We are the Computer Chess Community. We control ourselves.

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.
noobpwnftw
Posts: 465
Joined: Sun Nov 08, 2015 10:10 pm

Re: We are the Computer Chess Community. We control ourselves.

Post by noobpwnftw » Mon Feb 22, 2021 9:31 pm

the_real_greco wrote:
Mon Feb 22, 2021 7:57 pm
I'm confused about this. More supply arising due to increased demands makes sense, but why the street drugs exactly? Are you arguing that having engines on these ratings lists encourages cloning?
Yes.
the_real_greco wrote:
Mon Feb 22, 2021 7:57 pm
I'm sure that right now, Silver and ChessBase would prefer that FF2 not be on the CCRL or Pohl's. That was never their goal, and now it's just hurting them.
You mean like this?
https://en.chessbase.com/post/fat-fritz ... tereotypes

User avatar
hgm
Posts: 25824
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: We are the Computer Chess Community. We control ourselves.

Post by hgm » Mon Feb 22, 2021 9:32 pm

Wouldn't FF2 gain approximately equal Elo from using Stockfish 13 to run it instead of Stockfish 12, as Stockfish itself does?

noobpwnftw
Posts: 465
Joined: Sun Nov 08, 2015 10:10 pm

Re: We are the Computer Chess Community. We control ourselves.

Post by noobpwnftw » Mon Feb 22, 2021 9:36 pm

First, it did use SFdev as base, which is newer than SF12, second, the FF2 net is about 20~50 elo weaker than SF's own best net so that pretty much counteracts the gaining, therefore they had to compare it with SF12 to show an advantage.

As for itself, they could've used CFish to run the net, which would be even stronger, but that is probably too much work because it has architecture-specific optimizations that AS unable to invent.

the_real_greco
Posts: 62
Joined: Thu Sep 19, 2019 3:55 am
Full name: Andy!

Re: We are the Computer Chess Community. We control ourselves.

Post by the_real_greco » Mon Feb 22, 2021 9:40 pm

noobpwnftw wrote:
Mon Feb 22, 2021 9:31 pm
the_real_greco wrote:
Mon Feb 22, 2021 7:57 pm
I'm confused about this. More supply arising due to increased demands makes sense, but why the street drugs exactly? Are you arguing that having engines on these ratings lists encourages cloning?
Yes.
the_real_greco wrote:
Mon Feb 22, 2021 7:57 pm
I'm sure that right now, Silver and ChessBase would prefer that FF2 not be on the CCRL or Pohl's. That was never their goal, and now it's just hurting them.
You mean like this?
https://en.chessbase.com/post/fat-fritz ... tereotypes

User avatar
Graham Banks
Posts: 35040
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: We are the Computer Chess Community. We control ourselves.

Post by Graham Banks » Mon Feb 22, 2021 9:45 pm

noobpwnftw wrote:
Mon Feb 22, 2021 9:36 pm
First, it did use SFdev as base, which is newer than SF12, second, the FF2 net is about 20~50 elo weaker than SF's own best net so that pretty much counteracts the gaining, therefore they had to compare it with SF12 to show an advantage.

As for itself, they could've used CFish to run the net, which would be even stronger, but that is probably too much work because it has architecture-specific optimizations that AS unable to invent.
I suspect that the FF2 net that comes with the purchase is fairly close in strength to SF13, perhaps 10-15 Elo either way.

My 8CPU match between the two only has 50 games at 2 hours with 30 second increments, but there was 1 FF2 win and 49 draws.
gbanksnz at gmail.com

noobpwnftw
Posts: 465
Joined: Sun Nov 08, 2015 10:10 pm

Re: We are the Computer Chess Community. We control ourselves.

Post by noobpwnftw » Mon Feb 22, 2021 9:51 pm

People did some tests and here are the results:
https://discord.com/channels/4359437104 ... 0529749014

Code: Select all

Official test by SF team, at fishtest conditions, of Stockfish 13 vs
the sold version of CB/AS (ff2) net and the corresponding Stockfish player.
===========================================================================

TC 10+0.1
=========

games at: https://drive.google.com/file/d/14TVxuyOGWAVc2kJyR6K7DEUOetZfWplm/view?usp=sharing

../ordo -a 0 -p out-sf13-vs-ff2-TC10+0.1.pgn -A sf13 -s 100 -q  -J -D -W

   # PLAYER    :  RATING  ERROR   POINTS  PLAYED   (%)  CFS(%)
   1 sf13      :     0.0   ----  34445.5   65536    53     100
   2 ff2       :   -18.1    1.3  31090.5   65536    47     ---

White advantage = 29.29 +/- 0.67
Draw rate (equal opponents) = 77.35 % +/- 0.18

TC 60+0.6
=========

games at: https://drive.google.com/file/d/10yOOoCqt12sssmKy3jSI7og-M-nNs_Xs/view?usp=sharing

../ordo -a 0 -p out-sf13-vs-ff2-TC60+0.6.pgn -A sf13 -s 100 -q  -J -D -W

   # PLAYER    :  RATING  ERROR   POINTS  PLAYED   (%)  CFS(%)
   1 sf13      :     0.0   ----  33774.0   65536    52     100
   2 ff2       :   -10.8    1.0  31762.0   65536    48     ---

