FCP Tourney-KI started January 02nd, 2022

[kranium]

Hi Norman,

what should I do?

You are the programmer of Fire ...

Rename to Fire 8 MC.3 NNSV
"Fire includes 'Raptor', a top reinforcement learning network trained by Sergio Vieri"

But if I do that I have to rename:
Dragon x NN (Komodo) to Dragon x NNDK (Komodo).
After all I understand it comes from Dietrich Kappe (maybe I am wrong here).

And the long day I have to rename engine names again and again and all the statistic *.txt files from the tourney.
All the round-by-round files ... and many others need for statistics.

For changed Stash 32 to Stash 32.0 in my databases and *.txt files (I found the mistake last week) I need around 5 hours.

Best
Frank

I don't care Frank, but while you're at it...
you could also name SF NNUE: Stockfish NN-LC0 and rename LC0: LC0 NN-SF
because they believe in the 'open & shared data' philosophy
and share data liberally for creation of their NNs

https://lczero.org/blog/2021/06/the-imp ... open-data/

I do know that there's nothing nefarious going on with Fire's README.md, as suggested by Guenther
All Readme.md changes have been for fixing typos, crediting others as much as humanly possible, updating features, version nos., etc.
It's public, anyone can read it themselves, follow his link

The suggestion or innuendo otherwise is simply hateful and mean-spirited.

[Frank Quisinsky]

Hi Norman,

at the moment I have no interest to change that.
A lot of private work in the next days and I have at the moment no ideas to find a solution for more silence. Must thinking about the name of engines again. Perhaps the idea with NN"Sf" or NN"Re" was not a good solution.

BTW:
Two of my three "tourney systems" lost interest, go sleeping, and do three hours nothing ... after the TalkChess power.
Joking, two of my systems lost the network connection for unkown reasons.

Now all is working fine and I restarted!

Enough for today ... to late ... 02:17.

Best ...
and for the readers of my tourney news?!

Engines works fine.
No engine in "run-6" produced a crash after more as 5.000 new 80-min. games.

From time to time my Wlan network have problems, in the last half year now 3times.
A network cable to the other room can solve my problem.

Frank

PS: I am thinking often that programmers like to do the following.
Breaking the causal chain on a position they like and add GPL rules.

A chess friend, know him from Arena times (Lawyer in New-York, a phone call ... 1 minute with him = 2.700 Euro) gave me the information for some months after I asked him different things I need for a professional work I do. GPL is't free of problems!

[Modern Times]

I do not know who imposed the idea that each engine should use its own evaluations to create the network and could not use evaluations from other engines, this was imposed as a must. Who decided it and why? If someone wants to organize a tournament under those circumstances I am fine with it. But I don't feel the need for someone to tell me how to make my engine.

Well said. It is a small group of people who are bullying testers and enthusiasts by telling them what they can and can't test, how to name the engines on their lists etc. That bullying has to stop, otherwise there will be no testers left at all.

[Graham Banks]

..........Stockfish decided that they are going to use the evaluation data generated by Leela. If I had been the one to decide that my engine would not have been tested. If I am being questioned for using nodchip, why isn't there a bigger protest from other programmers with Stockfish? Of course I have no problem with Stockfish's decision.............

I've wondered this myself.
If they're happy to use each other's data, why is it so wrong for others?

Having said that, I like it when an engine author uses his/her own self-play games to generate a network, as to me, that maintains the uniqueness of their engine.

[Modern Times]

..........Stockfish decided that they are going to use the evaluation data generated by Leela. If I had been the one to decide that my engine would not have been tested. If I am being questioned for using nodchip, why isn't there a bigger protest from other programmers with Stockfish? Of course I have no problem with Stockfish's decision.............

I've wondered this myself.
If they're happy to use each other's data, why is it so wrong for others?

Indeed. Is it one rule for Stockfish and Lc0, and another rule for everyone else ?

[Frank Quisinsky]

Hi there,

yes, Pedro made the point in his posting.
One of the very cool programmers.

