New open-source engine, Bit-Genie

[Gabor Szots]

So you mean more specific to Arena? But this is already known, sorry.
(uci filter and changing some graphic display options / animations might help, but I never user Arena for good reasons)

No, it is specific to certain engines under Arena. Most engines never (or rarely) forfeit on time under Arena. Yes, there is a problem with Arena but not as serious as to abandon this GUI when it is so comfortable and flexible to use for tournaments.

I use the CuteChess GUI only for very short matches, just to obtain a view of some new engine. Until interrupting/resuming tournaments (in a user-friendly way) is not implemented, it is of no use for me.

[Gabor Szots]

As Guenther says; enable the UCI-filter

I have always used the UCI filter.

[Guenther]

So you mean more specific to Arena? But this is already known, sorry.
(uci filter and changing some graphic display options / animations might help, but I never user Arena for good reasons)

No, it is specific to certain engines under Arena. Most engines never (or rarely) forfeit on time under Arena. Yes, there is a problem with Arena but not as serious as to abandon this GUI when it is so comfortable and flexible to use for tournaments.

As long as it won't happen under other (reliable) GUIs for those 'specific' engines I will still claim it's more specific to Arena.

I am quite sure this are engines, which try to squeeze time close to the edge and Arenas known overhead is guilty.
The ones that don't lose on time, often already have implemented 'hacks' and saftety margins for this.
(the forum is full of hacks for Arena BTW)

[lithander]

Hi Aryan!

I also started chess programming only a few months ago and can relate to your addiction symptoms. But it's a very pandemic-friendly hobby, at least.

I gave your engine a try as it's pretty much exactly the same strength as the current build of my engine MinimalChess that Marcel mentioned. But your's is much faster. It reports around 8 Million NPS where my engine only manages to search 1.4 NPS! Well done!

I have encountered a few stability problems with Bit Genie, though.

My tournament against Bit Genie (bmi2 build) ended prematurely in CuteChess CLI:

Code: Select all

./cutechess-cli.exe -engine conf="MinimalChess 0.3.1 PeSTO" -engine conf="Bit-Genie_100" -each tc=3+0.5 book=varied.bin -pgnout mmc031_pesto_vs_bit-genie_100.pgn -rounds 100 -games 2 -repeat -concurrency 4
[...]
Terminating process of engine Bit-Genie_100(6)
Finished game 37 (MinimalChess 0.3.1 PeSTO vs Bit-Genie_100): 1-0 {Black disconnects}

[pgn]
[Event "?"]
[Site "?"]
[Date "2021.03.22"]
[Round "19"]
[White "MinimalChess 0.3.1 PeSTO"]
[Black "Bit-Genie_100"]
[Result "1-0"]
[ECO "A10"]
[GameDuration "00:00:43"]
[GameEndTime "2021-03-22T13:00:07.779 Mitteleuropäische Zeit"]
[GameStartTime "2021-03-22T12:59:23.814 Mitteleuropäische Zeit"]
[Opening "English Opening"]
[PlyCount "103"]
[Termination "abandoned"]
[TimeControl "3+0.5"]

