Fritz 17

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
dkappe
Posts: 273
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Fritz 17

Post by dkappe » Thu Nov 07, 2019 11:41 pm

AndrewGrant wrote:
Thu Nov 07, 2019 10:41 pm
Daniel Shawul wrote:
Thu Nov 07, 2019 4:42 pm

For years, you have been taunting Scorpio crashing etc when it none of your business, and now that it doesn't crash at all but
it actually crashes Ethereal you are worried. I thought this would be a big lesson for you that when you are trying new stuff
,instead of tuning to death, you are bound to introduce some bugs and it is OK...

Yes I do like to give it to you when Scorpio beats Ethereal given the extra length you go to taunt me (you still do), here is another one for you:

Code: Select all

Scorpio 3.0.4 MCTS+NN Maddex   - Ethereal 11.53 x64 1CPU          35.0 - 15.0    +22/=26/-2    70.00%
I think the "Leela Wave" has given you an inflated sense of self worth. If you knew anything about Ethereal (And you don't) you would understand the difference between the sort of patches that make it into Ethereal and those that make it into Stockfish.
Ah, that's the talkchess I know and love: engine developers insulting each other energetically. Good times.

Still, Andy, that match result had to sting. The wave has swamped your boat.

Daniel Shawul
Posts: 3762
Joined: Tue Mar 14, 2006 10:34 am
Location: Ethiopia
Contact:

Re: Fritz 17

Post by Daniel Shawul » Fri Nov 08, 2019 12:05 am

Leela wave
I was doing MCTS after A0 news before Leela was born so you got it backwards. Sure lc0 did use NN first but i was mostly focused on understanding how MCTS can cope with AB -- for which you don't need NN. Then came Komodo MCTS after I am done with all my investigation, that then came leela i think. We are all riding the A0 wave but have it your way.

You were constantly taunting me with the crashes at the TCEC while I was experimenting with a bunch of mcts algorithms you do not even bother to try. You hate anyone who dares to attempt things not aligned with your idea of how chess should be programmed, i.e. the "stockfish way".

Fact is I take crashes as a good sign that I am trying new stuff, but you the prodigy thinks Stockfish is the way. You do LMR with big tables like stockfish, you do parallel search like stockfish Lazy SMP, you do futility like stockfish ... etc While that is perfectly fine, it is so damn boring to me, accept it is not everyone's cup of tea. You are new to this chess programming, so i am sure you measure your skill's with your engine's elo, and now that Scorpio scored better it rocked your world. How dare him, I am a better programmer ... This prodigy interview thing must have gotten to your head! Chess programming is not about programming skills alone... and is infact was first meant for a testbed of AI, so once in a while read a paper and try to implement it in your engine.
I find it sad that your #1 criticism is this inane phrase "Tuned to death". Firstly, you don't have a clue what goes on in Ethereal. Secondly, I hardly ever tune and in fact I've never tuned the search, only the eval. Thirdly, in my estimation training a NN is the literal definition of "Tuned to death". Its one thing to be a fool, another to be a hypocrite knowingly.
Your way of doing stuff taking well known algorithms that are proven to work well with Stockfish or otherwise (see above for example) and then
running 30000 games to see if it is a +2 elo patch is one big tuning project. You like doing that, that is fine with me, but you criticize how I do things I critize you back. People who do programming and focus too much on code quality/style, while makeing little intellectual contributions, are sometimeconsidered "useful idiots" -- I am sure DeepMind also have lots of those.
I could not tell you the last time I checked out the SF framework. It is true that OpenBench is made in the vision of Stockfish's fishtest. In fact, it says that RIGHT in the repository. But you made it clear that you don't actually know much about Ethereal so its no surprise you did not know that. In my experience there has only been 1 (ONE) patch that wins for both Stockfish and Ethereal. Its a patch about ProbCut. That I wrote. That I commited to both engines.
Chill. I only meant to demonstrate your tuning to death approach to the point that you wrote a testing framework all by yourself.
You have everything Stockfish have -- though not copy pasted directly -- which you can brag about however you like.
I have also a self-play training pipeline for training neural networks -- that is also means to an end.
If you ever hop down from your high horse and try to understand something instead of parroting the same shallow critisms of others, you might begin to understand things more clearly. The fact that you are naive enough to think engines can just share patches verbatim proves you don't have a damn clue how things work
Where did I say you copied patches?? Ofcourse I image you re-write the idea your way and that is fine with me, tehn test each line with 30000 games...
That is a big achievement for you, and I am not judging, but it is boring process for for me. I don't care about Ethereal nor do follow how how you do your patches, but if you have contributed some key algorithmic improvements, I am all ears.

You run your mouth like below, expect me to point out what kind of "useful idiot" you are solely focused on programming (typical of young college graduate) :
other explanation do you have for a mid-tier engine with an unreadable code base and a rampant history of crashes, bugs, and other play limiting or preventing issues, suddenly shooting up almost overnight to become one of the most powerful engines featured in TCEC / CCC. I won't make you guess the explanation -- the explanation is that anyone can piggy back off the work of the Leela team and with minimal effort build an engine that is a couple hundred elo weaker than Leela.
The bold text tell you, the amount of arrogance in you!!

Did I misread that Ethereal went all TCEC divisions (div-4 to premier) and you were called a prodigy. This is the problem right there!!
Your "inflated ego" after that, and anybody who dares to compete with you must be attacked. I am sure you accused Xiphos/Laser etc at one point.

Dann Corbit
Posts: 10205
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Fritz 17

Post by Dann Corbit » Fri Nov 08, 2019 12:12 am

Time to take a breath here.
Andy and Daniel are both innovators.

I think Andy may not be aware of all the interesting things Daniel has done with gaming.

He did a really early and interesting distributed version of his program using MPI. I played with it and it worked.
He applied his gaming framework to other games, besides chess.

Sometimes, chess programmers have more things to do than just work on chess engines. Daniel's sweetie is a knockout, so I find it surprising that he has any time for chess at all! :lol:

Andy has also done some very interesting work with his engine and his framework. Since his engine is C, it more closely resembles Cfish than Stockfish. (Yes, of course I know that CFish is a port of Stockfish to C, but even that is a very big effort). Ed and Chris have been doing a similarity analysis study and Andy's program does not show any red flags for being a close copy of other programs.

I don't think our energies are being directed as well as possible in this thread. It seems that we are losing focus on the nature of the thread itself.

I don't think anyone has serious concerns about cloning or the posts would be in the other forum.

Daniel is a very good chess programmer. Andy is a very good chess programmer. Both programmers wrote working chess engines that are far stronger than average.

So {as a suggestion} let's try not to get offended and let's try not to offend.
IMO-YMMV
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

User avatar
Graham Banks
Posts: 33247
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: Fritz 17

Post by Graham Banks » Fri Nov 08, 2019 1:06 am

Dann Corbit wrote:
Fri Nov 08, 2019 12:12 am
Time to take a breath here.
Andy and Daniel are both innovators.

I think Andy may not be aware of all the interesting things Daniel has done with gaming.

He did a really early and interesting distributed version of his program using MPI. I played with it and it worked.
He applied his gaming framework to other games, besides chess.

Sometimes, chess programmers have more things to do than just work on chess engines. Daniel's sweetie is a knockout, so I find it surprising that he has any time for chess at all! :lol:

Andy has also done some very interesting work with his engine and his framework. Since his engine is C, it more closely resembles Cfish than Stockfish. (Yes, of course I know that CFish is a port of Stockfish to C, but even that is a very big effort). Ed and Chris have been doing a similarity analysis study and Andy's program does not show any red flags for being a close copy of other programs.

I don't think our energies are being directed as well as possible in this thread. It seems that we are losing focus on the nature of the thread itself.

I don't think anyone has serious concerns about cloning or the posts would be in the other forum.

Daniel is a very good chess programmer. Andy is a very good chess programmer. Both programmers wrote working chess engines that are far stronger than average.

