New Release of Francesca - 0.22

[silentshark]

Francesca 0.22 is released

This should be another step forward in strength, hopefully over +30 ELO

Change detail in the readme.txt file.

Available at www.silentshark.co.uk

I have some testing results which I might post later if time..

Enjoy

Regards,
Tom

[Graham Banks]

Thanks Tom.

[Dann Corbit]

Thanks, I snagged my copy

[silentshark]

Thanks, somewhere I should note that the version with the larger hash tables (uses over 1GB memory) is not suitable for many purposes. It seems at fast timescales that smaller hash is usually better. At least on my machine. Maybe cache related or something.

I'm currently running a blitz test between the version with 1GB hash and one with 64Mb hash. The latter is is about 40 ELO ahead after 200 games.. interesting..

[tpoppins]

Francesca sometimes claims a draw in positions that technically do not qualify as insufficient mating material, e.g. KBKN and gets awarded a zero for doing so.

[pgn][Event "Francesca MAD 0.22 64-bit 40/40 Gauntlet"] [Site "Dual E5-2690v2"] [Date "2019.03.23"] [Round "1"] [White "Francesca MAD 0.22 64-bit"] [Black "RuyDos 1.1.9 64-bit"] [Result "0-1"] [ECO "B02"] [GameDuration "01:14:31"] [GameEndTime "2019-03-23T21:37:57.525 Eastern Daylight Time"] [GameStartTime "2019-03-23T20:23:26.349 Eastern Daylight Time"] [Opening "Alekhine's defense"] [PlyCount "163"] [Termination "adjudication"] [TimeControl "40/1200"] [Variation "Saemisch attack"] 1. e4 {book} Nf6 {book} 2. e5 {book} Nd5 {book} 3. Nc3 {book} c6 {book} 4. Nf3 {book} d6 {book} 5. d4 {+0.73/25 89s} Nxc3 {-0.26/17 38s} 6. bxc3 {+0.85/25 28s} Nd7 {-0.17/18 15s} 7. exd6 {+1.18/26 38s} exd6 {-0.20/19 19s} 8. Bd3 {+1.23/26 47s} Be7 {-0.26/20 59s} 9. O-O {+1.26/26 35s} O-O {-0.25/20 62s} 10. Re1 {+0.49/25 36s} Re8 {-0.27/19 79s} 11. Rb1 {+0.50/25 77s} g6 {-0.20/18 54s} 12. Bh6 {+0.73/23 29s} Nb6 {-0.21/18 16s} 13. Ng5 {+0.60/26 26s} Bd7 {-0.16/20 101s} 14. Qd2 {+0.63/26 52s} f6 {-0.13/20 21s} 15. Ne6 {+0.65/31 36s} Bxe6 {-0.15/21 20s} 16. Rxe6 {+0.32/31 75s} Qd7 {-0.12/22 47s} 17. Rbe1 {+0.48/33 40s} Bf8 {-0.10/21 54s} 18. Rxe8 {+0.52/34 35s} Rxe8 {-0.10/20 112s} 19. Rxe8 {+0.51/34 70s} Qxe8 {-0.10/20 130s} 20. Bxf8 {+0.49/33 33s} Kxf8 {-0.09/18 10s} 21. c4 {+0.48/32 41s} Nd7 {-0.08/19 32s} 22. Qa5 {+0.62/24 28s} Qb8 {-0.19/19 34s} 23. Qd2 {+0.56/28 35s} Qe8 {0.00/22 8.7s} 24. Qb4 {+0.58/29 22s} Qb8 {-0.11/18 19s} 25. a3 {+0.55/28 