White advantage = 24.72 +/- 0.50
Draw rate (equal opponents) = 87.77 % +/- 0.14

Conclusion
==========

The ff2 net is weaker than SF13 in a proper testing setup.
Draw rates are comparable to fishtest results.
Setup and parameters
====================

Same setup as fishtest, using 8moves_v3.pgn book, run in 128 batches (command used for TC60+0.6):

../cutechess-cli -repeat -rounds 1 -games 512 -tournament round-robin -srand $RANDOM -resign movecount=3 score=400 -draw movenumber=34 movecount=8 score=20\
                 -concurrency 64 -openings file=../8moves_v3.pgn format=pgn order=random plies=16\
                 -each tc=60+0.6 option.hash=64 proto=uci -ratinginterval 128\
                 -engine cmd=../stockfish.sf13 name=sf13 \
                 -engine cmd=../stockfish.ff2 name=ff2 \
                 -pgnout out-sf13-vs-ff2.pgn

Corresponding to SF13 and FF2 as available Feb 21st

../stockfish.ff2 bench 2>&1 | grep "Nodes searched"
Nodes searched  : 4025201

../stockfish.sf13 bench 2>&1 | grep "Nodes searched"
Nodes searched  : 3766422

Compiled using exactly the same compiler and same settings

on master at SF13 commit
make clean && make -j ARCH=x86-64-avx2 profile-build && mv ./stockfish ./stockfish.sf13

on ff2 branch with the CB/AS net added to the src directory
make clean && make -j ARCH=x86-64-avx2 profile-build && mv ./stockfish ./stockfish.ff2
Note that by this standard of testing condition, SF13 is +36elo over SF12, while using some other book it can go as high as +70 elo.

Your mileage may vary but I don't see it flips the sign. Nevertheless, so unfortunate that we'll have to argue over the illegal(or disputed) clones in the first place.

User avatar
Graham Banks
Posts: 35040
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: We are the Computer Chess Community. We control ourselves.

Post by Graham Banks » Mon Feb 22, 2021 9:56 pm

noobpwnftw wrote:
Mon Feb 22, 2021 9:51 pm
People did some tests and here are the results:
https://discord.com/channels/4359437104 ... 0529749014

Code: Select all

Official test by SF team, at fishtest conditions, of Stockfish 13 vs
the sold version of CB/AS (ff2) net and the corresponding Stockfish player.
===========================================================================

TC 10+0.1
=========

games at: https://drive.google.com/file/d/14TVxuyOGWAVc2kJyR6K7DEUOetZfWplm/view?usp=sharing

../ordo -a 0 -p out-sf13-vs-ff2-TC10+0.1.pgn -A sf13 -s 100 -q  -J -D -W

   # PLAYER    :  RATING  ERROR   POINTS  PLAYED   (%)  CFS(%)
   1 sf13      :     0.0   ----  34445.5   65536    53     100
   2 ff2       :   -18.1    1.3  31090.5   65536    47     ---

White advantage = 29.29 +/- 0.67
Draw rate (equal opponents) = 77.35 % +/- 0.18

TC 60+0.6
=========

games at: https://drive.google.com/file/d/10yOOoCqt12sssmKy3jSI7og-M-nNs_Xs/view?usp=sharing

../ordo -a 0 -p out-sf13-vs-ff2-TC60+0.6.pgn -A sf13 -s 100 -q  -J -D -W

   # PLAYER    :  RATING  ERROR   POINTS  PLAYED   (%)  CFS(%)
   1 sf13      :     0.0   ----  33774.0   65536    52     100
   2 ff2       :   -10.8    1.0  31762.0   65536    48     ---

White advantage = 24.72 +/- 0.50
Draw rate (equal opponents) = 87.77 % +/- 0.14

Conclusion
==========

The ff2 net is weaker than SF13 in a proper testing setup.
Draw rates are comparable to fishtest results.
Setup and parameters
====================

Same setup as fishtest, using 8moves_v3.pgn book, run in 128 batches (command used for TC60+0.6):

../cutechess-cli -repeat -rounds 1 -games 512 -tournament round-robin -srand $RANDOM -resign movecount=3 score=400 -draw movenumber=34 movecount=8 score=20\
                 -concurrency 64 -openings file=../8moves_v3.pgn format=pgn order=random plies=16\
                 -each tc=60+0.6 option.hash=64 proto=uci -ratinginterval 128\
                 -engine cmd=../stockfish.sf13 name=sf13 \
                 -engine cmd=../stockfish.ff2 name=ff2 \
                 -pgnout out-sf13-vs-ff2.pgn

Corresponding to SF13 and FF2 as available Feb 21st

../stockfish.ff2 bench 2>&1 | grep "Nodes searched"
Nodes searched  : 4025201

../stockfish.sf13 bench 2>&1 | grep "Nodes searched"
Nodes searched  : 3766422

Compiled using exactly the same compiler and same settings

on master at SF13 commit
make clean && make -j ARCH=x86-64-avx2 profile-build && mv ./stockfish ./stockfish.sf13

on ff2 branch with the CB/AS net added to the src directory
make clean && make -j ARCH=x86-64-avx2 profile-build && mv ./stockfish ./stockfish.ff2
Note that by this standard of testing condition, SF13 is +36elo over SF12, while using some other book it can go as high as +70 elo.

