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
User avatar
MikeB
Posts: 4733
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Fat Fritz 2

Post by MikeB » Tue Feb 16, 2021 5:28 am

AndrewGrant wrote:
Tue Feb 16, 2021 5:17 am
dkappe wrote:
Tue Feb 16, 2021 3:55 am
kinderchocolate wrote:
Tue Feb 16, 2021 2:55 am
"... We created our own program, which we called Fat Fritz ..." that's exactly what Chessbase wrote on https://en.chessbase.com/post/chess-and ... lligence-1.

Chessbase is a big company, they surely have money to invest something on AI. They just need to hire a few data scientists...
ChessBase GMBH has 20 employees and $2 million in revenue per year. That’s not a big company.
Well thats pretty good for a company that seems to produce nothing
+1 LOL
Image

User avatar
MikeB
Posts: 4733
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Fat Fritz 2

Post by MikeB » Tue Feb 16, 2021 5:30 am

Official Statement by the Current Stockfish Maintainer

https://blog.stockfishchess.org/post/64 ... at-fritz-2
Image

dkappe
Posts: 900
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Fat Fritz 2

Post by dkappe » Tue Feb 16, 2021 6:20 am

AndrewGrant wrote:
Tue Feb 16, 2021 5:17 am
dkappe wrote:
Tue Feb 16, 2021 3:55 am
kinderchocolate wrote:
Tue Feb 16, 2021 2:55 am
"... We created our own program, which we called Fat Fritz ..." that's exactly what Chessbase wrote on https://en.chessbase.com/post/chess-and ... lligence-1.

Chessbase is a big company, they surely have money to invest something on AI. They just need to hire a few data scientists...
ChessBase GMBH has 20 employees and $2 million in revenue per year. That’s not a big company.
Well thats pretty good for a company that seems to produce nothing
$100k per employee? That not very good regardless. Stockfish has decimated the commercial chess software market.

User avatar
Tibono
Posts: 73
Joined: Sat Aug 01, 2015 4:16 pm
Location: France
Contact:

Re: Fat Fritz 2

Post by Tibono » Tue Feb 16, 2021 7:08 am

MikeB wrote:
Tue Feb 16, 2021 5:30 am
Official Statement by the Current Stockfish Maintainer

https://blog.stockfishchess.org/post/64 ... at-fritz-2
Thanks;
From this statement, I would point out
quote:"Furthermore, the engine Fat Fritz 2 fails to convince on independent rating lists, casting doubt on the usefulness of those modifications."
Indeed an additional argument worth mentioning, to undermine CB's doubtful communication.
so:
- rating the new chess entity matters
- naming it accordingly to what the average Joe buyer will clearly spot & understand does matter as well.
Kind regards,
Tibono

acepoint_de
Posts: 56
Joined: Mon Jun 10, 2013 11:14 pm

Re: Fat Fritz 2

Post by acepoint_de » Tue Feb 16, 2021 7:55 am

intrepid wrote:
Mon Feb 15, 2021 10:34 pm
Chessbase has a new article, "Chess and Artificial Intelligence (1)," and they made the mistake of allowing comments. I'm sure they're screening the messages for mention of "Stockfish" and the like, but there is a chance a meaningful comment might get through. The latest comment reads "It's just taking money out of people's pockets :-)" :lol:
They are already deleting some comments. I also wrote something now, let's wait and see.

Ciao

acepoint

pohl4711
Posts: 1545
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

Re: Fat Fritz 2

Post by pohl4711 » Tue Feb 16, 2021 8:34 am

AB-testrun of SF Fat Fritz 2 github finished.
Vondele (Stockfish team) made an avx2-compile of Fat Fritz 2 source code from github for me. I downloaded the free version of the FF2-net from GitHub. This is the freeware/GitHub version of Fat Fritz 2. Same conditions as in my testrun of Fat Fritz 2 commercial version. On my machine, Vondeles avx2-binary is around -7% slower, than the commercial FF2 avx2 binary. But that is no big deal, means an Elo-loss of (around) 3-4 Elo with my testing-conditions.