41s} c5 {0.00/17 16s} 26. Qa4 {+0.38/30 29s} Ke7 {+0.03/18 11s} 27. c3 {+0.46/32 17s} f5 {+0.03/18 47s} 28. Qc2 {+0.44/22 24s} Qe8 {+0.06/16 5.8s} 29. Qb1 {+0.33/24 27s} b6 {+0.06/17 7.0s} 30. Qe1+ {+0.29/28 18s} Kf7 {0.00/21 5.3s} 31. Qxe8+ {+0.28/29 16s} Kxe8 {-0.06/23 31s} 32. f3 {+0.25/29 20s} Ke7 {-0.04/23 5.0s} 33. Kf1 {+0.24/28 17s} Nf6 {0.00/20 6.0s} 34. a4 {+0.24/27 14s} Nh5 {+0.01/17 7.8s} 35. g3 {+0.24/28 13s} Nf6 {-0.04/20 9.0s} 36. Kf2 {+0.24/28 12s} d5 {-0.03/22 8.9s} 37. Ke2 {+0.23/28 10s} Ke6 {-0.03/21 14s} 38. Ke3 {+0.23/27 8.7s} Kd6 {-0.01/21 14s} 39. Kd2 {+0.23/27 7.8s} dxc4 {0.00/23 15s} 40. Bxc4 {+0.23/25 1.4s} h6 {-0.02/24 34s} 41. Bb5 {+0.23/31 69s} g5 {0.00/26 43s} 42. Bd3 {+0.23/31 47s} Ke6 {0.00/26 33s} 43. Bc4+ {+0.22/29 29s} Kd6 {0.00/30 39s} 44. Ke2 {+0.22/30 34s} Nd5 {0.00/26 34s} 45. Kd3 {+0.22/32 39s} cxd4 {0.00/28 28s} 46. cxd4 {+0.22/32 24s} h5 {0.00/29 76s} 47. Kd2 {+0.23/32 31s} h4 {0.00/30 41s} 48. Ba6 {+0.23/31 26s} Kd7 {0.00/29 35s} 49. Bb7 {+0.23/32 35s} Kd6 {0.00/29 24s} 50. Kd3 {+0.23/32 31s} h3 {0.00/27 23s} 51. Bc8 {+0.23/34 37s} Ne7 {0.00/30 48s} 52. Ba6 {+0.23/35 23s} Nd5 {-0.02/30 188s} 53. Bc4 {+0.23/35 23s} a6 {-0.03/29 29s} 54. Kd2 {+0.50/32 26s} Nc7 {-0.01/28 34s} 55. Ke3 {+0.50/33 25s} Nd5+ {0.00/28 16s} 56. Kd3 {+0.50/35 26s} Kc6 {0.00/28 17s} 57. Kd2 {+0.36/35 35s} b5 {-0.01/29 25s} 58. axb5+ {+0.36/36 31s} axb5 {0.00/32 17s} 59. Bf1 {+0.36/36 30s} g4 {0.00/33 12s} 60. fxg4 {+0.36/36 36s} fxg4 {0.00/34 16s} 61. Bd3 {+0.20/36 48s} Nf6 {0.00/35 14s} 62. Ke3 {+0.20/35 69s} b4 {0.00/34 12s} 63. Bc4 {0.00/33 24s} Nh5 {0.00/30 12s} 64. Kf2 {0.00/32 40s} Nf6 {0.00/38 10s} 65. Ke3 {0.00/33 23s} Nh5 {0.00/37 12s} 66. Be6 {0.00/31 24s} Nxg3 {0.00/29 11s} 67. Bxg4 {0.00/30 25s} Nf1+ {0.00/33 12s} 68. Kd3 {0.00/31 43s} b3 {0.00/41 13s} 69. Bxh3 {0.00/29 31s} Nxh2 {0.00/41 15s} 70. Bf5 {0.00/30 25s} Kb5 {0.00/40 15s} 71. Kc3 {0.00/32 30s} Nf3 {0.00/47 12s} 72. Bd7+ {0.00/33 26s} Kb6 {0.00/63 3.7s} 73. Kxb3 {0.00/34 24s} Ka7 74. d5 {0.00/34 21s} Ne1 75. d6 {0.00/35 17s} Nf3 76. Kb4 {0.00/36 16s} Ne1 77. Kc5 {0.00/33 12s} Nd3+ 78. Kc6 {0.00/36 12s} Ne5+ 79. Kc7 {0.00/39 9.9s} Ng6 80. Bc6 {0.00/36 6.6s} Ne5 81. d7 {0.00/41 58s} Nxd7 82. Ba4 {0.00/49 0.003s, Black wins by adjudication: Invalid result claim} 0-1 [/pgn]

Theoretically it is possible to reach a mate position in such endgames

[d]6bk/8/6NK/8/8/8/8/8 b - - 0 1
so Francesca should just play on.

