New engine: Tcheran

[jgilchrist]

v2.2 of Tcheran was released today with a couple of bug fixes for issues which caused the en passant target to be forgotten during search. This was pointed out during the 107th Amateur competition being run at the moment, which pushed me to release early.

There's some other refactoring included in the release, but nothing that gains strength. Elo gains are almost entirely a result of the bug fixes and are:

Code: Select all

Elo   | 23.11 +- 9.18
Conf  | 8.0+0.08s Threads=1 Hash=16MB
Games | N: 4096 W: 1655 L: 1383 D: 1058

Code: Select all

Elo   | 24.81 +- 17.94
Conf  | 60.0+0.60s Threads=1 Hash=128MB
Games | N: 1024 W: 397 L: 324 D: 303

https://github.com/jgilchrist/tcheran/releases/tag/v2.2
https://github.com/jgilchrist/tcheran/b ... ELOG.md#22

[Andrew Matthews]

What does strategy do, does it make sense to change it and how to do it?

Is there a way to disable logging? That file has already grown to 25 MB.

Hey Gabor, one trick I've used in the past to handle similar situations where I wasn't lucky enough to have the author make a change

Here is the trick: before running I just create a blank file with the log name (zappa.log for example), then modify permissions to make it read only, since the file name exists, the engine can't write to the file, but that permission shouldn't affect engine playing.

Let me know if that works!

[jgilchrist]

Hey Gabor, one trick I've used in the past to handle similar situations where I wasn't lucky enough to have the author make a change

Here is the trick: before running I just create a blank file with the log name (zappa.log for example), then modify permissions to make it read only, since the file name exists, the engine can't write to the file, but that permission shouldn't affect engine playing.

Let me know if that works!

Nice trick Andrew - for Tcheran specifically it shouldn't be required since logging was disabled by default (except for fatal crashes) back in v1.1 (https://github.com/jgilchrist/tcheran/b ... ELOG.md#11).

May well be useful for other engines though!

[Andrew Matthews]

Agreed 100%, it was just unfortunate that I saw that comment in your engines thread, the solution is indeed obsolete and not required for Tcheran, but just thought a good place to leave a solution id found many moons a go, as it should hold true for other places where we weren't lucky enough to have an option from authors to disable the log.

I just also want to say thank you for taking these quality of life feature requests from us and incorporating them in your creation!

[Ray]

Here is the trick: before running I just create a blank file with the log name (zappa.log for example), then modify permissions to make it read only, since the file name exists, the engine can't write to the file, but that permission shouldn't affect engine playing.

Unless of course the engine crashes because it can't write to the file

[Andrew Matthews]

Here is the trick: before running I just create a blank file with the log name (zappa.log for example), then modify permissions to make it read only, since the file name exists, the engine can't write to the file, but that permission shouldn't affect engine playing.

Unless of course the engine crashes because it can't write to the file

yeah, possible for sure, thankfully where I'd tried it worked just fine, good call out Ray!

[Gabor Szots]

What does strategy do, does it make sense to change it and how to do it?

Is there a way to disable logging? That file has already grown to 25 MB.

Hey Gabor, one trick I've used in the past to handle similar situations where I wasn't lucky enough to have the author make a change

Here is the trick: before running I just create a blank file with the log name (zappa.log for example), then modify permissions to make it read only, since the file name exists, the engine can't write to the file, but that permission shouldn't affect engine playing.

Let me know if that works!

Thanks Andrew, good idea. Not needed here any more but useful for other engines e.g. Beaches.

[jgilchrist]

v2.3 of Tcheran was released today. Elo gains are mainly due to a new lazy move generation implementation, but also a couple of minor bug fixes.

Code: Select all

Elo   | 25.75 +- 9.23
Conf  | 8.0+0.08s Threads=1 Hash=16MB
Games | N: 4096 W: 1685 L: 1382 D: 1029
Penta | [207, 358, 735, 421, 327]

Code: Select all

Elo   | 22.08 +- 8.85
Conf  | 60.0+0.60s Threads=1 Hash=128MB
Games | N: 4096 W: 1542 L: 1282 D: 1272
Penta | [145, 384, 837, 430, 252]

https://github.com/jgilchrist/tcheran/releases/tag/v2.3
https://github.com/jgilchrist/tcheran/b ... ELOG.md#23

[Graham Banks]

Thanks. I'll put it into Division 11 shortly and replay its games.

[jgilchrist]

v2.4 of Tcheran was released today. Various improvements are included, but the highest Elo gains are from fixes and optimisations of move scoring and picking. For the next version I'll be focusing on evaluation improvements.

Code: Select all

Elo   | 28.48 +- 8.10 (95%)
Conf  | 8.0+0.08s Threads=1 Hash=16MB
Games | N: 4096 W: 1686 L: 1351 D: 1059
Penta | [151, 311, 900, 424, 262]

Code: Select all

Elo   | 29.81 +- 14.57 (95%)
Conf  | 60.0+0.60s Threads=1 Hash=128MB
Games | N: 1028 W: 384 L: 296 D: 348
Penta | [17, 99, 227, 121, 50]

https://github.com/jgilchrist/tcheran/releases/tag/v2.4
https://github.com/jgilchrist/tcheran/b ... ELOG.md#24

[Gabor Szots]

I have put it in my testing queue.

[jgilchrist]

Thanks for testing 2.4 Gabor!

v2.5 was released today:

Code: Select all

Elo   | 56.13 +- 8.57 (95%)
Conf  | 8.0+0.08s Threads=1 Hash=16MB
Games | N: 4108 W: 1828 L: 1170 D: 1110
Penta | [131, 284, 793, 488, 358]

Code: Select all

Elo   | 74.40 +- 17.43 (95%)
Conf  | 60.0+0.60s Threads=1 Hash=128MB
Games | N: 1024 W: 468 L: 252 D: 304
Penta | [23, 76, 174, 140, 99]

https://github.com/jgilchrist/tcheran/releases/tag/v2.5
https://github.com/jgilchrist/tcheran/b ... ELOG.md#25