Page 1 of 1

SPCC: Testrun of LC Zero Net 529 finished

Posted: Tue Aug 07, 2018 3:40 pm
by pohl4711
Long thinking-time testrun of LC Zero v0.16.0 (Cuda) Net 529 is finished.

http://www.sp-cc.de


(Perhaps you have to clear your browsercache or reload the website)

Re: SPCC: Testrun of LC Zero Net 529 finished

Posted: Tue Aug 07, 2018 6:11 pm
by Milos
pohl4711 wrote: Tue Aug 07, 2018 3:40 pm Long thinking-time testrun of LC Zero v0.16.0 (Cuda) Net 529 is finished.

http://www.sp-cc.de


(Perhaps you have to clear your browsercache or reload the website)
You'll get around 70Elo more with test server net even though you selected the one deeply in regression (the strongest atm is 10445) ;).

Re: SPCC: Testrun of LC Zero Net 529 finished

Posted: Tue Aug 07, 2018 7:49 pm
by pohl4711
Dont just look on the selfplay graph on the testserver-site.
Here the net 10487 plays very strong vs. other engines: https://m.twitch.tv/potato_chess

Re: SPCC: Testrun of LC Zero Net 529 finished

Posted: Wed Aug 08, 2018 6:41 am
by pohl4711
pohl4711 wrote: Tue Aug 07, 2018 7:49 pm Dont just look on the selfplay graph on the testserver-site.
Here the net 10487 plays very strong vs. other engines: https://m.twitch.tv/potato_chess
Update: I restarted the testrun with Net 10480 instead of 10487, because Net 10480 will play in Division 3 of TCEC (Just found on discord (dev-log channel))

Re: SPCC: Testrun of LC Zero Net 529 finished

Posted: Wed Aug 08, 2018 4:11 pm
by Milos
pohl4711 wrote: Wed Aug 08, 2018 6:41 am
pohl4711 wrote: Tue Aug 07, 2018 7:49 pm Dont just look on the selfplay graph on the testserver-site.
Here the net 10487 plays very strong vs. other engines: https://m.twitch.tv/potato_chess
Update: I restarted the testrun with Net 10480 instead of 10487, because Net 10480 will play in Division 3 of TCEC (Just found on discord (dev-log channel))
In my tests 10480 and 10445 are close, but 10445 is still noticeably better (over 98% LoS).