 5.6s} 21. Qg2 {+1.29/31 7.1s} Nhf6 {-0.95/24 17s}
22. e5 {+1.50/31 7.0s} Nd7 {-1.14/24 5.0s} 23. Bd2 {+1.33/30 12s}
h5 {-1.08/22 3.9s} 24. Rc1 {+2.08/30 9.0s} h4 {-0.71/22 3.8s}
25. e6 {+2.23/32 6.5s} fxe6 {-1.93/25 12s} 26. dxe6 {+2.09/35 6.7s}
Nb8 {-2.18/26 26s} 27. Na4 {+2.13/31 6.2s} Bg7 {-2.36/26 12s}
28. f5 {+2.10/32 7.3s} Bd4+ {-1.61/24 8.8s} 29. Kf1 {+2.39/32 5.6s}
Qh5 {-2.05/24 9.5s} 30. Nb6+ {+2.32/33 6.4s} Kc7 {-2.41/21 2.3s}
31. Nd5+ {+2.30/35 9.4s} Rxd5 {-2.35/21 3.3s} 32. Bxd5 {+2.80/32 6.6s}
Kd8 {-3.19/24 5.4s} 33. Bxb7 {+3.40/31 6.0s} h3 {-4.12/24 3.0s}
34. Qe2 {+3.74/32 5.4s} h2 {0.00/29 2.0s} 35. Kg2 {-0.01/40 5.1s}
h1=Q+ {0.00/39 2.0s} 36. Rxh1 {-0.01/52 5.2s} Qxh1+ {0.00/47 2.1s}
37. Kxh1 {+0.01/54 5.6s} Rh8+ {0.00/49 18s} 38. Kg2 {+0.01/65 4.9s}
Rh2+ {0.00/60 1.5s} 39. Kf1 {-0.01/72 4.5s} Rh1+ {0.00/64 2.2s}
40. Kg2 {+0.01/70 5.1s} Rh2+ {0.00/69 2.1s} 41. Kf1 {-0.01/72 4.3s}
Rh1+ {0.00/69 2.4s} 42. Kg2 {+0.01/76 4.5s}
Rh2+ {0.00/75 2.7s, Draw by 3-fold repetition} 1/2-1/2

[ i'm on Xubuntu 22.04 ]

[vincentbab]

This could indeed be due to a hash collision in the transposition table. It would really help if you can provide the full input and ouput of the engine for the game to reproduce the bug. You can use the "Debug Log File" option for this purpose

[tissatussa]

..the full input and ouput of the engine for the game to reproduce the bug..

i can do that but this game wasn't logged. you never experienced such 'behaviour' yourself ? I use 128 Mb hash