1. c4 {book} b6 {book} 2. e4 {book} Bb7 {book} 3. Nc3 {book} e6 {book}
4. Nf3 {+0.57/5 0.14s} Bb4 {book} 5. Bd3 {book} Ne7 {book} 6. Ne2 {book}
Na6 {+0.18/7 0.57s} 7. a3 {+0.25/6 0.35s} Bd6 {+0.10/7 0.56s}
8. b4 {+0.51/5 0.59s} f5 {-0.47/7 0.56s} 9. e5 {+2.00/6 0.27s}
Bxf3 {-0.76/7 0.56s} 10. exd6 {+1.62/6 0.15s} Bxg2 {-1.02/7 0.56s}
11. dxe7 {+1.64/6 0.18s} Qxe7 {-0.62/7 0.56s} 12. Rg1 {+1.79/6 0.25s}
Qf6 {-1.23/7 0.55s} 13. Bb2 {+2.29/6 0.34s} Qxb2 {-1.76/7 0.55s}
14. Rxg2 {+2.23/6 0.53s} Qf6 {-1.73/7 0.55s} 15. c5 {+2.40/5 0.62s}
Nb8 {-1.40/7 0.55s} 16. Rc1 {+2.20/5 0.24s} Nc6 {-1.39/7 0.55s}
17. b5 {+2.01/4 0.62s} Ne5 {-0.94/7 0.55s} 18. Rc3 {+0.89/5 0.13s}
Nxd3+ {-0.06/7 0.54s} 19. Rxd3 {+0.95/6 0.13s} bxc5 {-0.30/7 0.54s}
20. Rdg3 {+0.97/5 0.64s} O-O {+0.16/7 0.54s} 21. Rxg7+ {+1.06/5 0.63s}
Qxg7 {-0.16/9 0.54s} 22. Rxg7+ {+1.55/7 0.30s} Kxg7 {-0.26/8 0.54s}
23. Qb3 {+1.78/6 0.33s} Kf7 {-0.52/8 0.54s} 24. Qc4 {+2.02/6 0.39s}
d6 {-0.39/8 0.54s} 25. Nf4 {+2.07/6 0.34s} Rae8 {-0.11/8 0.54s}
26. d4 {+2.13/6 0.31s} Rg8 {+0.04/8 0.53s} 27. Kf1 {+2.16/6 0.30s}
Rg4 {-0.04/8 0.53s} 28. dxc5 {+2.49/6 0.23s} Re7 {-0.44/8 0.53s}
29. c6 {+3.00/6 0.18s} Ke8 {+0.34/8 0.53s} 30. h3 {+3.10/6 0.14s}
d5 {-0.75/8 0.53s} 31. Qc3 {+3.75/6 0.28s} Rxf4 {-1.72/9 0.53s}
32. Qg3 {+4.57/7 0.48s} Rxf2+ {-2.01/9 0.53s} 33. Qxf2 {+4.81/7 0.41s}
e5 {-2.16/9 0.53s} 34. Qa2 {+4.95/7 0.47s} Kf7 {-2.31/9 0.53s}
35. Qxd5+ {+5.06/7 0.30s} Kf6 {-2.86/9 0.53s} 36. Qc5 {+5.24/7 0.64s}
e4 {-3.18/9 0.52s} 37. Qxa7 {+5.23/7 0.35s} e3 {-3.31/9 0.52s}
38. Ke2 {+5.37/7 0.16s} f4 {-3.91/9 0.52s} 39. Qd4+ {+5.64/7 0.23s}
Kf5 {-4.46/8 0.52s} 40. b6 {+6.20/7 0.64s} cxb6 {-6.81/9 0.52s}
41. c7 {+7.12/7 0.44s} Rxc7 {-7.18/9 0.52s} 42. Qd5+ {+7.81/7 0.20s}
Kf6 {-8.04/9 0.52s} 43. Qd6+ {+8.14/7 0.22s} Kf5 {-10.37/10 0.52s}
44. Qxc7 {+8.98/7 0.20s} Ke4 {-10.37/8 0.52s} 45. Qc4+ {+9.63/7 0.42s}
Ke5 {-11.15/9 0.52s} 46. Kf3 {+10.06/7 0.44s} Kf5 {-11.45/9 0.52s}
47. Qxf4+ {+10.57/7 0.63s} Kg6 {-12.20/9 0.52s} 48. Qxe3 {+10.66/6 0.73s}
b5 {-12.20/8 0.52s} 49. Qe5 {+10.87/6 0.72s} Kf7 {-12.49/9 0.52s}
50. Qxb5 {+11.00/6 0.72s} Kf6 {-12.66/8 0.52s} 51. a4 {+11.24/6 0.71s}
Kf7 {-20.49/8 0.51s} 52. a5 {+12.10/6 0.71s, Black disconnects} 1-0
[/pgn]

Maybe you can try to recreate the final position using the PGN and see if there's a reproducable crash?

Up to game 37 I saw no "time forfeits" despite the fast "3+0.5" time controls and our engines seemed to be of equal strength as predicted by Marcel.

However if I try to run a tournament in the CuteChess GUI then Bit-Genie_100 loses almost *all* games on time with "40/5+1" time controls!

MinimalChess 0.3 with PeSTO falls short by just 70 Elo, but it doesn't have a hash table yet. Adding this should add another 170 Elo. Thus... a basic engine with MVV-LVA, TT, TTMove ordering, and a tapered and tuned evaluation, could be around 1950 Elo.

I don't think Hash tables would add 170 ELO to MinimalChess. I don't have TT move ordering (obviously) and no killer and history heuristics but I do play the best move from the principal variation first which provides a lot of the cutoffs you'd otherwise get from TT tables. Also my search depth is constrained by the slow speed of the naive move generator. Both factors together cause the expected gain of adding TT tables to be much lower than what you saw with Rustic.

