Odonata author contact

[akanalytics]

Hi. I'm the author!

Thanks everyone for taking an interest. My email address is andy at webcalculator dot com. And no problems with it being posted here.
I'll add it to github to make it easier for others to find.

These forums are great btw - I'm learning so much - and realise I have so much more to learn...

The binary versions of odonata on github are very old - I'll try and get some more recent versions uploaded soon, and will post when I've done so.

Thanks again,
Andy.

[Gabor Szots]

Thank you Andy. We are watching!

[akanalytics]

Hi I've posted a new release of Odonata v0.3.12.

Details along with a windows binary at
https://github.com/akanalytics/odonata/releases

Unsure of the ELO now - but suspect its much improved, though probably 1600-1700 (guess?)

Looking to produce Mac and Linux bins in future releases.

Andy

[Gabor Szots]

Hi I've posted a new release of Odonata v0.3.12.

Details along with a windows binary at
https://github.com/akanalytics/odonata/releases

Unsure of the ELO now - but suspect its much improved, though probably 1600-1700 (guess?)

Looking to produce Mac and Linux bins in future releases.

Andy

Thanks Andy. I notice there are a lots of UCI parameters. Does it mean you are not sure the default values are best?

[akanalytics]

Hi Gabor. I think I'm comfortable with the defaults for now - though really many of them are guesses
The idea is to have uci setoptions for many evaluation parameters, and extend the uci protocol with some commands for loading pgn files and evaluating positions in bulk - tweaking the eval params to get Odonata's evals to best line up with the each pgn's outcome - won/drawn/lost. This would be used for tuning eval parameters. A step I hopefully perform before next release. I thought performing this through uci would make it easier to script in python.... But we'll see.

I've not been able to find any "standard naming" for uci protocol extensions - certainly stockfish offers some ('d' for displaying board etc) but nothing around tuning.

Andy

[akanalytics]

Odonata 0.4.0 is available

Link: https://github.com/akanalytics/odonata/ ... tag/v0.4.0
ELO improved to around 2300 maybe.

Still struggling to ensure the "modern" version actually runs on Haswell CPUs - but hopefully this release fixes this issue. Let me know!

Started using Rémi Coulom's CLOP for tuning non-eval parameters - with some success - though the process is very slow.

Andy