If you were ...

[Cornfed]

Hello Cornfed,

Nibbler, Banksia, CuteChess ... new wonderful GUIs.

Banksia currently has more bugs than a corn field, but I love the way it looks. When stable, it will be great...for what it is - a 'chess tournament manager', probaly the best! But it offers me nothing as I am not interested in running engine vs engine tourneys.

Nibbler...is kind of 'stand alone' isn't it? I mean, it doesn't interface with a database management program like Chessbase? I downloaded it a while back...liked it, but could not find easy use for it.

CuteChess - really just for engine matches, right?

And see, that's what I am saying - these are all good for what they do...but limited for the competitive (non-'cybord') chess player who simply needs more.

You mention all the 'bells and whistles' with Chessbase and how many are not used. True that... but someone uses them (except perhaps the Ratraced board/pieces waste of time...) and so that is why it is go wildly popular with competitive chess players compared to other programs.

I do wish they would develop certain things - like a feature akin to DecodeChess (that should be useful to many lower rated players) and integrating a 'fritz like' thing into the program so one could run basic engine vs engine monte carlo testing of positions ...maybe letting you run your played games thru a process to spit out in a user friendly where your weaknesses are, etc BUT I've owned CB since version 7 and CB16 is the best iteration of any that I can recall with a number of useful stuff CB15 did not.

[mclane]

I am not interested in todays engines. They all play like stockfish. How boring.

[towforce]

If you were Ubisoft, Ripstone, or Chessbase for example and looking for a chess engine to license for use in your next chess software project to release. Who would you be contacting?

If I were the point man for Ubisoft or Ripstone, I think I would be trying to contact Jonathan Kreuzer (Slowchess). For Chessbase I think I would go with Andrew Grant (Ethereal)

In think Andrew Grant works for chess.com (CC), so licensing his program to Chessbase might not look too well to his employer. I don't know what he does there, but writing the bots and the analysis page would be a reasonable guess.

Apologies if I'm wrong, but I think that CC's analysis engine runs in JavaScript on your own browser, whereas the opponents ("bots") run on CC's server. For my personal chess computer needs, CC is more than good enough: I can play human opponents easily and conveniently, I can play against a bot, and when I look at another game and ask "why didn't they play the move that looks good to me?", I just copy the position fen into CC's analysis page, and it quickly shows me how my "good idea" gets crushed.

[supersharp77]

If you were Ubisoft, Ripstone, or Chessbase for example and looking for a chess engine to license for use in your next chess software project to release. Who would you be contacting?

If I were the point man for Ubisoft or Ripstone, I think I would be trying to contact Jonathan Kreuzer (Slowchess). For Chessbase I think I would go with Andrew Grant (Ethereal)

Well I think Mike B's Harmon Chess Engine would work quite well in any of the commercial Gui's plays a quite interesting game with Human.nnue or with Harmon.nnue files.....'Quite Human Like'.....

https://github.com/MichaelB7/Stockfish/ ... x86-64.exe

The other possibility would be a interesting project like "Crazy Ara" (Classic Ara)

https://github.com/QueensGambit/CrazyAr ... in_MKL.zip
Needs some adjustments for time losses (Fritz Gui) but plays a interesting game indeed....

[JohnW]

If you were Ubisoft, Ripstone, or Chessbase for example and looking for a chess engine to license for use in your next chess software project to release. Who would you be contacting?

If I were the point man for Ubisoft or Ripstone, I think I would be trying to contact Jonathan Kreuzer (Slowchess). For Chessbase I think I would go with Andrew Grant (Ethereal)

Why would they need to? Just create the gui and tell the customer to do a google search for free chess engines.
Personally I like the look and feel of the Fritz GUI the best. I like setting up engine vs engine matches and tournaments. The gui always seems buggy though.
For actual playing I much rather play against older engines like the old Mephisto ROMS. Being able to play them on my DGT or Certabo board is perfect, I almost never play just looking at the computer screen. I have also collected about a dozen+ of older dedicated chess computers, partly for nostalgia but it's just more satisfying to me to beat a computer that I know is playing it's best instead of beating a uci engine that you intentionally dumbed down so that you have a chance. It's like when you were a kid and an adult let you beat them, it was much more satisfying when you learned and improved and the adult could no longer beat you if they wanted.

[Frank Quisinsky]

Hi Thorsten,

I am sure that many programmers today like to used Stockfish ideas. If I am looking in playing-style stats from TOP-41 end of the year 2020 (FCP-Tourney-2021) I can see enough different engines. A hand full of TOP-41 I used for FCP Toruney-2021 only I have stomache ache. Will not give the names of engines.

I am thinking that you miss the last 20-25 years computerchess and you life "a bit" in the past. To like nostalgic chess computers or PC-programs is fully OK for me (I like it too) but you should not fail all the current interesting developments for the reason that Stockfish is freely available with sources. Try to learn how nice Stockfish is.