[Guenther]

Up to game 37 I saw no "time forfeits" despite the fast "3+0.5" time controls and our engines seemed to be of equal strength as predicted by Marcel.

However if I try to run a tournament in the CuteChess GUI then Bit-Genie_100 loses almost *all* games on time with "40/5+1" time controls!

May be it doesn't understand that 'mixed' tc currently? Time management was added just recently at all.

[lithander]

May be it doesn't understand that 'mixed' tc currently? Time management was added just recently at all.

I just looked at the PGN log and curiously Bit Genie always runs out of time in the 36. move as white and 37. move as black!

[aryan1508]

Hi Aryan!

I also started chess programming only a few months ago and can relate to your addiction symptoms. But it's a very pandemic-friendly hobby, at least.

I gave your engine a try as it's pretty much exactly the same strength as the current build of my engine MinimalChess that Marcel mentioned. But your's is much faster. It reports around 8 Million NPS where my engine only manages to search 1.4 NPS! Well done!

I have encountered a few stability problems with Bit Genie, though.

My tournament against Bit Genie (bmi2 build) ended prematurely in CuteChess CLI:

Code: Select all

./cutechess-cli.exe -engine conf="MinimalChess 0.3.1 PeSTO" -engine conf="Bit-Genie_100" -each tc=3+0.5 book=varied.bin -pgnout mmc031_pesto_vs_bit-genie_100.pgn -rounds 100 -games 2 -repeat -concurrency 4
[...]
Terminating process of engine Bit-Genie_100(6)
Finished game 37 (MinimalChess 0.3.1 PeSTO vs Bit-Genie_100): 1-0 {Black disconnects}

[pgn]
[Event "?"]
[Site "?"]
[Date "2021.03.22"]
[Round "19"]
[White "MinimalChess 0.3.1 PeSTO"]
[Black "Bit-Genie_100"]
[Result "1-0"]
[ECO "A10"]
[GameDuration "00:00:43"]
[GameEndTime "2021-03-22T13:00:07.779 Mitteleuropäische Zeit"]
[GameStartTime "2021-03-22T12:59:23.814 Mitteleuropäische Zeit"]
[Opening "English Opening"]
[PlyCount "103"]
[Termination "abandoned"]
[TimeControl "3+0.5"]

1. c4 {book} b6 {book} 2. e4 {book} Bb7 {book} 3. Nc3 {book} e6 {book}
4. Nf3 {+0.57/5 0.14s} Bb4 {book} 5. Bd3 {book} Ne7 {book} 6. Ne2 {book}
Na6 {+0.18/7 0.57s} 7. a3 {+0.25/6 0.35s} Bd6 {+0.10/7 0.56s}
8. b4 {+0.51/5 0.59s} f5 {-0.47/7 0.56s} 9. e5 {+2.00/6 0.27s}
Bxf3 {-0.76/7 0.56s} 10. exd6 {+1.62/6 0.15s} Bxg2 {-1.02/7 0.56s}
11. dxe7 {+1.64/6 0.18s} Qxe7 {-0.62/7 0.56s} 12. Rg1 {+1.79/6 0.25s}
Qf6 {-1.23/7 0.55s} 13. Bb2 {+2.29/6 0.34s} Qxb2 {-1.76/7 0.55s}
14. Rxg2 {+2.23/6 0.53s} Qf6 {-1.73/7 0.55s} 15. c5 {+2.40/5 0.62s}
Nb8 {-1.40/7 0.55s} 16. Rc1 {+2.20/5 0.24s} Nc6 {-1.39/7 0.55s}
17. b5 {+2.01/4 0.62s} Ne5 {-0.94/7 0.55s} 18. Rc3 {+0.89/5 0.13s}
Nxd3+ {-0.06/7 0.54s} 19. Rxd3 {+0.95/6 0.13s} bxc5 {-0.30/7 0.54s}
20. Rdg3 {+0.97/5 0.64s} O-O {+0.16/7 0.54s} 21. Rxg7+ {+1.06/5 0.63s}
Qxg7 {-0.16/9 0.54s} 22. Rxg7+ {+1.55/7 0.30s} Kxg7 {-0.26/8 0.54s}
23. Qb3 {+1.78/6 0.33s} Kf7 {-0.52/8 0.54s} 24. Qc4 {+2.02/6 0.39s}
d6 {-0.39/8 0.54s} 25. Nf4 {+2.07/6 0.34s} Rae8 {-0.11/8 0.54s}
26. d4 {+2.13/6 0.31s} Rg8 {+0.04/8 0.53s} 27. Kf1 {+2.16/6 0.30s}
Rg4 {-0.04/8 0.53s} 28. dxc5 {+2.49/6 0.23s} Re7 {-0.44/8 0.53s}
29. c6 {+3.00/6 0.18s} Ke8 {+0.34/8 0.53s} 30. h3 {+3.10/6 0.14s}
d5 {-0.75/8 0.53s} 31. Qc3 {+3.75/6 0.28s} Rxf4 {-1.72/9 0.53s}
32. Qg3 {+4.57/7 0.48s} Rxf2+ {-2.01/9 0.53s} 33. Qxf2 {+4.81/7 0.41s}
e5 {-2.16/9 0.53s} 34. Qa2 {+4.95/7 0.47s} Kf7 {-2.31/9 0.53s}
35. Qxd5+ {+5.06/7 0.30s} Kf6 {-2.86/9 0.53s} 36. Qc5 {+5.24/7 0.64s}
e4 {-3.18/9 0.52s} 37. Qxa7 {+5.23/7 0.35s} e3 {-3.31/9 0.52s}
38. Ke2 {+5.37/7 0.16s} f4 {-3.91/9 0.52s} 39. Qd4+ {+5.64/7 0.23s}
Kf5 {-4.46/8 0.52s} 40. b6 {+6.20/7 0.64s} cxb6 {-6.81/9 0.52s}
41. c7 {+7.12/7 0.44s} Rxc7 {-7.18/9 0.52s} 42. Qd5+ {+7.81/7 0.20s}
Kf6 {-8.04/9 0.52s} 43. Qd6+ {+8.14/7 0.22s} Kf5 {-10.37/10 0.52s}
44. Qxc7 {+8.98/7 0.20s} Ke4 {-10.37/8 0.52s} 45. Qc4+ {+9.63/7 0.42s}
Ke5 {-11.15/9 0.52s} 46. Kf3 {+10.06/7 0.44s} Kf5 {-11.45/9 0.52s}
47. Qxf4+ {+10.57/7 0.63s} Kg6 {-12.20/9 0.52s} 48. Qxe3 {+10.66/6 0.73s}
b5 {-12.20/8 0.52s} 49. Qe5 {+10.87/6 0.72s} Kf7 {-12.49/9 0.52s}
50. Qxb5 {+11.00/6 0.72s} Kf6 {-12.66/8 0.52s} 51. a4 {+11.24/6 0.71s}
Kf7 {-20.49/8 0.51s} 52. a5 {+12.10/6 0.71s, Black disconnects} 1-0
[/pgn]

Maybe you can try to recreate the final position using the PGN and see if there's a reproducable crash?

Up to game 37 I saw no "time forfeits" despite the fast "3+0.5" time controls and our engines seemed to be of equal strength as predicted by Marcel.

However if I try to run a tournament in the CuteChess GUI then Bit-Genie_100 loses almost *all* games on time with "40/5+1" time controls!

MinimalChess 0.3 with PeSTO falls short by just 70 Elo, but it doesn't have a hash table yet. Adding this should add another 170 Elo. Thus... a basic engine with MVV-LVA, TT, TTMove ordering, and a tapered and tuned evaluation, could be around 1950 Elo.

I don't think Hash tables would add 170 ELO to MinimalChess. I don't have TT move ordering (obviously) and no killer and history heuristics but I do play the best move from the principal variation first which provides a lot of the cutoffs you'd otherwise get from TT tables. Also my search depth is constrained by the slow speed of the naive move generator. Both factors together cause the expected gain of adding TT tables to be much lower than what you saw with Rustic.

Hi! Can you please tell me which branch you have tested?

[aryan1508]

I've implemented it here https://github.com/Aryan1508/Bit-Genie/ ... rch_thread. What happens is A thread is used for the UCI loop since reading the stdin in C++ is 'blocking'. The other 'worker' thread demands cpu since it is used for searching

OK, I'm going to use that for my testing.

Sorry for my late reply, can you please use the master branch for testing? I just implemented the search thread branch this morning and haven't tested it myself for bugs.

[lithander]

Hi! Can you please tell me which branch you have tested?

I downloaded the "Bit-Gnie_100bmi2.exe" from your "Bit-Genie 1" release.

[aryan1508]

Hi! Can you please tell me which branch you have tested?

I downloaded the "Bit-Gnie_100bmi2.exe" from your "Bit-Genie 1" release.

Can you also paste in your `go` command? I haven't seen any crash in 1000+ games so im really confused