FCP-Tourney-2024-MA ... game in 66 min. + 6 sec. / move

[Frank Quisinsky]

Hi there,

I am 95% finished with the tournament organisation.
I will only be working on the opening lines for the next 2 days.

At first the list of participanting engines:

Code: Select all

FCP-Tourney-2024-MA
44 of the available ~TOP-75 CPU-Engines at the start!
It is to 95% the move-average that is decisive for the selection of engines.

Please note (all games with resign = off):
Move-average FCP-Rating-List  = 88,39
Move-average FCP-Tourney-2020 = 87,30
Move-average FCP-Tourney-2021 = 87,03
Move-average FCP-Tourney-2022 = 88,98
Move-average KI-Ratings 2023  = 90,37
Move-average FCP-Tourney-2024 = 95,40 (unfortunately, I have to changed different things).

List of participating engines (changes possible):
Fastest available compile is used!
NN = Neural-Network
NNSV = Neural-Network file based on the work by Sergio Vieri (for Stockfish)


    Name of engine                engine date  commentaries
01. Alexandria 5.1.0 NN           2023-11-26
02. Altair 6.0.0 NN               2023-12-05
03. Andscacs 0.95.123             2019-03-22   HCE = Hand Crafted Evaluation
04. Arasan 24.0 NN                2023-08-12
05. BlackCore 6.0 NN              2023-03-02
06. Booot 6.4                     2020-04-05   HCE = Hand Crafted Evaluation, superstar from FCP-Tourney-2020
07. Caissa 1.15 NN                2023-12-14
08. Carp 3.0.1 NN                 2023-09-25
09. CSTal 2.00 NN                 2023-06-29
10. Chess.cpp 4.0 NN              2023-07-07
11. Chiron 5.01                   2021-06-25   HCE = Hand Crafted Evaluation
12. Clover 6.1 NN                 2023-12-12
13. DanaSah 9.1 NN                2023-06-20
14. Devre 4.0 NN                  2022-08-23
15. Dragon 3.3 NN (aggr.)         2023-10-02   Personality: Aggressive
16. Fire 9.2 NN                   2023-11-12
17. Fizbo 2.0 NN                  2017-12-18
18. Fritz 19 NN (Gingko)          2023-09-11
19. Hakkapeliitta 3.0 NNSV        2023-06-08   With "Raptor" Neural-Network by Sergio Vieri (Indonesian)
20. Hiarcs 15.2 (aggr.)           2023-07-01   HCE = Hand Crafted Evaluation, Personality: Aggressive
21. Igel 3.5.0 NN                 2023-06-22
22. Laser 1.7                     2019-02-07   HCE = Hand Crafted Evaluation
23. Mantissa 3.7.2 NN             2022-08-19
24. Marvin 6.2.0 NN               2023-07-09
25. Midnight 8 NN                 2023-09-26
26. Minic 3.39 NN                 2023-10-29
27. Nalwald 18 NN                 2023-08-13
28. Nemorino 6.11 NN dev          2022-08-26   Net16 (same NN file used for older versions since 6.05)
29. Pawn 2.0 NN                   2023-10-12
30. Rebel EAS NN                  2023-08-20
31. Revenge 3.0 NN                2022-07-22   DrawScore = 0 (version with adjustable contempt)
32. RubiChess 20230918 NN         2023-09-18
33. Seer 2.7.0 NN                 2023-10-28
34. Shredder 13                   2016-10-20   HCE = Hand Crafted Evaluation
35. SlowChess Blitz 2.9 NN        2022-06-01   3380 Elo (FCP-Rating-List, ending on June 29th, 2022)
36. Stockfish 13 NN               2021-02-19   Contempt = 0 (last version with adjustable contempt)
37. Texel 1.10 NN                 2023-09-25
38. Tucano 11.00.1 NN             2023-07-14
39. Uralochka 3.40a NN            2023-11-03
40. Velvet 6.0.0 NN               2023-12-21
41. Wasp 6.63 NN dev              2023-12-02   Current dev version from Wasp
42. Wasp 6.50 NN                  2023-02-28
43. Winter 2.0 NN                 2023-01-28
44. Xiphos 0.6                    2019-11-07   HCE = Hand Crafted Evaluation



History:
--------

December 02nd, 2023
- optimization UCI settings for 6-man syzygy
  if possible:
  SyzygyProbeDepth = 1
  SyzygyProbeLimit = 6
- Hiarcs 15.2 and Dragon 3.3 NN will play with "Personality: Aggressive"
- for Hakkapeliitta 3.0 NNSV: PawnHash=16

*** A new rule ***
After round 03: Engine with highest move-average is disqualificated for the next rounds = 44-1
After round 04: Engine with highest move-average is disqualificated for the next rounds = 43-1
After round 05: Engine with highest move-average is disqualificated for the next rounds = 42-1
                41 engines remain until the end of the tourney = 13.120 games

Please note:
The list of participating engines is still unclear. Booot 6.4 has a better move average than Booot 6.5 and plays a
more aggressive style of chess, but is 40 Elo weaker. I don't know the move averages of possibly newer versions of
the engines! A bit risky to have added newer versions of the engines. The move average should not be higher than
87 moves, the new rule helps! But all in all, it is possible to change this and that before the tournament starts!


