CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Rebel, chrisw

Vernon Crawford
Posts: 73
Joined: Wed Sep 01, 2021 2:05 am
Location: London, England
Full name: Vernon Crawford

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Vernon Crawford »

Guenther wrote: Mon Sep 13, 2021 7:26 pm
Rebel wrote: Mon Sep 13, 2021 4:24 pm Well, well, your 6th post and what a language use.

I have nothing but respect for CCRL, they know. But criticism is allowed.

Using an old outdated net for versions 26,27 and 28 can't be good.
Well actually it is his 7th post, one was deleted, because he called me 'social misfit',
after I finally called 'alexchess' a moron for again and again polluting my new engine releases
thread with crap engines.
(BTW he did it again today! I will remain with full confidence at my moron calling).

For me it is completely clear he (Vernon X, or whatever his real name is) came here for trolling.
Also at the same day he appeared here, he opened his own 'forum' which offers all kind of illegal derivates
and clones for download, even the old Patriot! one of only two illegal clones/derivates, which ever tried
to be commercial.

I don't think there is much more to say regarding him...
https://engine-chess.com/index.php
Posted the day I opened the forum:

"These Engines
Post by Vernon Crawford » Sun Sep 05, 2021 1:27 pm

I believe the engines presented here for download do not violate copyright, etc.
If I'm mistaken about one or another, please let me know and I'll remove it ASAP.

Vernon"

Guenther if there's something there that has got your panties in a knot...
just letting me know in a respectful friendly manner would be appreciated.

(but I do realize from what I've read here so far, that "respectful friendly manner" is going to be difficult for you.)
User avatar
Guenther
Posts: 4605
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Guenther »

Vernon Crawford wrote: Mon Sep 13, 2021 8:41 pm
Guenther wrote: Mon Sep 13, 2021 7:26 pm
Rebel wrote: Mon Sep 13, 2021 4:24 pm Well, well, your 6th post and what a language use.

I have nothing but respect for CCRL, they know. But criticism is allowed.

Using an old outdated net for versions 26,27 and 28 can't be good.
Well actually it is his 7th post, one was deleted, because he called me 'social misfit',
after I finally called 'alexchess' a moron for again and again polluting my new engine releases
thread with crap engines.
(BTW he did it again today! I will remain with full confidence at my moron calling).

For me it is completely clear he (Vernon X, or whatever his real name is) came here for trolling.
Also at the same day he appeared here, he opened his own 'forum' which offers all kind of illegal derivates
and clones for download, even the old Patriot! one of only two illegal clones/derivates, which ever tried
to be commercial.

I don't think there is much more to say regarding him...
https://engine-chess.com/index.php
Posted the day I opened the forum:

"These Engines
Post by Vernon Crawford » Sun Sep 05, 2021 1:27 pm

I believe the engines presented here for download do not violate copyright, etc.
If I'm mistaken about one or another, please let me know and I'll remove it ASAP.

Vernon"

Guenther if there's something there that has got your panties in a knot...
just letting me know in a respectful friendly manner would be appreciated.

(but I do realize from what I've read here so far, that "respectful friendly manner" is going to be difficult for you.)
To put it politely, I don't think your account will live that long here...

You don't grasp that respect has to be earned by something - showing up here and immediately defending another
verified troll doesn't qualify for that. You won't silence the truth.
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy
connor_mcmonigle
Posts: 530
Joined: Sun Sep 06, 2020 4:40 am
Full name: Connor McMonigle

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by connor_mcmonigle »

Vernon Crawford wrote: Mon Sep 13, 2021 8:38 pm Yes, a friendly comment or suggestion certainly would very likely be appreciated...
but this link that was posted was an accusation of incompetence or deliberate sabotage on LC0 rating
and it was posted without any discretion or 2nd thought.

There's a big difference..
Maybe Ed could have avoided linking to the discussion on his forum. Regardless, your reaction was laughably extreme. It would be nice to have an explanation from CCRL as to why they keep replacing the networks bundled with the Lc0 releases with a very old, much weaker, network.
User avatar
Guenther
Posts: 4605
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Guenther »

connor_mcmonigle wrote: Mon Sep 13, 2021 9:33 pm
Vernon Crawford wrote: Mon Sep 13, 2021 8:38 pm Yes, a friendly comment or suggestion certainly would very likely be appreciated...
but this link that was posted was an accusation of incompetence or deliberate sabotage on LC0 rating
and it was posted without any discretion or 2nd thought.

There's a big difference..
Maybe Ed could have avoided linking to the discussion on his forum. Regardless, your reaction was laughably extreme. It would be nice to have an explanation from CCRL as to why they keep replacing the networks bundled with the Lc0 releases with a very old, much weaker, network.
Actually it is nearly 21 months old. Imagine SF (since NNUE appeared) would always be tested with the first appeared NNUE by someone.
It somehow makes the goofy impression it is unknown that the playing strength comes mainly from the network and not from the LC0 binary.

