Did Rofchade adopt NNUE?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Madeleine Birchfield
Posts: 512
Joined: Tue Sep 29, 2020 4:29 pm
Location: Dublin, Ireland
Full name: Madeleine Birchfield

Did Rofchade adopt NNUE?

Post by Madeleine Birchfield »

People in TCEC and on this forum have said that Rofchade now uses NNUE, but when I checked the Rofchade website, there was no mention of NNUE at all. The latest version of Rofchade dates back to April of this year, which was before the computer chess community became aware of NNUE.
Alayan
Posts: 550
Joined: Tue Nov 19, 2019 8:48 pm
Full name: Alayan Feh

Re: Did Rofchade adopt NNUE?

Post by Alayan »

See http://talkchess.com/forum3/viewtopic.php?f=7&t=75219
Ronald wrote: Sun Sep 27, 2020 3:15 pm A nice game indeed!

I think that this hairy game was played with the regular rofChade version, but I'm not totally sure, because I had to fight some battles with xboard and directories etc. The version that is playing right now is the nnue version with the "latest" net from Sergio Vieri. On the Pi the speed loss of the net is bigger than on my Threadripper, so the difference looks a little bit less. The gain on my Threadripper seems immense, around 200 elo. A testmatch against Stockfish 11 (8000 games 60-0.6) resulted in minus 14 elo difference. Like I said earlier, after 70 years of experience with traditional evaluation functions, it looks like we still don't really understand what such a function should look like...
Raphexon
Posts: 476
Joined: Sun Mar 17, 2019 12:00 pm
Full name: Henk Drost

Re: Did Rofchade adopt NNUE?

Post by Raphexon »

Seems like my prediction that RofChade with NNUE > Leela wasn't even off mark.
(I made that statement before Jhorthos went berserk with his new nets)

Near SF11 at 60+0.6 probably translates to (easily) taking over at a more extended TC.