Leela data worth +20 elo

[Vernon Crawford]

Hi Vernon,

I'm only posting here to confirm that although Andrew, Connor and some other engine authors don't approve of certain engines in the CCRL lists, they've not tried to directly pressure us into removing them.

In fact, I'm on good terms with them, just as I am with you.

Graham.

Graham, what does being on "good terms with them, just as I am with you", have to do with anything?

I realize that you're apt to have the back of old-timers, like Guenther for ex, who appears to do nothing other than pretend to be programmer, attack individuals whose posts he doesn't like, and maintain a spreadsheet, along with a huge database of people he doesn't agree with, his so-called talkchess 'trolls', in an effort to get recognition. This actually makes sense due to your years-long friendship with him.

But are you refuting my assertation that CCRL is influenced by pressure from extremists with clear agendas, like Andrew, Connor, Ed, etc. ?

[dkappe]

OK, I’ve let this attempted thread hijacking go on long enough. I’ve asked the moderators to quash Vernon. If everyone can stop feeding the troll, that would be appreciated.

[Graham Banks]

Hi Vernon,

I'm only posting here to confirm that although Andrew, Connor and some other engine authors don't approve of certain engines in the CCRL lists, they've not tried to directly pressure us into removing them.

In fact, I'm on good terms with them, just as I am with you.

Graham.

Graham, what does being on "good terms with them, just as I am with you", have to do with anything?

I realize that you're apt to have the back of old-timers, like Guenther for ex, who appears to do nothing other than pretend to be programmer, attack individuals whose posts he doesn't like, and maintain a spreadsheet, along with a database of his so-called talkchess 'trolls'. This actually makes sense due to your years-long friendship with him.

But are you refuting my assertation that CCRL is influenced by pressure from extremists with clear agendas, like Andrew, Connor, Ed, etc. ?

I can't really see how we're being influenced, as we test what we choose to test regardless.

[Vernon Crawford]

OK, I’ve let this attempted thread hijacking go on long enough. I’ve asked the moderators to quash Vernon. If everyone can stop feeding the troll, that would be appreciated.

Expressions of opposition to the establishment, against the status quo, should never be 'quashed' ...at least not in any free society.
Sounds totalitarian...have you seen the Russian domestic situation in the news recently?

But I get your point, according to you opposing/dissenting viewpoints are not welcome and may not be permitted here.

Quite frankly, I think Graham is probably a mature adult and can (and should be allowed to) answer the question I directed to him by himself.
Why you need to chime in is beyond me. (wait...you're running for moderator and need to appear to be in control?)

[dkappe]

Expressions of opposition to the establishment, against the status quo, should never be 'quashed' ...at least not in any free society.
Sounds totalitarian...have you seen the Russian domestic situation in the news recently?

But I get your point, according to you opposing/dissenting viewpoints are not welcome and may not be permitted here.

Quite frankly, I think Graham is probably a mature adult and can (and should be allowed to) answer the question I directed to him by himself.
Why you need to chime in is beyond me.

This is a thread about the use of leela data in training nnue nets. Your brand of mental diarrhea is irrelevant noise. You are free to express yourself, but don’t do it in this thread.

[kranium]

This policy is not officially stated anywhere, though it does seem to be the policy CCRL is employing (as a consequence of testers not being interested in testing such engines). Hopefully that clarifies what I meant.

I have received emails from Graham explicitly stating that they only run engines whose nets have been trained by the engine itself.
That is why they don't test Mayhem for example.

Is that 'official' enough?

Vernon has a good point, why is it ok then to test Stockfish and Leela who share data?
It's a double standard.

But we've already had the explanation:

I can't really see how we're being influenced, as we test what we choose to test regardless.

ie no policy at all, we just 'wing' it, and we're entitled to do that
I'm OK with that as long as it's clear that's it's misguided and completely arbitrary

[AndrewGrant]

they only run engines whose nets have been trained by the engine itself.

The Engine itself, or the engine author themselves, or the network is trained FOR the engine itself?

Three variations there:
1. Network was trained on the engine's own data (SF/Leela mixing runs afoul, Fire runs afoul, Ethereal is not afoul)
2. Engine authors are the network authors (SF/Leela mixing does not run afoul, Fire runs afoul, Ethereal is not afoul)
3. The network was trained specifically for the engine (SF/Leela mixing does not run afoul, Fire runs afoul, Ethereal is not afoul)

(Engines are for example). I think most engines fall into [2]+[3] being true since they are solo ventures. [1] is true for "most"

[Graham Banks]

Must admit I'd prefer if networks were trained on self-play games from the engine itself.

I can see why some aren't all that comfortable with the SF/Leela collaboration in using each other's games.

[Vernon Crawford]

they only run engines whose nets have been trained by the engine itself.

The Engine itself, or the engine author themselves, or the network is trained FOR the engine itself?

Three variations there:
1. Network was trained on the engine's own data (SF/Leela mixing runs afoul, Fire runs afoul, Ethereal is not afoul)
2. Engine authors are the network authors (SF/Leela mixing does not run afoul, Fire runs afoul, Ethereal is not afoul)
3. The network was trained specifically for the engine (SF/Leela mixing does not run afoul, Fire runs afoul, Ethereal is not afoul)

(Engines are for example). I think most engines fall into [2]+[3] being true since they are solo ventures. [1] is true for "most"

Yeah yeah, puke... I think we've all had enough of you single-minded crusade against Fire. Please give us a a break.

[AndrewGrant]

they only run engines whose nets have been trained by the engine itself.

The Engine itself, or the engine author themselves, or the network is trained FOR the engine itself?

Three variations there:
1. Network was trained on the engine's own data (SF/Leela mixing runs afoul, Fire runs afoul, Ethereal is not afoul)
2. Engine authors are the network authors (SF/Leela mixing does not run afoul, Fire runs afoul, Ethereal is not afoul)
3. The network was trained specifically for the engine (SF/Leela mixing does not run afoul, Fire runs afoul, Ethereal is not afoul)

(Engines are for example). I think most engines fall into [2]+[3] being true since they are solo ventures. [1] is true for "most"

Yeah yeah, puke... I think we've all had enough of you single-minded crusade against Fire. Please give us a a break.

Truth hurts, Norm---Vernon Crawford.