Code: Select all

1	Lc0 0.28.0 t40-1541 RTX2080	3669	+15	−15	64.4%	−88.0	59.3%	1389	 
 	Lc0 0.26.3 t40-1541 RTX2080	3651	+16	−16	59.7%	−61.2	62.4%	1094
 	Lc0 0.27.0 t40-1541 RTX2080	3647	+13	−13	60.1%	−62.4	64.8%	1649
 	Lc0 0.25.1 t40-1541 RTX2080	3645	+15	−15	64.4%	−94.4	59.4%	1389
 	Lc0 0.23.1 w42850 RTX2080	3620	+13	−13	62.1%	−74.3	67.8%	1789
 	Lc0 0.22.0 w49921 RTX2080	3596	+15	−15	63.8%	−82.8	65.5%	1292
OTH I bet Stoofvlees is always tested with a different net. (But actually it is unclear what 'a X' means?)

Code: Select all

1	Stoofvlees II a18 RTX2080	3664	+14	−14	59.8%	−60.9	61.3%	1524	 
 	Stoofvlees II a16 RTX2080	3636	+15	−15	54.3%	−29.1	62.6%	1327
 	Stoofvlees II a14 RTX2080	3568	+12	−12	49.7%	−3.9	57.8%	2202
 	Stoofvlees II a13 RTX2080	3554	+17	−17	49.1%	−10.7	58.6%	1017
 	Stoofvlees II a12 RTX2080	3530	+16	−16	50.9%	−10.6	53.5%	1103
 	Stoofvlees II a11 RTX2080	3484	+30	−30	60.3%	−72.3	36.2%	370
 
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy
Albert Silver
Posts: 3019
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Albert Silver »

Guenther wrote: Mon Sep 13, 2021 9:48 pm
OTH I bet Stoofvlees is always tested with a different net.
OTOH I bet Stoofvlees has an author who povides them with all the components to test, and highlights when and what comprises a new version. Leela used to have a clear net to test: the last net of T10, T30, and T40. Nowadays the only clear upgrades representing it are the new binaries. The nets are all over the place, sprouting up in Discord, forums and more. Good luck trying to figure out which one presumably is the proper representative of the Leela project.

Have any of their key developers come forward to declare net XYZ as the current flagship of Leela? Demanding the tester go around digging through that is ridiculous. It is up to Leela's devs to provide a clear package to test. It is worth adding that because all the testers are free to choose and test what they want, there is no 'CCRL' policy on any engine or net.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
User avatar
Rebel
Posts: 6991
Joined: Thu Aug 18, 2011 12:04 pm

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Rebel »

Albert Silver wrote: Tue Sep 14, 2021 8:32 pm
Guenther wrote: Mon Sep 13, 2021 9:48 pm
OTH I bet Stoofvlees is always tested with a different net.
OTOH I bet Stoofvlees has an author who povides them with all the components to test, and highlights when and what comprises a new version. Leela used to have a clear net to test: the last net of T10, T30, and T40. Nowadays the only clear upgrades representing it are the new binaries. The nets are all over the place, sprouting up in Discord, forums and more. Good luck trying to figure out which one presumably is the proper representative of the Leela project.

Have any of their key developers come forward to declare net XYZ as the current flagship of Leela? Demanding the tester go around digging through that is ridiculous. It is up to Leela's devs to provide a clear package to test. It is worth adding that because all the testers are free to choose and test what they want, there is no 'CCRL' policy on any engine or net.
https://lczero.org/dev/wiki/best-nets-for-lc0/
90% of coding is debugging, the other 10% is writing bugs.
connor_mcmonigle
Posts: 530
Joined: Sun Sep 06, 2020 4:40 am
Full name: Connor McMonigle

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by connor_mcmonigle »

Albert Silver wrote: Tue Sep 14, 2021 8:32 pm
Guenther wrote: Mon Sep 13, 2021 9:48 pm
OTH I bet Stoofvlees is always tested with a different net.
OTOH I bet Stoofvlees has an author who povides them with all the components to test, and highlights when and what comprises a new version. Leela used to have a clear net to test: the last net of T10, T30, and T40. Nowadays the only clear upgrades representing it are the new binaries. The nets are all over the place, sprouting up in Discord, forums and more. Good luck trying to figure out which one presumably is the proper representative of the Leela project.