After "run-8" I stopped my tourney and delete all information about the tourney from my site.
In around 6 weeks I think.

So, if people like to download the games, more as 100.000 80-min games ... they should do that in the next 6 weeks.

I must not create an official new FEOBOS or an official tournament.
Can do that for myself and people where like the work.

With the final results:
More silence for ... restless / troubled ghosts.

Will be my contribution for the group of programmers, like it to write dirty messages about my person ... if the day is long, and no other things are to do. But absolutely clear is ... I will have a lot of fun with computer chess in the next years, no doubt about it.

I wrote that last year, the most problems we build in computer chess comes all the years from the group of programmers, not from harmless users of chess programs.

Best
Frank

[Frank Quisinsky]

Hi there,

back to my tourney:

Code: Select all

After round 13 out of 30 ...

FCP Tourney-KI
run-6 after round 13 out of 30

         Name                       Games     Won     Draw     Lose       Pts         S-B         %

01. Stockfish 020622 NN dev       :   520  :  312+  :  208=  :    0-  :   416.0  : 103541.25  :  80.00%
02. Dragon 3 NN (Komodo)          :   520  :  294+  :  226=  :    0-  :   407.0  : 101290.50  :  78.27%
03. Koivisto 8.9 NN               :   520  :  223+  :  283=  :   14-  :   364.5  :  88699.75  :  70.10%
04. Fire 8 MC.3 NN                :   520  :  212+  :  301=  :    7-  :   362.5  :  89287.25  :  69.71%
05. SlowChess Blitz 2.9 NN        :   520  :  215+  :  291=  :   14-  :   360.5  :  88048.25  :  69.33%
06. Berserk 9 NN dev3             :   520  :  215+  :  289=  :   16-  :   359.5  :  87669.75  :  69.13%
07. Revenge 2.0 NN                :   520  :  200+  :  297=  :   23-  :   348.5  :  84555.25  :  67.02%
08. rofChade 3.0 NN               :   520  :  200+  :  288=  :   32-  :   344.0  :  83721.50  :  66.15%
09. RubiChess 20220223 NN         :   520  :  182+  :  309=  :   29-  :   336.5  :  81421.00  :  64.71%
10. Seer 2.5.0 NN                 :   520  :  177+  :  297=  :   46-  :   325.5  :  78408.50  :  62.60%
11. Arasan 23.3 NN                :   520  :  135+  :  339=  :   46-  :   304.5  :  73741.00  :  58.56%
12. Minic 3.22 NN                 :   520  :  140+  :  325=  :   55-  :   302.5  :  73153.50  :  58.17%
13. Igel 3.1.0 NN                 :   520  :  106+  :  367=  :   47-  :   289.5  :  70881.75  :  55.67%
14. Halogen 10.23 NN dev          :   520  :  126+  :  308=  :   86-  :   280.0  :  66493.75  :  53.85%
15. Ethereal 13.07                :   520  :  109+  :  333=  :   78-  :   275.5  :  66206.75  :  52.98%
16. Lc0 0.28.2 752187 CPU         :   520  :  110+  :  325=  :   85-  :   272.5  :  66301.00  :  52.40%
17. Clover 3.1 NN                 :   520  :  112+  :  321=  :   87-  :   272.5  :  64951.50  :  52.40%
18. Nemorino 6.09 NN dev          :   520  :  110+  :  322=  :   88-  :   271.0  :  64971.25  :  52.12%
19. Wasp 5.54 NN dev              :   520  :  111+  :  317=  :   92-  :   269.5  :  64766.00  :  51.83%
20. Booot 7.0 NN dev              :   520  :   93+  :  342=  :   85-  :   264.0  :  63388.75  :  50.77%
21. Tucano 10.00 NN               :   520  :   84+  :  342=  :   94-  :   255.0  :  60715.00  :  49.04%
22. Rebel 15x2 NN                 :   520  :   91+  :  318=  :  111-  :   250.0  :  59825.25  :  48.08%
23. Uralochka 3.36c NN            :   520  :   89+  :  320=  :  111-  :   249.0  :  59716.50  :  47.88%
24. Fritz 18 NN (Ginkgo)          :   520  :   85+  :  307=  :  128-  :   238.5  :  56071.25  :  45.87%
25. Velvet 3.3.0 NN               :   520  :   73+  :  298=  :  149-  :   222.0  :  52089.25  :  42.69%
26. Zahak 10.0 NN                 :   520  :   72+  :  295=  :  153-  :   219.5  :  51545.50  :  42.21%
27. Xiphos 0.6                    :   520  :   68+  :  284=  :  168-  :   210.0  :  49665.00  :  40.38%
28. Combusken 2.0.0               :   520  :   61+  :  295=  :  164-  :   208.5  :  49416.00  :  40.10%
29. Weiss 2.0                     :   520  :   61+  :  287=  :  172-  :   204.5  :  48952.25  :  39.33%
30. Schooner 2.2 XB               :   520  :   45+  :  302=  :  173-  :   196.0  :  46248.75  :  37.69%
31. Defenchess 2.3 dev            :   520  :   41+  :  305=  :  174-  :   193.5  :  46382.25  :  37.21%
32. Bit-Genie 9.19 dev            :   520  :   62+  :  255=  :  203-  :   189.5  :  44763.00  :  36.44%
33. Shredder 13                   :   520  :   53+  :  270=  :  197-  :   188.0  :  44140.50  :  36.15%
34. Laser 1.7                     :   520  :   50+  :  275=  :  195-  :   187.5  :  44095.50  :  36.06%
35. Hiarcs 15.1                   :   520  :   45+  :  283=  :  192-  :   186.5  :  44531.50  :  35.87%
36. Fizbo 2.0                     :   520  :   54+  :  254=  :  212-  :   181.0  :  42975.00  :  34.81%
37. Stash 33.0                    :   520  :   46+  :  270=  :  204-  :   181.0  :  42911.50  :  34.81%
38. DanaSah 9.0 NN                :   520  :   32+  :  297=  :  191-  :   180.5  :  43587.75  :  34.71%
39. Marvin 5.2.0 NN               :   520  :   40+  :  266=  :  214-  :   173.0  :  40652.50  :  33.27%
40. Chiron 5                      :   520  :   44+  :  251=  :  225-  :   169.5  :  39186.25  :  32.60%
41. Expositor 2WN29 NN            :   520  :   25+  :  252=  :  243-  :   151.0  :  36593.25  :  29.04%