December 05th, 2023
- I made an Elo average calculation for the 44 programs on start.
  The Elo average is around = ~3280 (for Shredder's tournament *.html table).
  

December 06th, 2023
- time control changed to: 66 Minutes/Game + 6 Seconds/Move



Frank Quisinsky, Germany / Gutweiler

At second the conditions:

Code: Select all

FCP-Tourney-2024-MA, hardware

Operating system: Windows 10 Professional
GUI:              Shredder 12 by Stefan Meyer-Kahlen
Processor:        AMD Ryzen™ 9 5950X / RTX 3070-Ti = 16 cores
Ghz, RAM:         4.0Ghz overclocked, DDR4 3200Mhz
Bench:            Total time= 87,18 seconds, nodes= 126164000, nps= 1447112
                  Bench results by Wasp 6.50 NN with 1 core, 768Mb hashtables
TalkChess:        U.S. Forum, thread 83065



FCP-Tourney-2024-MA, parameters / statistics

Start of tourney: December 27th, 2023 (23:45)
End of tourney:   unclear

FTP updates:      Every 10 minutes, the current table is updated with Fling Plus 5.04.
Tournament table: https://www.amateurschach.de/fling/fcp-tourney-2024.html

FTP updates:      Every 10 minutes, the current game-plan is updated with Fling Plus 5.04.
Game-Plan         https://www.amateurschach.de/fling/fcp-tourney-2024.sto

HTTPS updates:    After every or every other round. If I work on it more often (irregular)
Games & Stats     https://www.amateurschach.de/download/_fcp-tourney-2024.zip

                  ***********************************************************
                  * Current statistics after round 0 (Dec. 27th, 2023)      *
                  * Move-average: 0,00 = 0,0 moves                          *
                  * Draws: 0,00%                                            *
                  * ~150 games per day                                      *
                  ***********************************************************

Elo:              ~3282 (avg. of 44 participate engines) setting for Shredder 12 GUI
                  SlowChess Blitz 2.9 NN is playing with ~3380 Elo (reference)

Tournament type:  Everyone against everyone, 41 engines, 16 games per match,
                  640 games per engine, 820 games per round, 16 x 820 = 13.120 games total
                  Have a look under: *** A new rule ***

Engine-Updates:   Possible for Wasp development versions only
                  Possible during round 01 is still running

Time control:     66 Minutes/Game + 6 Seconds/Move

Endgame 6-man:    syzygy
Endgame 5-6-man:  gaviotabases & 6-man syzygy for Texel 1.10 NN
                  gaviotabases & 6-man syzygy for Chiron 5.01
Endgame 5-man:    gaviotabases & egbb for DanaSah 9.1 NN
Endgame 5-man:    shredderbases for Shredder 13

Opening book:     feobos-6m-v1.0.bkt for Shredder GUI
                  *.pgn can be found in this download file!

Cores:            1 core, Hyperthreading = off, Large-Pages = off
Hashtables:       1Gb
Contempt:         Default engine settings changed to contempt = 0
Other settings:   resign = off, ponder = off, learning = off

*** A new rule ***
After round 03: Engine with highest move-average is disqualificated for more rounds = 44-1
After round 04: Engine with highest move-average is disqualificated for more rounds = 43-1
After round 05: Engine with highest move-average is disqualificated for more rounds = 42-1
                41 engines remain until the end of the tourney = 13.120 games
				


FCP-Tourney-2024-MA, this and that

dev:              Developer builds from available engines are allowed.
Derivat:          Derived engines are not allowed.



Frank Quisinsky, Germany / Gutweiler

Best
Frank

[Frank Quisinsky]

Links will be the same (no changes from my last tournament).
Again, the tournament will start on 27 December 2023.
Engine updates possible during round 1 is still running.


URLs for:
1. FCP-Tourney-2024
2. FCP-Tourney-2024-MA (MA = Move-Average)

1. Shredder GUI, tournament table:
https://www.amateurschach.de/fling/fcp- ... -2024.html (could also be saved in the browser)
Every 10 minutes, the current tournament table is updated with Fling Plus 5.04.