Have any of their key developers come forward to declare net XYZ as the current flagship of Leela? Demanding the tester go around digging through that is ridiculous. It is up to Leela's devs to provide a clear package to test. It is worth adding that because all the testers are free to choose and test what they want, there is no 'CCRL' policy on any engine or net.
You're spouting misinformation.
The network understood to be strongest at the time is packaged with each release: https://github.com/LeelaChessZero/lc0/r ... ag/v0.28.0. The networks are not "all over the place". It's completely reasonable to expect an engine to be tested with the network literally included in the release.
Albert Silver
Posts: 3019
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Albert Silver »

connor_mcmonigle wrote: Tue Sep 14, 2021 9:25 pm
Albert Silver wrote: Tue Sep 14, 2021 8:32 pm
Guenther wrote: Mon Sep 13, 2021 9:48 pm
OTH I bet Stoofvlees is always tested with a different net.
OTOH I bet Stoofvlees has an author who povides them with all the components to test, and highlights when and what comprises a new version. Leela used to have a clear net to test: the last net of T10, T30, and T40. Nowadays the only clear upgrades representing it are the new binaries. The nets are all over the place, sprouting up in Discord, forums and more. Good luck trying to figure out which one presumably is the proper representative of the Leela project.

Have any of their key developers come forward to declare net XYZ as the current flagship of Leela? Demanding the tester go around digging through that is ridiculous. It is up to Leela's devs to provide a clear package to test. It is worth adding that because all the testers are free to choose and test what they want, there is no 'CCRL' policy on any engine or net.
You're spouting misinformation.
The network understood to be strongest at the time is packaged with each release: https://github.com/LeelaChessZero/lc0/r ... ag/v0.28.0. The networks are not "all over the place". It's completely reasonable to expect an engine to be tested with the network literally included in the release.
I'm afraid you're misinformed. The net that is packaged is a smaller demo net. It is not even the official strongest net. Much less the strongest net.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
Albert Silver
Posts: 3019
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Albert Silver »

Rebel wrote: Tue Sep 14, 2021 8:47 pm
Albert Silver wrote: Tue Sep 14, 2021 8:32 pm
Guenther wrote: Mon Sep 13, 2021 9:48 pm
OTH I bet Stoofvlees is always tested with a different net.
OTOH I bet Stoofvlees has an author who povides them with all the components to test, and highlights when and what comprises a new version. Leela used to have a clear net to test: the last net of T10, T30, and T40. Nowadays the only clear upgrades representing it are the new binaries. The nets are all over the place, sprouting up in Discord, forums and more. Good luck trying to figure out which one presumably is the proper representative of the Leela project.

Have any of their key developers come forward to declare net XYZ as the current flagship of Leela? Demanding the tester go around digging through that is ridiculous. It is up to Leela's devs to provide a clear package to test. It is worth adding that because all the testers are free to choose and test what they want, there is no 'CCRL' policy on any engine or net.
https://lczero.org/dev/wiki/best-nets-for-lc0/
Did you even read that link? Because there is no clear nets designated to represent Leela in it. Instead it just says that there are a number of nets that can be recommended for different purposes. It isn't up to the tester to try to figure out which one should be the one to represent Leela. It is the developers we should be doing that. If they can't be bothered, then they can't complain when someone doesn't use the one they prefer.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
User avatar
Guenther
Posts: 4605
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: CCRL 40/15, 40/2 and FRC lists updated (11th September 2021)

Post by Guenther »

Albert Silver wrote: Tue Sep 14, 2021 9:48 pm
connor_mcmonigle wrote: Tue Sep 14, 2021 9:25 pm
Albert Silver wrote: Tue Sep 14, 2021 8:32 pm
Guenther wrote: Mon Sep 13, 2021 9:48 pm
OTH I bet Stoofvlees is always tested with a different net.
OTOH I bet Stoofvlees has an author who povides them with all the components to test, and highlights when and what comprises a new version. Leela used to have a clear net to test: the last net of T10, T30, and T40. Nowadays the only clear upgrades representing it are the new binaries. The nets are all over the place, sprouting up in Discord, forums and more. Good luck trying to figure out which one presumably is the proper representative of the Leela project.

Have any of their key developers come forward to declare net XYZ as the current flagship of Leela? Demanding the tester go around digging through that is ridiculous. It is up to Leela's devs to provide a clear package to test. It is worth adding that because all the testers are free to choose and test what they want, there is no 'CCRL' policy on any engine or net.
You're spouting misinformation.
The network understood to be strongest at the time is packaged with each release: https://github.com/LeelaChessZero/lc0/r ... ag/v0.28.0. The networks are not "all over the place". It's completely reasonable to expect an engine to be tested with the network literally included in the release.
I'm afraid you're misinformed. The net that is packaged is a smaller demo net. It is not even the official strongest net. Much less the strongest net.
No and no and no. No one knows what the strongest net is that is not the point, but LC0 is delivered with a recent 'strong' net and you have actively discard it for instead testing it with a net from 2020-01-01. What is your beef in this BTW for returning again after the FF disgrace?
Is it the same motivation as for dkappe?
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy