Heh heh. Obviously my tongue-in-cheek post has suddenly come reality! Lc0 is leading TCEC superfinal even with some 5-7 clear blunders .
I wonder if these tactical blunders can't be prevented with doing single CPU AB search by any decent engine?
Jouni wrote: ↑Sun Feb 17, 2019 10:43 pm
Heh heh. Obviously my tongue-in-cheek post has suddenly come reality! Lc0 is leading TCEC superfinal even with some 5-7 clear blunders .
I wonder if these tactical blunders can't be prevented with doing single CPU AB search by any decent engine?
It is a probable result. Latest tests in Leela discord estimated that SF vs Leela were within 20 elo, with SF in favour. It is now very likely that Leela scale better with specific cput setting for Sufi.
So much tests has been done in Leela discord to get optimized setting for this Final although SF developers dont even test contempt/without contempt against Leela.
Looks like those optimizations helps for another +10 to +20 elo in VLTC setting.
Nay Lin Tun wrote: ↑Tue Feb 19, 2019 8:20 amSo much tests has been done in Leela discord to get optimized setting for this Final although SF developers dont even test contempt/without contempt against Leela.
That on its own is stretching the meaning of "zero". (But it's already non-zero.)
Should just call it Lc1.
CMCanavessi wrote: ↑Thu Feb 14, 2019 8:44 pm
Leela about to go +3, SF fans already using excuses like "hardware disadvantage" or "no book"... gimme a break.
The same guys who made fun about the argument, Lc0 would be better without book?
You don't think that lc0 plays better from the start position than from a 8-16-move predefined opening? You should check current bonus matches at TCEC...