2. Shredder GUI, game-plan:
https://www.amateurschach.de/fling/fcp-tourney-2024.sto
Every 10 minutes, the current game-plan is updated with Fling Plus 5.04.

3. This download file:
https://www.amateurschach.de/download/_ ... y-2024.zip
Available after every or every other round, if I am work on it more often.

Links will not be changed while FCP-Tourney-2024 / FCP-Tourney-2024-MA is running.
Good luck to all the participating programmers and thank you for the engines!

[Frank Quisinsky]

The most important thing for me is the move average:
So the name of the tournament is: FCP-Tourney-2024-MA (MA for Move Average).

So, I create a new rule.
The tournament starts with 44 engines, but 3 engines are out after round 05:

*** A new rule ***
After round 03: Engine with highest move-average is disqualificated for more rounds = 44-1
After round 04: Engine with highest move-average is disqualificated for more rounds = 43-1
After round 05: Engine with highest move-average is disqualificated for more rounds = 42-1
41 engines remain until the end of the tourney = 13.120 games

What is important to me?

Code: Select all

1st Priority	lower average of the moves	
2nd Priority	Wasp test (2x Wasp on start)	
3rd Priority	test-set optimisations (FEOBOS database search)	
4th Priority	midgame with time-control manipulation	
5th Priority	more short wins, more powerful tactical engines are inside	
6th Priority	lower average of the draws	
7th Priority	more HCE engines	
8th Priority	FEOBOS opening book with smaller depth

[Frank Quisinsky]

Here a bit to my reasons of the new list of participating engines!



Please note:
I forgot to set Contempt=0 for SlowChess 2.9 NN and forgot to add the version the programmer of Revenge gave me after his release (with Contempt = 0).



Best
Frank

[Frank Quisinsky]

Hi there,

the tourney have an Elo-average from around 3282!
I await after a test-tourney with 3+1, runs in the last months ...

1. Move-average from 86-87 moves
2. Draw quote from 59-60%

And all will be wonderful here in Gutweiler ... Germany, a town around 700 persons life.
I am the once with computer-chess in the brain ... and the once completly crazy.



Have a look here:

https://www.amateurschach.de/common/ma-3.jpg

Enough ...
I am ready for take-off on runway ... MA

Best
Frank

[Frank Quisinsky]

Ah ...
I forgot ... sorry!

Merry christmas to the readers.

[Frank Quisinsky]

December 24th, 2023
I have finished the work on the opening lines.
There are 4.131 lines in feobos-6m-v2.0.bkt.
The new *.bkt opening book with *.pgn file can be found in this download file.

v0.16.0.0.3 = 38,360Mb
https://www.amateurschach.de/download/_ fcp-tourney-2024.zip

With test tournaments (time control: 3+1) the current versions of Wasp, Velvet, Revenge, SlowChess, Dragon &
RubiChess reduced the draws by 4.41%, when I compare the results with my standard FEOBOS book:
feobos-v20.1-contempt_3-5_tuned-v2.bkt



Here the readme:

Code: Select all

No magic!
feobos-6m-v2.0.bkt (opening book for Shredder GUI by Stefan Meyer-Kahlen)

- collected games, 1:0 games (below move number 60 to mate) and 0:1 games (below move number 68 to mate)
  games played with feobos-20.1-contempt_3-5_tuned-v2.bkt
  games are from FCP-Tourney-2021, KI-Ratings, FCP-Tourney-2024 (round 01-15)
- games with an eval higher as 0,60 (white) / -0,75 (black) deleted.
  Tool: universal-pgn-epd-tool v2 by Ferdinand Mosca
- truncate to 6 moves
  Tool: truncate by Norm Pollock
- game results set to 1/2:1/2, different other optimizations (have a look in the *.pgn header)
- move transposition with pgn-extract deleated
  pgn-extract --fuzzydepth 0 --duplicates dupes.pgn --output unique-beta-1.pgn beta-1.pgn
  Tool: pgn-extract by David J. Barnes
  
final results = 4.131 lines (without doubles)
I hope this small and easy opening book reduced the draws with 3-4%!



Frank Quisinsky, Germany / Gutweiler

Please note:
I am working with balanced lines and using lines from the own FEOBOS project.

