Fizbo version 1.3 is released

[ymatioun]

Fizbo version 1.3 has just been released. It can be downloaded from
https://sites.google.com/site/fizbochessengine/.

This release increases playing strength by around 80 ELO.

Please use it for the 50th amateur series (division 2).

Here are release notes:
Changes in version 1.3, released November 11, 2014:
1. improved transposition table management
fixed a bug
reduce size of TT entry to 8 bytes
started using TT in Q search
added prefetch
better aging
2. added aspiration window
3. various small modifications to search and evaluation.

Single CPU improvement over version 1.2 is around +80 ELO (based on self-play at 1 sec/move).
Estimated strength with 1 CPU is 2,920 ELO. Strength with 4 CPUs should be around 3,010.


Thank you, Youri Matiounine.

[Graham Banks]

Thanks Youri.

[Graham Banks]

Youri - did you change the version number inside the exe? Apparently it still opens up saying Fizbo 1.2.

[ymatioun]

Sorry, i forgot about version number that the engine reports.

I fixed it now - if you download the engine, it will now say that it is version 1.3.

Youri.

[Graham Banks]

Thanks.

[Graham Banks]

Question from Werner (CEGT):

Hi Graham,

can you please ask Youri which version we shall test, as the 1.3 compiles (old one showing 1.2 and new one) have different evaluations.

best wishes
Werner

[ymatioun]

Please use the latest version, the one that displays version number 1.3. The one posted on the website right now.

It has different evaluation, and should play slightly better.

Youri.

[Graham Banks]

Okay. Thanks again.

[Graham Banks]

Werner (CEGT) reports that he is having crashes.
I have also had a couple of crashes.


[Event "Fizbo3"]

[Site "AGANDO_HP_PC"]

[Date "2014.11.18"]

[Round "27"]

[White "Fizbo 1.3 x64"]

[Black "Protector 1.40 w32 1CPU"]

[WhiteElo "2200"]

[BlackElo "2200"]

[ECO "D43"]

[Result "0-1"]



1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 e6 5. Bg5 h6 6. Bh4

dxc4 7. e4 g5 8. Bg3 b5 9. Be2 Bb4 {-0.08/14 12} 10. Qc2

{+0.36/18 13} g4 {0.00/16 11} 11. Nd2 {+0.14/19 20} Qxd4

{+0.15/14 7} 12. O-O-O {+0.20/19 19} Qb6 {+0.19/14 8}

13. h3 {+0.02/18 7} gxh3 {+0.03/14 14} 14. Rxh3 {-0.06/18

7} Nbd7 {+0.05/14 7} 15. e5 {-0.09/18 10} Nd5 {-0.12/15 15}

16. Nde4 {-0.18/17 11} Qa5 {-0.12/14 18} 17. Kb1 {0.00/17

24} Be7 {-0.31/13 16} 18. Nxd5 {+0.88/17 37} cxd5 {-0.39/15

8} 19. Nd6+ {+0.79/19 10} Bxd6 {-0.43/16 10} 20. exd6

{+0.58/20 41} b4 {-0.43/16 12} 21. Qd2 {+0.48/19 37} c3

{-0.39/16 18} 22. Qd4 {+0.83/18 10} c2+ {Weiss

überschreitet die Zeit} 0-1



and here is the Shredder ucilog

Wed Nov 19 11:22:26 2014: to Fizbo 1.3 x64 (0): ucinewgame

Wed Nov 19 11:22:26 2014: to Fizbo 1.3 x64 (0): isready

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): readyok

Wed Nov 19 11:22:26 2014: to Fizbo 1.3 x64 (0): position startpos moves d2d4 d7d5 c2c4 c7c6 g1f3 g8f6 b1c3 e7e6 c1g5 h7h6 g5h4 d5c4 e2e4 g7g5 h4g3 b7b5 f1e2 f8b4 d1c2 g5g4 f3d2 d8d4 e1c1 d4b6 h2h3 g4h3 h1h3 b8d7 e4e5 f6d5 d2e4 b6a5 c1b1 b4e7 c3d5 c6d5 e4d6 e7d6 e5d6 b5b4 c2d2 c4c3 d2d4 c3c2

Wed Nov 19 11:22:26 2014: to Fizbo 1.3 x64 (0): go movestogo 18 wtime 240000 btime 240000

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 1 score cp 1

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 3 score cp 221 time 0 nodes 3574 nps 3574000

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 4 score cp 256 time 10 nodes 30393 nps 3039300

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 5 score cp 278 time 10 nodes 51120 nps 5112000

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 6 score cp 243 time 10 nodes 59481 nps 5948100

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 7 score cp 232 time 10 nodes 66201 nps 6620100

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 8 score cp 205 time 40 nodes 113846 nps 2846150

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 9 score cp 188 time 40 nodes 145110 nps 3627750 pv b1c2 a5c5

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 10 score cp 208 time 90 nodes 234818 nps 2609088 pv b1c2 a5c5

Wed Nov 19 11:22:26 2014: from Fizbo 1.3 x64 (0): info depth 11 score cp 187 time 140 nodes 370951 nps 2649650 pv b1c2 a5c5

Wed Nov 19 11:22:27 2014: from Fizbo 1.3 x64 (0): info depth 12 score cp 161 time 320 nodes 833123 nps 2603509 pv b1c2 a5c5

Wed Nov 19 11:22:27 2014: from Fizbo 1.3 x64 (0): info depth 13 score cp 142 time 570 nodes 1493495 nps 2620166 pv b1c2 a5c5

Wed Nov 19 11:22:27 2014: from Fizbo 1.3 x64 (0): info depth 14 score cp 145 time 870 nodes 2242718 nps 2577836 pv b1c2 a5c5

Wed Nov 19 11:22:28 2014: from Fizbo 1.3 x64 (0): info depth 15 score cp 129 time 1660 nodes 4291815 nps 2585430 pv b1c2 a5c5

Wed Nov 19 11:23:10 2014: to Fizbo 1.3 x64 (0): stop

Wed Nov 19 11:23:13 2014: to Fizbo 1.3 x64 (0): stop

Wed Nov 19 11:23:17 2014: to Fizbo 1.3 x64 (0): stop

[ymatioun]

I looked into this, and here is what i found:

1. i reproduce the problem; starting with FEN of "r1b1k2r/p2n1p2/3Pp2p/q2p4/1p1Q4/6BR/PPp1BPP1/1K1R4 w kq - 0 0" crashed the engine at search depth 15.
2. unfortunately, i did not save the source code that vent into the release. Since then i made a few changes to evaluation function and coefficients.
3. current version does not crash in that position! And it also never crashed in all the games i played on my computer recently.
4. so, i could not fix this bug in version 1.3, but current code seems to have this corrected somehow. So what i did is i released current code as version 1.31 - it is posted at https://sites.google.com/site/fizbochessengine/. And i made sure to retain the source code, in case bugs come up in the future.

So, version 1.31 does not have that particular bug. And while it does have slightly different evaluation, in my self-play it shows essentially the same strength as version 1.3. So i suggest that you use version 1.31 to avoid these crashes.

Also, v 1.31 still states that it is "version 1.3", and not 1.31; i just noticed this but i don't think it makes sense to recompile and reload the executables.

Thanks,
Youri.

[Graham Banks]

Thanks Youri.

[Graham Banks]

Hi Graham,

sometimes I have crashes always in the same position like here:

Regards

Werner

[Event "Fizbo2"]

[Site "I7"]

[Date "2014.11.21"]

[Round "38"]

[White "MinkoChess x64"]

[Black "Fizbo13x64"]

[Result "*"]

[BlackElo "2200"]

[ECO "B94"]

[Opening "Sizilianisch (Najdorf-Variante mit 6. Lg5) 6.Bg5 Nbd7 7.f4"]

[Time "16:40:20"]

[WhiteElo "2200"]

[TimeControl "40/480:40/480:40/480"]

[Termination "unterminated"]

[PlyCount "39"]

[WhiteType "program"]

[BlackType "program"]



1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 Nbd7 7. f4 e5 8.

Nf5 Qb6 9. Qd2 Qxb2 {0.00/18 12} 10. Rb1 {-0.29/18 20} Qa3 {-0.14/18 10}

11. Ne3 {-0.19/18 19} h6 {+0.13/17 11} 12. Bxf6 {0.00/20 19} Nxf6 {+0.01/18

7} 13. Ned5 {+0.02/21 18} Qc5 {+0.16/17 6} 14. Nxf6+ {+0.02/18 18} gxf6

{+0.19/17 7} 15. Bd3 {+0.06/17 17} exf4 {+0.38/18 13} 16. Na4 {+0.60/21 17}

Qe3+ {+0.01/21 21} 17. Qxe3 {+0.67/22 17} fxe3 {-0.16/23 14} 18. Nb6

{+0.71/22 17} Rb8 {-0.43/22 9} 19. Nxc8 {+0.67/23 16} Rxc8 {-0.11/21 15}

20. Rxb7 {+0.68/24 16} *

[Graham Banks]

Hi Graham,

Arena repeated this game, and Fizbo crashes on the same move:

When I let Fizbo alone calculate on this position it moves h6 h5

best wishes

Werner


[Event "Fizbo2"]

[Site "I7"]

[Date "2014.11.21"]

[Round "38"]

[White "MinkoChess x64"]

[Black "Fizbo13x64"]

[Result "*"]

[BlackElo "2200"]

[ECO "B94"]

[Opening "Sizilianisch (Najdorf-Variante mit 6. Lg5) 6.Bg5 Nbd7 7.f4"]

[Time "16:52:29"]

[WhiteElo "2200"]

[TimeControl "40/480:40/480:40/480"]

[Termination "unterminated"]

[PlyCount "39"]

[WhiteType "program"]

[BlackType "program"]



1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 Nbd7 7. f4 e5 8.

Nf5 Qb6 9. Qd2 Qxb2 {0.00/18 12} 10. Rb1 {-0.29/18 20} Qa3 {-0.14/18 10}

11. Ne3 {-0.29/18 19} h6 {+0.13/17 11} 12. Bxf6 {+0.09/19 19} Nxf6

{+0.01/18 7} 13. Ned5 {0.00/20 19} Qc5 {+0.16/17 6} 14. Nxf6+ {-0.01/17 18}

gxf6 {+0.19/17 7} 15. Bd3 {+0.22/19 18} exf4 {+0.38/18 13} 16. Na4

{+0.65/22 18} Qe3+ {+0.01/21 20} 17. Qxe3 {+0.72/23 17} fxe3 {-0.16/23 14}

18. Nb6 {+0.65/24 17} Rb8 {-0.43/22 9} 19. Nxc8 {+0.73/24 17} Rxc8

{-0.11/21 15} 20. Rxb7 {+0.79/23 16} *

[ymatioun]

OK, i finally found the bug and fixed it. Now version 1.31 posted at https://sites.google.com/site/fizbochessengine/ should not crash. Aside from bug fix, it is unchanged from what it was, except that now it reports version number 1.31. And it is still different from version 1.3, but playing strength is about the same.

Thanks for giving me details of the crash - without them i would have never found this bug!

Youri.



PS the bug was that when i checked if castling move is allowed, i did not check that opponent's king will not be attacking my king after castling. This situation never came up in my testing before, because (i think) this happens extremely rarely. But this is exactly the problem what crashes the engine in the position that Werner identified.