Fat Fritz 2

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

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Locked
Jouni
Posts: 2432
Joined: Wed Mar 08, 2006 7:15 pm

Fat Fritz 2

Post by Jouni » Tue Feb 09, 2021 11:23 am

https://de.chessbase.com/post/fat-fritz ... e-nummer-1

Commercial version of SF? I smell SF dev here :) .
Jouni

Modern Times
Posts: 2773
Joined: Thu Jun 07, 2012 9:02 pm

Re: Fat Fritz 2

Post by Modern Times » Tue Feb 09, 2021 11:28 am

Initial results are on the CCRL lists. Chess960 was its strongest performance so far.

Yes it is modified Stockfish as it says in the description, and they say the source is available.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

marsell
Posts: 68
Joined: Tue Feb 07, 2012 10:14 am

Re: Fat Fritz 2

Post by marsell » Tue Feb 09, 2021 11:59 am

Jouni wrote:
Tue Feb 09, 2021 11:23 am


Commercial version of SF? I smell SF dev here :) .
I should try this too. Train a net, rename Stockfish and then sell it.

User avatar
Guenther
Posts: 3944
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Fat Fritz 2

Post by Guenther » Tue Feb 09, 2021 11:59 am

Modern Times wrote:
Tue Feb 09, 2021 11:28 am
Initial results are on the CCRL lists. Chess960 was its strongest performance so far.

Yes it is modified Stockfish as it says in the description, and they say the source is available.
Typical Chessbase nonsense in the text too. Calling it the new number one with complete lack of independent testing is plain BS.
Also their own biased test is based on a match vs. old SF 12 and not a current SF dev version.
Typically Silver driven - at least with FatFritz(1) aka LC0 (+self-trained NN) they were still too shy to sell it and just added it as gift to another commercial package.

From CB Germany
https://de.chessbase.com/post/fat-fritz ... e-nummer-1
Die neue Nr. 1:

Fat Fritz 2.0 basiert auf der Engine Stockfish 12, verwendet aber ein völlig neues neuronales Netzwerk mit besseren Bewertungen und größerem Leistungsumfang.

Dank neuer NNUE-Technologie:

Fat Fritz 2.0 erreicht auch ohne high-end Grafikkarte Top-Leistung!

Fat Fritz 2.0 klar vorne!

Im Vergleichsmatch über 1552 Partien besiegt Fat Fritz 2.0 Stockfish 12 deutlich (286 Siege bei 99 Niederlagen, Rest Remis) und übertrifft die bisherige Spitzenengine um über 40 Elo-Punkte (Stand Februar 2021).
Translation (by me) of the title and the second part after the biz bla-bla:
The new number one!
...
Fat Fritz 2.0 clearly ahead!

In a comparison match over 1552 games FF 2.0 wins clearly vs. SF12 (286 wins, 99 losses, others draw) and is now more than 40 rating points
in front of the previous top engine. (As of February 2021)
Hopefully CCRL lists it as one of the dozens SF derivates...

I guess the (again nearly unmodified) source again lies in a repo of Daniel Uranga ...( not yet checked)
https://rwbc-chess.de
'chessqueen' 2018-present, aka: 'George' 2013-2016, 'pichy' 2006-2013, 'Jorge Pichard' 2000-2006 (old forum)
Troll barometer:
https://docs.google.com/spreadsheets/d/ ... KSptBx9AUs

Frank Quisinsky
Posts: 5592
Joined: Wed Nov 18, 2009 6:16 pm
Location: Trier, Germany
Full name: Frank Quisinsky
Contact:

Re: Fat Fritz 2

Post by Frank Quisinsky » Tue Feb 09, 2021 12:18 pm

Hi there,

same opinion!
Rybka, Houdini and now this one!
The question is:
Should I delete the links to Chessbase from the webpage?

Other products are really good (example: Chessbase database).

Best
Frank
I like computer chess!

marsell
Posts: 68
Joined: Tue Feb 07, 2012 10:14 am

Re: Fat Fritz 2

Post by marsell » Tue Feb 09, 2021 12:36 pm

Frank Quisinsky wrote:
Tue Feb 09, 2021 12:18 pm

Rybka, Houdini and now this one!
Rybka and Houdini were at least always number one engines in those days!