And the current ratings ...
as *.jpg file ...
https://www.amateurschach.de/t-ki/common/ki-ratings.jpg

or as *.txt file ...
https://www.amateurschach.de/t-ki/ratin ... atings.txt

Best
Frank

[connor_mcmonigle]

I'm really rather clueless as to why Danasah was brought up in this discussion. The process Pedro used to generate networks for Danasah is somewhat novel and interesting, involving bootstrapping the RL training process from CCRL game results. This is completely different from Fire as Fire uses an old SF network taken directly from FishTest which was trained on SF evaluations.

Frank, my understanding was that your "NNSf" label is intended to denote engines which use networks trained by Stockfish contributors for Stockfish. Note that this amounts to copy pasting Stockfish's evaluation function into one's own engine. This is also not remotely the same as Stockfish contributors using Lc0 data in training networks for Stockfish using training tools, inference code and architectures which were all the fruits of their own development efforts. Hopefully this clarifies the situation.

[Frank Quisinsky]

Hi Conner,

At first:
I understand what you wrote here and it must not a secret that I like your comments in TalkChess.
Read it and try to understand. I think you should know that at first.
I have no interest to go with the next strong programmer in confrontation.

To explain my view in German language is much more easy as in "Frankish".

OK ...

1998 or 1999 (Dutch-op).
A programmer from The Netherlands bought the sources of the program "Nightmare". Note: Two programs with the name Nightmare played different computer chess tourneys before. A program from Germany and a program from The Netherland ... after all I know ... completely different programs.

OK ...