Fat Fritz 2 github has a suprisingly strong result in my testrun, compared to Fat Fritz 2 commercial. Of course, the github-version is weaker, but the difference is smaller as I expected.

https://www.sp-cc.de

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

User avatar
Ozymandias
Posts: 1323
Joined: Sun Oct 25, 2009 12:30 am

Re: Fat Fritz 2

Post by Ozymandias » Tue Feb 16, 2021 9:17 am

pohl4711 wrote:
Tue Feb 16, 2021 8:34 am
AB-testrun of SF Fat Fritz 2 github finished.
Vondele (Stockfish team) made an avx2-compile of Fat Fritz 2 source code from github for me. I downloaded the free version of the FF2-net from GitHub. This is the freeware/GitHub version of Fat Fritz 2. Same conditions as in my testrun of Fat Fritz 2 commercial version. On my machine, Vondeles avx2-binary is around -7% slower, than the commercial FF2 avx2 binary. But that is no big deal, means an Elo-loss of (around) 3-4 Elo with my testing-conditions.

Fat Fritz 2 github has a suprisingly strong result in my testrun, compared to Fat Fritz 2 commercial. Of course, the github-version is weaker, but the difference is smaller as I expected.

https://www.sp-cc.de

(Perhaps you have to clear your browsercache or reload the website)
If a sim test places the github net farther away from SF, it could be the one to have. No kidding.

User avatar
MikeB
Posts: 4733
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Fat Fritz 2

Post by MikeB » Tue Feb 16, 2021 9:19 am

Code: Select all

#########################################################################################################
###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/02160247.pgn
Time Control: Time Control-> base+inc: 30+0.500
Games: 2000
Threads: 1
Hash: 128

Current date : time (EDST)
Date: 02/16/21 : 04:02:25

Projected-> Time: 1h:15m:41s
     Run -> Time: 1h:14m:36s

2000 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 Oki-Maguro-12-R2-11-Eval   3505   0.0    8    8  2000 1028.0  51.4  340  284 1376  17.0  68.8  3495
   2 Fat-Fritz-2                3495   9.7    8    8  2000  972.0  48.6  284  340 1376  14.2  68.8  3505
---------------------------------------------------------------------------------------------------------

  Δ = delta from the next higher rated opponent
  # = number of games played
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
                          Ok Fa
Oki-Maguro-12-R2-11-Eval     98
Fat-Fritz-2                1

#########################################################################################################
###                                                End                                                ###
#########################################################################################################
The commercial version:
Image

To steal a line from the CB site, Big Tuna (Oki Maguro) is arguably the strongest entity that has ever played chess. ;>)
And It costs $0. Totally FREE, as it should be. Also if you do test Oki-Maguro, it stays with the Stockfish engines in the ratings list because it is a stockfish engine.. just like FF2.

Imagine if I said Oki-Maguro should be listed separately , can you imagine the backlash??? ... and it would be justifiable.

Will release this version tomorrow.

https://en.chessbase.com/post/fat-fritz ... oth-worlds
From CB site :
"Score of Fat Fritz 2 vs Stockfish 12:

286 wins / 99 losses / 1167 draws

Elo difference: 42.1 +/- 8.5, LOS: 100.0 %, DrawRatio: 75.2 %
1552 of 1552 games finished."
Image

JohnWoe
Posts: 347
Joined: Sat Mar 02, 2013 10:31 pm

Re: Fat Fritz 2

Post by JohnWoe » Tue Feb 16, 2021 9:38 am

I think we need a website called clonestockfish.com or smt.

As having authored a 3700 ELO chess engine is part of the human rights( clean water, freedom of speech, not to get tortured ... ) these days.

People can input a name (AI could suggest a catchy name if busy or on mobile phone).
A few sliders to select aggressiveness, style, default hash size etc.

After 5 min of hard clicking it would automatically upload the fresh 3700 ELO wonder to github. Job well done.

Stockfish w/o NNUE is still 3600 ELO. NNUE only gained ~100 ELO (not hundreds) as SF has a very good HCE. Even if you you fail to setup EvalFile you still have a 3600 ELO engine. Calling it just a "NNUE reader" is plain wrong.

User avatar
hgm
Posts: 26128
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Fat Fritz 2

Post by hgm » Tue Feb 16, 2021 9:52 am

It is not 'just a NNUE player'. But NNUE player is definitely one of its functions. Which does not use the built-in HCE. FF2 only uses SF in its capacity of NNUE player; it doesn't use the HCE.

Locked