Again, I like more the closed sources engines. Without all the free and very strong open-source engines some own ideas more by chess programmers are available for us. But fact ist we have all the free sources and we have to life with it.

Best
Frank

[Frank Quisinsky]

Hello Cornfeld,

that's right what you wrote. Same for Arena I am working with Martin.
To added more and more GUI features have a price ... stability of the GUI.
Martin Blume is working more time in optimazation processes as in development of new features.

So, it make sense if current GUI developments have a main-topic!

If I am looking in Fritz GUI in detail I can see a lot of features for mass-market with a lot of bugs, in reality nobody need. That's right what Thorsten wrote here. I must collect again and again the UPD versions of ChessBase and Fritz GUI. So I can put an older version if for a newer version the list of bugs is again higher.

Such easy problems like to add the right ECO code for big databases. First ChessBase 16 versions crashed (one example only). For many GUI-options the programmers have to pay a very high price. I wish me that GUI programmers try to make the GUIs more stable before new features will be added. More stable ChessBase GUIs are much more important as to thinking about FatFritz. A _new commercial engine_ without any GUI changes with a program we all can download very easy.

With FatFritz II the possiblity to selling engines with a GUI ended in my humble opinion.
An own-goal.

I fully agreed with your other words (Monte-Carlo for an example).
I wish me that 3-hirn features will be develop for the future.
Much more great stats to playing-styles of players and engines.

Or that the engines can analyze games more stable without all the bugs and missunderstanding stats the gui produced.

I think we can try to start with simulations the styles of great players with the material we have. 4.000 Kortschnoi games are available for us. What we can learn from Kortschnoi and how we can give an engine the order to play like the style of Kortschnoi.

The GUI development is today in start position.
Engine developers are in waiting position of GUI developers.

I think that in GUI developments we lost some important years.

Best
Frank

[smatovic]

Frank, it seems you welcome now our new AI overlords, when did you flip??

Hi,

programmer of Stockfish is Tord Romstad (sucesscor of Glaurung).
With Marco and Joona and many other helpers.

I nice project!
https://www.schach-welt.de/schach/compu ... stalba-eng

LCO is no name, nothing, lesser as nothing!
It's a part of KI, a nice KI experiment, not more not less.

KI can't be innovative.
A human can be innovative!
That's completly nonsens!

Waste of time to test it.
So I think, every 800 Elo strong chess computer from the beginning of that area is exactly 100% more interesting as to test nothing?!

Sorry, I am a human and not a slave from a computer!
The computers are the slaves from the humans, this is a different!

Best
Frank

Maybe KI will help us in chess programming, that would be great but I can't see any good chess project for it.
Sure it will be come!

[Frank Quisinsky]

Topic Lc0
In FCP Tourney-2021 I can saw ...

Maybe Rybka is the engine with more as 3000 Elo can top Lc0 in boring mid-game chess, blunders and mistakes and quantity of fast lost games. I think with KI we can do much more interesting as we do. Stockfish and a lot of other engines are available can play the gaming phases, people like most, mid-games spectacular.

Example:

Code: Select all

  769 of 41.000 = 1,876% : Fast won/lost games below 50 moves (with mate ended)
  452 of 41.000 = 1,103% : Fast draw games below 30 moves