So {as a suggestion} let's try not to get offended and let's try not to offend.
IMO-YMMV
Agreed. Sad to see two of my computer chess 'friends' fighting. :cry:
My email addresses:
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz

User avatar
Ovyron
Posts: 2830
Joined: Tue Jul 03, 2007 2:30 am

Re: Fritz 17

Post by Ovyron » Fri Nov 08, 2019 1:13 am

Dann Corbit wrote:
Thu Nov 07, 2019 11:25 pm
I don't really understand what you are saying. You may notice by examination that each page in the preview says, "Copyrighted material".
So if Fat Fritz's weights were printed in a book and it said "Copyrighted material" then it wouldn't be fine to distribute them?

The claim is that if distributing a set of random numbers is piracy, so is distributing the weight file of Fat Fritz.

Dann Corbit
Posts: 10205
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Fritz 17

Post by Dann Corbit » Fri Nov 08, 2019 1:19 am

Ovyron wrote:
Fri Nov 08, 2019 1:13 am
Dann Corbit wrote:
Thu Nov 07, 2019 11:25 pm
I don't really understand what you are saying. You may notice by examination that each page in the preview says, "Copyrighted material".
So if Fat Fritz's weights were printed in a book and it said "Copyrighted material" then it wouldn't be fine to distribute them?
I did not claim that.
A copyright notice is not necessary. And a video, photo, song, whatever can be copyrighted. Any creative work can be copyrighted. And it is as simple as that.

The claim is that if distributing a set of random numbers is piracy, so is distributing the weight file of Fat Fritz.
I did not claim that distributing a set of random numbers is piracy. But it could be. It depends on a lot of things.
I expect that laws also differ from nation to nation.
I think that intellectual property rights is one of the least understood things in this forum.

Even with copyrighted information, you can make fair use of it.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

AndrewGrant
Posts: 494
Joined: Tue Apr 19, 2016 4:08 am
Location: U.S.A
Full name: Andrew Grant
Contact:

Re: Fritz 17

Post by AndrewGrant » Fri Nov 08, 2019 3:27 am

Daniel Shawul wrote:
Fri Nov 08, 2019 12:05 am
You were constantly taunting me with the crashes at the TCEC
You have a bizarre understanding of the word "taunting". Pointing out that your engine crashes is not taunting. Pointing out that your engine continued to crash for multiple seasons is not taunting. Pointing out that its shameful that you told TCEC you had fixed the issues, only to crash out again, is not taunting.
Daniel Shawul wrote:
Fri Nov 08, 2019 12:05 am
Fact is I take crashes as a good sign that I am trying new stuff, but you the prodigy thinks Stockfish is the way. You do LMR with big tables like stockfish, you do parallel search like stockfish Lazy SMP, you do futility like stockfish
I don't know why you keep throwing around the word prodigy, but its not something I think. I also don't like the way Stockfish does things. That is why Ethereal exists, and why I'm not continuing to add to Stockfish. Are you daft enough to think I don't have a nuanced view of Stockfish? That I treat Stockfish as the word of God? Weird, and wrong.
Daniel Shawul wrote:
Fri Nov 08, 2019 12:05 am
Your way of doing stuff taking well known algorithms that are proven to work well with Stockfish or otherwise (see above for example) and then
running 30000 games to see if it is a +2 elo patch is one big tuning project.
Again you show the world just how clueless you are about the dev process of Ethereal. And you still refuse to properly refute the idiotic hypocrisy of the "tuned to death" expression. Ill restate since apparently it went over you the last time: NN is tuning to death.
Daniel Shawul wrote:
Fri Nov 08, 2019 12:05 am
I don't care about Ethereal nor do follow how how you do your patches, but if you have contributed some key algorithmic improvements, I am all ears.
Okay, so an admission that you don't have a damn clue what you are talking about. Ethereal has the best non NN (ACTUAL) Tuning process of all elite engines. That's an algorithmic improvement that comes with a set of criteria for the sorts of patches Ethereal accepts. Where as the SF criteria is simply elo.
Daniel Shawul wrote:
Fri Nov 08, 2019 12:05 am
You run your mouth like below, expect me to point out what kind of "useful idiot" you are solely focused on programming (typical of young college graduate)
And you don't know me. I do not have a college degree, nor am I a college student at this point.
Daniel Shawul wrote:
Fri Nov 08, 2019 12:05 am
The bold text tell you, the amount of arrogance in you!!
Did I misread that Ethereal went all TCEC divisions (div-4 to premier) and you were called a prodigy. This is the problem right there!!
Your "inflated ego" after that, and anybody who dares to compete with you must be attacked. I am sure you accused Xiphos/Laser etc at one point.
Yeah, I totally hate Xiphos and Laser. Thats why Laser was the first engine I put onto OpenBench after my own. Thats why I fixed an SMP bug in Laser. Thats why I commited some speedups to Laser. Thats why I helped Laldon learn to program patches for Laser. Thats why Jeffrey and I have talked at length. The only reason all of the previous does not apply to Xiphos is because I'm just not as familiar with the author. Milos and I have only chatted briefly, which I believe was when Xiphos made its debut here on Talkchess. I'de happily push Xiphos onto OpenBench. And I would contribute code if I though Milo was interested and if I cared to familiarize myself with his code base and his methodology.