I would also propose raising the hash size of the mid-tier exec, fran.exe, to 512 MB or at least to double the current 160/175 MB; that could be beneficial for longer time controls, like our 40/40.
It would be also nice if it could run on chips supporting POPCNT/SSE 4.2 only. Currently it appears to be either BMI2 or no optimizations at all.

[silentshark]

Hi Tirsa,

This is a great spot. I've fixed this endgame bug, which seems to give just a few ELO.

I'll look at your other suggestions, too.

If you have any other comments, or have seen any other strange moves or evals, do let me know

Thanks again,

Tom

[tpoppins]

Sorry, Tom, after I made the previous post I ran into the following passage in CECP protocol specs (towards the end of Section 9):

Note that (in accordance with FIDE rules) only KK, KNK, KBK and KBKB with all bishops on the same color can be claimed as draws on the basis of insufficient mating material. The end-games KNNK, KBKN, KNKN and KBKB with unlike bishops do have mate positions, and cannot be claimed.

I hope your fix takes all of the above-mentioned endgames into account, not just KBKN.

[carldaman]

Sorry, Tom, after I made the previous post I ran into the following passage in CECP protocol specs (towards the end of Section 9):

Note that (in accordance with FIDE rules) only KK, KNK, KBK and KBKB with all bishops on the same color can be claimed as draws on the basis of insufficient mating material. The end-games KNNK, KBKN, KNKN and KBKB with unlike bishops do have mate positions, and cannot be claimed.

I hope your fix takes all of the above-mentioned endgames into account, not just KBKN.

I think that in the US the USCF rules are more relaxed, and all the above situations are treated as insufficient material, which is a lot more reasonable, imo, than the FIDE rules. You could argue FIDE is more of an authority on such rules and should therefore prevail, but for me it is common sense that resonates harder.

[tpoppins]

Sorry, Tom, after I made the previous post I ran into the following passage in CECP protocol specs (towards the end of Section 9):

Note that (in accordance with FIDE rules) only KK, KNK, KBK and KBKB with all bishops on the same color can be claimed as draws on the basis of insufficient mating material. The end-games KNNK, KBKN, KNKN and KBKB with unlike bishops do have mate positions, and cannot be claimed.

I hope your fix takes all of the above-mentioned endgames into account, not just KBKN.

I think that in the US the USCF rules are more relaxed, and all the above situations are treated as insufficient material, which is a lot more reasonable, imo, than the FIDE rules. You could argue FIDE is more of an authority on such rules and should therefore prevail, but for me it is common sense that resonates harder.

You are mistaken. The USCF rules are more detailed on the subject but essentially boil down to the same thing:

USCF:

14D. Insufficient material to continue.
The game is drawn when one of the following endings exists as of the most recently determined legal move, in which the possibility of a win is excluded for either side (effective 1-1-19). See also 15H, Reporting of results:
TD TIP: Remember a 14D draw claim is first a draw offer (Rule 14, The Drawn Game).
14D1. King vs. king.
King vs. king.
14D2. King vs. king with bishop or knight.
King vs. king with bishop or knight.
14D3. King and bishop vs. king and bishop.
King and bishop vs. king and bishop, with both bishops on diagonals of the same color.
14D4. No legal moves leading to checkmate by opponent.
There are no legal moves that could lead to the player being checkmated by the opponent

FIDE:

9.6 The game is drawn when a position is reached from which a checkmate cannot occur by any possible series of legal moves. This immediately ends the game, provided that the move producing this position was legal.

Would you claim to possess more common sense than both FIDE and USCF?

Anyway, "common sense" is not particularly relevant here; it's an empty phrase without a context given. I will manually adjudicate as draw any KBKN (and the like) endgame lost by a strong engine (2200+ Elo) like Francesca due to an invalid draw claim; and replay any such game where the claimant is under 2000 Elo without even looking at the final position, as there's no guarantee that a low-rated engine would be able make its claim good. Common sense used in both cases, action taken quite different.

[carldaman]

It looks like I was wrong about the USCF rules. That's how I remembered them.

Anyway, 'common sense' can be somewhat subjective and opinion-driven and may actually vary based on the entities competing. As a chess player I feel it's wrong to adjudicate certain endgames that are virtually unwinnable as 'wins on time' because an extremely unlikely and easily avoidable checkmate position exists. Doesn't necessarily mean that I'm right, of course.

Use your own common sense as you see fit.