Modern Times
Posts: 2773
Joined: Thu Jun 07, 2012 9:02 pm

Re: Fat Fritz 2

Post by Modern Times » Tue Feb 09, 2021 1:18 pm

Guenther wrote:
Tue Feb 09, 2021 11:59 am
Typical Chessbase nonsense in the text too. Calling it the new number one
That is indeed a bold statement, but it came very very close on our chess960 list at least.

However chess engines are about more than just ratings lists. Perhaps it brings something extra to the table in analysis use ?
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

dkappe
Posts: 901
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Fat Fritz 2

Post by dkappe » Tue Feb 09, 2021 2:10 pm

marsell wrote:
Tue Feb 09, 2021 11:59 am
Jouni wrote:
Tue Feb 09, 2021 11:23 am


Commercial version of SF? I smell SF dev here :) .
I should try this too. Train a net, rename Stockfish and then sell it.
You could just train a net, or three, like me. Sell it. Don’t sell it. Run it on your own engine(s). Freedom.

pohl4711
Posts: 1547
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

Re: Fat Fritz 2

Post by pohl4711 » Tue Feb 09, 2021 2:20 pm

Jouni wrote:
Tue Feb 09, 2021 11:23 am
https://de.chessbase.com/post/fat-fritz ... e-nummer-1

Commercial version of SF? I smell SF dev here :) .
Not really.
I am testing Fat Fritz 2 right now. Fat Fritz 2 uses a nnue net, which is 2x bigger, than the Stockfish nnue-nets. Because of this, the binary is more than 40MB in size and the nps are measureable lower on my system (AMD Ryzen 12core CPU):

singlethread nps:
Fat Fritz 2 avx2: 1.15 MN/s
Stockfish 210131 avx2: 1.8 MN/s

22 Threads (11 cores, with huge hashtables (8 GB)):
Fat Fritz 2 avx2: 16 MN/s
Stockfish 210131 avx2: 18 MN/s

Fat Fritz uses the SF binary, true, but a complete different nnue-net (double size!). So, FF2 is definitly more than a Stockfish clone like Eman, RaubFish, CorChess and all the others...


More information here:
https://en.chessbase.com/post/92134

"Fat Fritz 2 is trained and developed using this NNUE technology, running inside the open source Stockfish binary, but with a completely new and different neural network that differs significantly from Stockfish's on a couple of key points:
The first is that the evaluations and positions did not come from a brute force engine such as Stockfish or Komodo, but rather came from our original neural network Fat Fritz.
The second key difference is that the neural network in Fat Fritz 2 is double the size of Stockfish 12, using 512 neurons instead of 256, offering the opportunity for double the knowledge and understanding."

IMHO these 2 differences makes Fat Fritz 2 a very interesting new approach. Of course based on Stockfish and the nnue-papers. But definitly more than a simple Stockfish Clone. As far as I know, until now, nobody built such a big nnue-net and nobody built a nnue-net based on evaluations given by a Lc0 neural net.
For me Fat Fritz 2 is much more interesting and innovative than Fat Fritz 1.

Modern Times
Posts: 2773
Joined: Thu Jun 07, 2012 9:02 pm

Re: Fat Fritz 2

Post by Modern Times » Tue Feb 09, 2021 2:39 pm

pohl4711 wrote:
Tue Feb 09, 2021 2:20 pm

"Fat Fritz 2 is trained and developed using this NNUE technology, running inside the open source Stockfish binary, but with a completely new and different neural network that differs significantly from Stockfish's on a couple of key points:
The first is that the evaluations and positions did not come from a brute force engine such as Stockfish or Komodo, but rather came from our original neural network Fat Fritz.
The second key difference is that the neural network in Fat Fritz 2 is double the size of Stockfish 12, using 512 neurons instead of 256, offering the opportunity for double the knowledge and understanding."

IMHO these 2 differences makes Fat Fritz 2 a very interesting new approach. Of course based on Stockfish and the nnue-papers. But definitly more than a simple Stockfish Clone. As far as I know, until now, nobody built such a big nnue-net based on evaluations given by a Lc0 neural net.
Yes I agree. It was inevitable that "Stockfsh clone" would be the response from a few people, but this is very much more than that.

I will be interested in your results. But I think ratings lists aren't the whole picture. I'd like to see how it performs in test suites compared to Stockfish and Komodo Dragon for example.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

Locked