SF progression test page

524 views
Skip to first unread message

Leonardo Ljubičić aka Dragon Mist

unread,
May 24, 2021, 7:26:56 AMMay 24
to FishCooking
Updated with single core test of May 22nd.

stephanenico...@gmail.com

unread,
May 25, 2021, 12:01:36 PMMay 25
to FishCooking
Thanks a lot, Leonardo :-)

stephanenico...@gmail.com

unread,
Jun 15, 2021, 4:54:45 PMJun 15
to FishCooking
We have made new regression/progression tests here, with good results. Probably back to the improvement rate of SF8 in the last few weeks :-)

Yery Spark

unread,
Jun 16, 2021, 1:59:37 AMJun 16
to FishCooking
Excellent progress, SF also for the first time hit the +400 elo mark compared to  SF7: https://nextchessmove.com/dev-builds !!

Rob Ovenden

unread,
Jun 16, 2021, 12:03:12 PMJun 16
to FishCooking
 Elo doesn't really mean anything above 400, shouldn't NCM be basing the elo on Stockfish 10 or later instead of Stockfish 7?

Rob

Wolfgang de Cauter

unread,
Jun 16, 2021, 5:36:40 PMJun 16
to FishCooking
Indeed, SF 7 is too bad an opponent already.
A newer version as opponent might deliver better accuracy.

Damas Clásicas

unread,
Jun 17, 2021, 1:03:26 AMJun 17
to FishCooking

SF7 cannot be changed because ALL NCM's page stadistics are BASED in games of SF7 against all other SF versions.

If you change SF7 to whatever other version you want, this literally DESTROY all NCM's stadistics.

Wolfgang de Cauter

unread,
Jun 17, 2021, 2:49:32 AMJun 17
to FishCooking
Actually all old statistics can be kept as historical reference.
A second block of statistics can be build up let's say starting with the current dev version back to all dev versions since SF 11 against SF 12 as opponent.
I guess that the graphs of the historical block of statistics and of the newly built block of statistics would match more or less even on the new scale.
With SF 7 as opponent and a win/loss rate of ~200:1 or more and a win/draw rate of ~4,7:1 or more the old scale is getting useless fast.
Reply all
Reply to author
Forward
0 new messages