I'm going to bold all of this since it seems like you are one of few who don't understand what I think Ethereal means, and what my role is in the chess world. I think of Ethereal as a more palatable and workable Stockfish. Ethereal's main purpose is to serve as a learning tool for other aspiring engine authors. That is why the source code is verbose. That is why I respond and work with other authors in private. OpenBench is an extension of this. I found fishtest to be too heavy for my needs. So I wrote a few of my own. At first they were nonSPRT frameworks, but eventually I settled on SPRT. I decided that I could generalize what I wrote and share it with others. Now a number of engines are using OpenBench in public or in private to support their development. I care far less about Ethereal than you think. I care more about supporting chess authors than you think.

Ask yourself this: Is it possible I hold my views about Scorpio without holding any sort of personal grievance with you? The answer is yes, but I think if I ask you that question about me and Ethereal, your answer is no.

dkappe
Posts: 273
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Fritz 17

Post by dkappe » Fri Nov 08, 2019 4:50 am

Let it go, Daniel. He’s not worth it. Andy’s personality defect is punishment enough. Let him grouse with his fellow misanthropes.

AndrewGrant
Posts: 494
Joined: Tue Apr 19, 2016 4:08 am
Location: U.S.A
Full name: Andrew Grant
Contact:

Re: Fritz 17

Post by AndrewGrant » Fri Nov 08, 2019 4:53 am

dkappe wrote:
Fri Nov 08, 2019 4:50 am
Let it go, Daniel. He’s not worth it. Andy’s personality defect is punishment enough. Let him grouse with his fellow misanthropes.
This is the kind of banter I like to read. Not the drivel that Daniel writes. Well put Dkappe.

smatovic
Posts: 949
Joined: Wed Mar 10, 2010 9:18 pm
Location: Hamburg, Germany
Full name: Srdja Matovic
Contact:

Re: Fritz 17

Post by smatovic » Fri Nov 08, 2019 9:22 am

Ovyron wrote:
Fri Nov 08, 2019 1:13 am
Dann Corbit wrote:
Thu Nov 07, 2019 11:25 pm
I don't really understand what you are saying. You may notice by examination that each page in the preview says, "Copyrighted material".
So if Fat Fritz's weights were printed in a book and it said "Copyrighted material" then it wouldn't be fine to distribute them?

The claim is that if distributing a set of random numbers is piracy, so is distributing the weight file of Fat Fritz.
The Fat Fritz NN files are not just random numbers, they were created by an creative human act, by selecting games.
I can see the point that argues that it should be covered by copyright.

Nevertheless I guess that science will get into 'Devil's kitchen' if you can copyright or patent a set of numbers.

Where does the copyright start? 1,2,3,4,5? Is an XOR MLP copyrightable? Do I need millions of weights for copyright? What about self-play NNs without human interaction?

Time will tell I guess....

--
Srdja

Post Reply