New PolyGlot X

[Fabien Letouzey]

Hi all,

Dann should have an experimental PolyGlot ready tonight.

UCI parsing and engine init have been completely rewritten. They are now more correct than before: engine is fully initialised before going on, only options declared by the engine are sent by PolyGlot (if the value has changed).

Hopefully this will allow more engines to work (all UCI-compliant ones). Unfortunately I might have added a new bug or some engines might prefer the "old way".

For now I am seeking data. Try in priority engines that did not work with PolyGlot 1.1

I expect problems, if any, to show up very early: at init or during the first few moves.

In case something is wrong, please send me log files in zipped form, thanks.

Obviously, don't use this version in tournaments before we have tested it seriously.

Thanks,

Fabien Letouzey (fabien_letouzey@hotmail.com).

[Fabien Letouzey]

Me again,

A detail: if the option "Engine Name" is omitted (you can use ";" to comment a line out), the UCI name picked up at init time should appear.

Fabien.

[Peter Fendrich]

Thank you Fabien for PolyGlot!
Thanks to PolygGot I only develop my new engine with the UCI protocol and there has never been any problems with winboard comunication in my tests.
My only wish is a better resign option. An option with score > x for y moves would be great!

Regards
Peter

[Fabien Letouzey]

As a matter of fact, now is a good time for feature requests.

Very small features, like the resign-option improvement, might be included in the release planned for Chrismas.

More-ambitious proposals (like opening-book handling) are welcome as well and will be considered for future versions.

Note that the announced prototype version has no such addition but tries to do things in a better way.

Fabien.

[Fabien Letouzey]

Hi again,

Dann sent me a binary. I suppose he also put it somewhere on his FTP server ... As of now I have no feedback whatsoever, maybe it does not work.

Today I am sending an updated version with multi-move resign.

The corresponding new option is "ResignMoves", the number of consecutive moves with bad score before PolyGlot resigns for the engine. The default value is 3 so it is not necessary to change INI files unless you want to experiment with the option.

For engine authors: positions with a single legal move are ignored, so the score does not matter in this case. If the engine plays one "losing" move, then single legal move then two more "losing" moves, PolyGlot will resign.

Of course a move with non-losing score resets the counter (except single legal move).

The default value of 3 should make the Glaurung bug (apparently overwriting of globally-stored score) almost invisible. I have not tested in many games ...

Fabien.

[Peter Fendrich]

Dann sent me a binary. I suppose he also put it somewhere on his FTP server ... As of now I have no feedback whatsoever, maybe it does not work.

Today I am sending an updated version with multi-move resign.

Thanks Fabien!
Why not also send a copy to Leo so he can put it on his Download pages.
There is Version 1.1 already.
/Peter

[Peter Fendrich]

Dann sent me a binary. I suppose he also put it somewhere on his FTP server ... As of now I have no feedback whatsoever, maybe it does not work.

I tried to find it (you meant Dann Corbit's ftp site I suppose) but couldn't.
Maybe I'm an idiot but it isn't easy to find things there.

/Peter

[Joachim Rang]

you don't have to look, just click here:

http://iwanuschka.de/schutthalde/polyglot.exe.bz2

regards Joachim