Examples:

Code: Select all

[Event "FEOBOS"]
[Site "Gutweiler"]
[Date "2021.04.04"]
[Round "?"]
[White "N.N."]
[Black "N.N."]
[Result "1/2-1/2"]
[WhiteElo "3000"]
[BlackElo "3000"]
[ECO "A01"]
[PlyCount "12"]
[Source "Frank Quisinsky"]
[SourceDate "2023.12.24"]

1. b3 b5 2. e3 Bb7 3. Nf3 a6 4. Bb2 e6 5. c4 b4 6. d4 Nf6 1/2-1/2

[Event "FEOBOS"]
[Site "Gutweiler"]
[Date "2022.03.20"]
[Round "?"]
[White "N.N."]
[Black "N.N."]
[Result "1/2-1/2"]
[WhiteElo "3000"]
[BlackElo "3000"]
[ECO "A05"]
[PlyCount "12"]
[Source "Frank Quisinsky"]
[SourceDate "2023.12.24"]

1. b3 b6 2. Bb2 Bb7 3. e3 Nf6 4. Nf3 e6 5. Nc3 d5 6. a4 a6 1/2-1/2

[Event "FEOBOS"]
[Site "Gutweiler"]
[Date "2021.12.09"]
[Round "?"]
[White "N.N."]
[Black "N.N."]
[Result "1/2-1/2"]
[WhiteElo "3000"]
[BlackElo "3000"]
[ECO "A30"]
[PlyCount "12"]
[Source "Frank Quisinsky"]
[SourceDate "2023.12.24"]

1. b3 b6 2. Bb2 Bb7 3. e3 Nf6 4. Nf3 e6 5. d4 g6 6. c4 Bg7 1/2-1/2

Best
Frank

[Frank Quisinsky]

Hi there,

forgotten ...
I am working since a while on a balanced test-set ... depth = 8-10 moves max.

So, participating engines from this "MP" tournament can help me a bit to find new interesting lines.
The book I created for FCP-Tourney-2024-MP have a depth = 6 moves.

Best
Frank

[Frank Quisinsky]

Hi there,

The pgn4web site is ready.
Wish from Norman Schmidt.
https://www.amateurschach.de/fling/index.html

I think this is a good idea, because the move-average will be significantly reduced with my next tournament and chess players will have more fun replaying games online.

1.
At the moment you can see the last games of the running tournament (I am testing the site). There are over 12,000 games loaded. I changed this for the "MP" tournament. Later you can see the games from the running round here. And the time for loading is clearly faster. All other material about the tournament can be found on the site too (Links). The links will also be changed before I start the tournament on 27 December 2023 (23:45). In the web directory I only put the logos for the next tournament. So not all logos are visible at the moment.

2.
Stockfish 16 NN is currently checking the 4,113 lines I have in the book version 2.0 (ready tomorrow morning). Each position is checked with an analysis of 15 seconds on 10-cores 4.4Ghz i9 10900k. The EPD and new PGN can be found tomorrow in the download file. Also a readme how I do this. Book version 2.1 will be available tomorrow and is ready for tournament start. Book will be reduced with around 400 positions ... think so.

3.
I changed Stockfish 13 NN to Stockfish 200731 dev (classic eval). I read on the Stockfish dev site that Contempt does not work perfectly with NN versions. With longer time-controls the last classic-eval version from Stockfish is around 160 Elo weaker, not more. So Stockfish is strong enough with around 3350 Elo. The level Fritz 19 NN have.

I am working ...
Will be a wonderful tournament, absolutely sure.



Best
Frank

[Frank Quisinsky]

Hi there,

I can see nothing wrong with my work of the last weeks and start the tournament on 27 December 2023 (20:00).

In cross table you can see ...
https://www.amateurschach.de/fling/fcp- ... -2024.html

0 out of 946 games played
Level: 66 Minutes/Game + 6 Seconds/Move
Games: C:\Tourney\fcp-tourney-2024.pgn ... that is a problem with Shredder-GUI, link goes to the PC

You can find all the information here ...
https://www.amateurschach.de/fling/index.html

- link to cross table
- link to game-plan
- link to original *.pgn from Shredder
- link to conditions
- link to participating engines / history
- link to my *.zip download file

0 out of 946 games played
I will restart the tournament table for every new round again with zero.
Same for the replay-zone (time for loading the games is much faster as to load again all games from all rounds again and again).

Replay-Zone and cross-table are synchron for each round.
The final results round-by-round can be found later in my *.zip download file.
So more fun for the visitors during the tourney for the next months, if others have interest on it.

Best
Frank

Interesting for me in the next hours for the first games are the time management of the engines with game in 66 minutes + 6 seconds and I hope my small and new 6-moves book has enough material in it. In the first round DanaSah vs. Rubi ... after a long time DanaSah gives the honour on my PC, a wonderful engine and I have to look in the fight Arasan - Wasp 6.50 ... the long time thriller for me. Wasp 6.63 dev. vs. Andscacs 0.95 is also running and many other interesting matches.

Thanks for the motivation I need so far to Jörg Burwitz and to Stefan Pohl with his tipps to syzygy and the time-control.