Your mileage may vary but I don't see it flips the sign. Nevertheless, so unfortunate that we'll have to argue over the illegal(or disputed) clones in the first place.
Not meaning to argue with you, but I assume they were using the purchased net and not the free one?
gbanksnz at gmail.com

noobpwnftw
Posts: 465
Joined: Sun Nov 08, 2015 10:10 pm

Re: We are the Computer Chess Community. We control ourselves.

Post by noobpwnftw » Mon Feb 22, 2021 10:01 pm

Graham Banks wrote:
Mon Feb 22, 2021 9:56 pm
Not meaning to argue with you, but I assume they were using the purchased net and not the free one?
Yes, as shown in bench numbers. The free one(bench: 4511552) is obviously weaker.

The problem however is that people have to legally obtain FF2 to do a test, almost feels like a painter has to buy the fake work of his own.

BTW:
There is a also net trained after AS's "innovative" arch, judging by the numbers I say it is probably also stronger than FF2 but still not quite there to beat SF13. Results: https://tests.stockfishchess.org/tests/ ... 561bc49984

He said that people called it a dead-end, I say people went further than him even on that same dead-end.

User avatar
towforce
Posts: 10745
Joined: Wed Mar 08, 2006 11:57 pm
Location: Birmingham UK

Re: We are the Computer Chess Community. We control ourselves.

Post by towforce » Mon Feb 22, 2021 11:17 pm

Guenther wrote:
Mon Feb 22, 2021 6:16 pm
towforce wrote:
Mon Feb 22, 2021 2:58 pm
pohl4711 wrote:
Mon Feb 22, 2021 1:50 pm
Totally wrong Imagination. I always paid for the commercial engines in my ratinglist. Only exception was Fat Fritz 2, I got a free testversion from Albert Silver. But otherwise, I would have bought it, because I think the concept of the Fat Fritz 2 net is interesting.
I bought Houdini from version 2 on. And I am a Komodo Subscriber since years. I bought Fritz 16, Fritz 17 and Fat Fritz 1.
Because, I am deciding, what I am interested in and what I am testing. And nobody can and will tell me, what I am allowed to test. Thats why I am paying for engines: The only way to stay really free and independent as a tester is the willingness to pay for commercial engines. I am tester and not a judge. If anybody doesnt like my testwork - no problem. Just do not visit my website. But nobody will tell me, what I am allowed to test. For a tester, independence and transparency are the most important virtues. Not pretending being a judge, like so many others here on talkchess and in the computerchess community do.

If anyone believes, a software is illegal: go to court and win. Case closed. Everything else is trashtalk (best case) or a witch hunt (worst case). And thats not my business. Ending this is the change, the community badly needs. My 2 cents.

Spirited and classy. The test community fights back! :twisted:
No. FatFritz2 testers obviously are a different brand and need to defend FatFritz2 by all means - not surprising.
(and they are always non-programmers too, a coincidence?)

In some ways, coding and testing require a somewhat different mentality.

If I was a tester, I would test FF2: it seems to be overhyped and immoral, and you obviously disapprove of it, but there is public interest in its playing strength, so absolutely I would test it.

IMO, there are too many programs on test lists now, so I would offer multiple filters on a test list that I produced. Not sure how I would classify FF2 - maybe something like "celebrity fork".

I could easily reel off several examples of a brand name reselling the products of an unknown brand (I'm wearing one on my wrist right now - a fitness tracker made by Huami but sold under the Xiaomi brand name. Huami also make Amazon's "Amazfit" trackers) - though usually, of course, both parties benefit. It is possible that Stockfish could one day benefit: in the unlikely event of the "Chessbase" brand outlasting the "Stockfish" brand, products like FF2 would remind people of Stockfish's existence, and the important role it's playing at the present time.

Does anyone understand what witch hunting has to do with the facts about FF2?
(and the lies in the beginning and the marketing scam, which goes down to pieces now slowly)
Probably he still believes in a witchhunt vs. Houdini too?

The sentence 'I am a tester not a judge' leaves some bitter aftertaste, it is quite analogue to a lot of horrible wrongdoings in mankind.
Writing is the antidote to confusion

dkappe
Posts: 817
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: We are the Computer Chess Community. We control ourselves.

Post by dkappe » Tue Feb 23, 2021 1:59 am

noobpwnftw wrote:
Mon Feb 22, 2021 9:36 pm
First, it did use SFdev as base, which is newer than SF12, second, the FF2 net is about 20~50 elo weaker than SF's own best net so that pretty much counteracts the gaining, therefore they had to compare it with SF12 to show an advantage.

As for itself, they could've used CFish to run the net, which would be even stronger, but that is probably too much work because it has architecture-specific optimizations that AS unable to invent.
You’re talking through your hat as usual. Even the fishtest scam didn’t show that.

Post Reply