After all I understand the programmer bought the program for one reason only. He search a way for experiments with Neural Network ideas and wrote a book about it. This book he gave to the programmers participate in Leiden. I am with Roland Pfister (Patzer) in Leiden and gave my book to the friend of Roland (Dr. Ulrich Türke, chess program Comet). Different of the programmers "laughing behind hidden hand" about the fantastic ideas around it. Walter Ravenck (very talented programmer, chess program Arthur - Mac). Walter was a specialist for knowledge about knight maneuver in Mid-games for a stronger positional style ... he like Karpov a lot.

OK ...

Walter told me about the situation with the book about Neural Network ideas that it's not uninteresting but it need 20 years to thinking about it in detail. With current programming ideas and hardware possibilities he can not see a directly innovation. With the comment ... to many games for 5 Elo more ... no reason to think about it.

Today different Go people are thinking that the idea comes from Go!
Unfortunately, a have not read the book.
Had enough work with special opening lines for Patzer during the tourney and gave the topic also a small view. Furthermore, for me not very interesting as non programmer.

And exactly this one is the situation I dislike.
An idea goes in a program like Stockfish many people are working on it.
A person from a big team created an neural network file to an idea comes not from Stockfish development.

I have a very big problem with GPL3.
Programmers should keep interesting ideas in the own hand.

For that reason ...
I like ... SlowChess, Revenge, Wasp, rofChade and others ... sources are not available.
OK, all of these programs have really a great style but for exactly that reason I like the program a lot more.

OK ...

After all my experience in so many years computer chess with programmers ...
Very often I got years later the comment: It was a mistake to give my sources free!

---

Nobody speak about the great years of developments in computer chess.
So many great ideas, from persons are working a long time on it, are today available in x chess programs.

If we today have the opinion:
The NN file the Fire programmer used is a part of Stockfish ...
We can say ... the ideas the Stockfish programmer used is a part from x others, often a long time before Stockfish are available.

Read that interview:
https://www.schach-welt.de/schach/compu ... stalba-eng

Interesting is Marco Costalba ...

Believe me Conner:
I have a lot of contacts to programmers (from bigger companies, also 100% profis) and the opinion is a complete other. Opinion about GPL 3 … you will not know that.

For that reason, I am thinking often …
All the power by younger programmers goes in Nirvana. Years later opinion will be changed.

---

OK back to Stockfish:
I am thinking that Stockfish is the best project ever in computer chess.
If in foreground the thoughts ... to help with the work.
Again, 1000 programmers before Stockfish are working in chess programming’s.
It can't be the goal ... to judge others if they using ideas ... but to collect all the available knowledge and ideas in big groups.

In my bad English ...
Need a coffee.
I am not satisfied with this message.
Much more easy for me to explain my view in German language.

A good chance now to go again in confrontation with my person.
Different people like it to use each chance!

Best
Frank

[Frank Quisinsky]

In 20 years ...

nobody will speak about Conner, Jay the two young Koivisto programmers and all the others.
An other group of persons are working on chess programs.

And this group of programmers will give the information:
Like the Houdini programmer after release 1.0 ... All is mine with the final results ...

The commercials used the chance directly!

And this isn't in my interest after all the bad situation I had in Winboard, in Arena times with commercials.

I am very sadly after a mail from Fabien LETOUZEY after I read in detail what he do some years after Fruit and was other do with all the ideas from him. A situation in my brain I have often if I wrote here messages or if I read that with 100% power younger programmers are working, for the moment very proud of his own work and perhaps in the future very disappointed what others do with it.

That's is the point:
To accept in computer chess is a big topic.
Younger people must learn that and the older people have to learn that social media, chess server and all this are the new way.
The new way is often the wrong way.

Shit, thinking on my father and his lectures.
Years later ... my father is death ... I wish me a lectures from him.

Now I am more satisfied with my message!
But I am sure, the most here will not understand it.

Enough ...
For the moment I will changed nothing with name of Fire neural network.
For the moment I have more interest in looking Wasp games.
You should do that also ... Wasp won two nice games vs. Seer, OK lost vs. Seer other games.
Seer is playing a great tournament here!