                                  won49 lost49 draw29        won49w won49b lost49w lost49b
01. Stockfish 110121 NN BMI2 x64    118      0     10            79     39       0       0
02. Dragon by Komodo NN AVX2 x64     87      0     10            54     33       0       0
03. SlowChess 2.5 NN AVX2 x64        45      0     14            30     15       0       0
04. Pedone 3.0 NN BMI2 x64           39      1     22            32      7       1       0
05. Xiphos 0.6 BMI2 x64              29      1     22            21      8       0       1
06. Booot 6.4 POPCNT x64             29      1     33            21      8       0       1
07. Defenchess 2.3 dev BMI2 x64      29      5     29            21      8       2       3
08. Ethereal 12.75 PEXT x64          27      1     22            19      8       0       1
09. Wasp 4.50 Modern x64             25      1     27            21      4       1       0
10. Fizbo 2.0 BMI2 x64               24      1     26            21      3       1       0
11. Andscacs 0.95.123 x64            23      2     25            16      7       0       2
12. Laser 1.7 BMI2 x64               23      4     20            16      7       3       1
13. Schooner 2.2 XB SSE x64          21      2     23            12      9       1       1
14. Igel 2.9.0 NN BMI2 x64           20     14     35            16      5       4      10
15. Seer 1.2.1 NN Skylake x64        20     20     29            14      6       9      11
16. RubiChess 1.9 NN BMI2 x64        18      2     21            10      8       1       1
17. Nemorino 6.04 NN PEXT x64        14     10     21            12      3       2       7
18. Marvin 5.0.0 NN AVX2 x64         12      9     25            10      2       1       8
19. Fritz 17 (Ginkgo) x64            12     10     26            10      2       1       9
20. Arasan 22.2 BMI2 x64             11     11     24             9      2       2       9
21. rofChade 2.3 BMI2 x64            11     11     25             6      5       3       8
22. Critter 1.6a x64                 11     28     21             9      2       8      20
23. Shredder 13 POPCNT x64           10      4     23             8      2       2       2
24. Vajolet2 2.8 BMI2 x64            10      8     17             9      1       2       6
25. GullChess 3.0 Sy BMI2 x64        10     21     20             9      1       9      12
26. Chiron 4 x64                      9     12     16             7      2       4       8
27. iCE 4.0 v853 Modern x64           9     29     24             8      1       8      21
28. Nirvanachess 2.5 POPCNT x64       8     20     27             3      5       5      15
29. Koivisto 4.19 AVX x64             8     25     22             5      3       5      20
30. Orion 0.8 NN FMA x64              8     40     24             8      0      13      27
31. Winter 0.9 NN BMI2 x64            7     35     21             5      2       7      28
32. Combusken 1.4.0 AMD x64           7     38     29             4      3      11      27
33. Halogen 9 NN PEXT x64             7     42     19             5      2      11      31
34. Lc0 0.26.3 NN CPU x64             7     54     11             4      3      11      43
35. Demolito 2020-12-24 PEXT x64      5     25      9             5      0       6      19
36. chess22k 1.14 JAVA x64            4     22     18             3      1       6      16
37. Protector 1.9.0 x64               3     11     32             2      1       3       8
38. Texel 1.08a13 BMI2 x64            3     23     16             3      0       9      14
39. Topple 0.8.0 Modern x64           2     93     19             2      0      29      64
40. Hannibal 1.7 x64                  1     27     21             1      0       7      20
41. pirarucu 3.3.5 JAVA x64           1    117     26             0      1      29      78

Place 34, with 54 fast loses out of 2.000 games.
Sent a grandmaster the lost games and his comment is ...
Frank, this one is Lc0 on CPU?
What for an chaos engine!

The grandmaster asked me for the the fast won games ...
Sure, I sent him the 7 games!

If you simulate now the strength to GPU hardware you will get the same result. The endgame is maybe the strongest from all but with many pieces on board the engine is not very interesting if I compare the results with TOP-20 (last place). To many other nice programs are available. So, waste on time to play with Lc0. But the ideas from Lc0 I think can be interesting for much other things in computer chess.

Best
Frank

[Frank Quisinsky]

My message is from the last year.
But my opinion about Lc0 is all the time the same.

Shortly, a chess friend, unhappy with CPU-results in FCP-Tourney 2021 from Lc0 can't believe (better will not believe) that and used Nvida 3070 GPU and replayed with my configuration the 2000 FCP Tourney-games. Now Lc0 is on place 17 in short game-stats (much better, but again ... 17 others, 16 much weaker in Elo produced more interesting mid-game results). Stockfish won 5.5x more games very fast as Lc0 on 3070er GPU hardware. Move average is in addition very high, only Demolito have an higher move average. Not that what humans like to see.

Endgame stats: Lc0 is on first position and Elo ... Stockfish 110121 is around 35 Elo stronger.

I hope we can use KI in the near feature for simulation from Grandmaster playing-styles much better as we can do today.

If we try to simulate human styles we can see that best players are playing most blunders in for chess programs easy endgames. We can see that strongest players have fantastic ideas in mid-games and a lot of knowledge in opening theory.

Chess engines have to produced this one for a strong simulation of human chess ...

- mid-game, +++++
- late-mid-game, ++++
- earlier endgame, ++ - +++
- endgame, +++

From the view as strongest chess engines today!

LcO produced this one ...

- mid-game, ++ - +++
- late mid-game, +++
- earlier endgame, +++++
- endgame, +++++

That's the problem!
From my view a chess program have to simulate exactly the same strengths best humans have. And for self playing such a chess program is exactly genial.

Wasp do that exactly (for an example).
Lc0 is playing the wirrwarr (view ... strength for game-phases) the older chess computers plays because the older chess computer produced this one ...

- mid game, ++
- late mid-game, ++
- earlier endgame, +
- endgame, nothing or lesser as 1x +

Can you give a club player with 1400 Elo the chesss computer MMV and you can looking what the club player will make with MM5 in endgames. In reality the MMV have around 1800 Elo but the endgame is 1200 Elo. Very bad for humans, same for Lc0 ... the combination from strengths for the playing-phases are very bad for humans.

Both, the older chess computer (most, not all) and software like Lc0 plays chess, it have nothing to do with human chess on highest niveau. And for me as "computer chess freak" the situation is very clear. I am searching for engines, simulate human chess of highest niveau. I will have the Super GM at home and I will not have a computer-brain at home give me results I can nothing do with it.

It is a big order for chess programmers to make engines weaker for products like chess computer and to give engines an interesting style, humans like.

Best
Frank