FCP Tourney-2021, 41.000x 45min. games on i9-10900k, 13xNN, now still running!

[Frank Quisinsky]

Hi there,

for the next 5 1/2 - 6 months I started FCP-Tourney-2021, yesterday 00:30 (German time).

In live mode you can see the current standings (*.html file from Shredder GUI). The *.html file will be updated each 30 minutes.
The games will be updated each 30 minutes and the *.sto Shredder tournament file (here you can see the next pairings, with X on the end the still running games).

Klaus Wlotzka is working again on Excel files during the tourney.
First will be available soon, after round 1.

From round to round the statistic files are available (one round with 820 games need around 3 days, 50 rounds to play, 245-275 games per day).

From TOP-20 of available engines only Scorpio 3.011 missed (not working with NN and Shredder GUI).
From TOP-20-30 of available engines only Fritz 16 (Rybka) missed. The style is to boring, produced to many fast lost games.
Lc0 is playing with the CPU version.

Elo average to FCP Tourney-2020 goes 45 points higher (from 3035 to 3080 after my calculation).
28 new engines / engine-version on start (if I compare it to FCP Toruney-2020, started for 6 months and ended for around 35 days).

A lot of information around the tourney can be found on my site.
Example: The detail page have after FCP Qualify Tourney-2021 ... 23-News!

Best solution is to start my main page!

Frank's Chess Page
http://www.amateurschach.de (all important links can be see on my start page)!

Note:
I am working with unmodern frames and do that on purpose.
So, users can save sites he like most without my main frame (Chess Computer WIKI like to do that for an example).

We wish all participate programmers good luck!

Best
Frank

PS-1: Fire 7.1 and Houdini 6.03 not on start (clone suspiction).
PS-2: The 41 participate programmers / programs produced "Tournament Category XXXIV (34)".

[Wolfgang]

Thanks Frank!

[Frank Quisinsky]

CEGT and CCRL testers ...
I have to say thank you.

So, CEGT and CCRL testers can compare and have also a bit for looking during all the work they do for us.
All is OK Wolfgang, but I am happy if the "Thanks" comes from yourself.

The system I am using I am order special for theses tourneys.
If I have all three days a bit to do (round by round) my wife is most happy because ...



Best
Frank

[Vinvin]

Thanks for this great tournament, Frank !!

[Frank Quisinsky]

Hi Vincent,

please, I read your message to FCP Qualify Tourney-2021 too.
Thanks for your interest!!

Have fun with the tourney, better ... have fun with the excel stats Klaus is working.
I am sure we can make, during the tourney is still running, the excel statistic with new ideas again better.

Best
Frank

[Angle]

Nice tournament, Frank! So many new versions of interesting engines.

[Frank Quisinsky]

Hi Angle,

different missed:

1. Houdini, clone suspicition (TOP 10 engine).
2. Fire, clone suspiction (TOP 10 engine).
3. Scorpio (TOP 10 engine, Shredder 12/13 GUI can not start batch files for the NN files).
4. Rybka, (TOP 20-30 engine, engine I don't like most from all available since many years, to many fast loses, bad opening style).
5. Equinox (today TOP 30-40 engine), based on IPP, fastest and most interesting IPP GullChess is under the list of participate engines, is enough!
6. ChessBrainVB (TOP 30-40 engine, to many lose on time games with ponder = off and without resign-mode).
7. Beef (TOP 30-40 engine, to many lose on time games with ponder = off and without resign-mode).
8. Hannibal (maybe TOP 30-40 engine, enough games played, to passive style).
9. Senpai (maybe TOP-30-40 engine, enough games played, to many fast loses).

I hope the Stefan Meyer-Kahlen fix sooner or later the GUI problem and that ChessBrainVB and Beef programmers will be solvd the "lose on time" problems. So I can add the engines for the next ...

All in all ...
Places for the others / newer engines are free in the group of TOP-41 engines.

Hope that the list of participate engines will be interesting enough for such a long-time tourney.

Without any more words ... most interesting for me and John = Wasp 4.50.
We are thinking ... tourneys are great, why not test all in a tourney.
Better as to test Wasp alone vs. all the others, so all the participate programmers can find interesting information!
John and myself have here the same opinion!

My tourneys are more or less a Wasp test, John and myself have more fun for looking in results and statistics.

Have fun with it too!

Best
Frank

Honestly:
I am looking most of time Wasp, Koivisto, Booot, Andscacs, Fizbo and SlowChess Blitz games here in LIVE mode on a big TV ... crime time in Trier, Germany. Koivisto can be the new Fizbo ... very speculative chess, what humans like me like a lot. Go Koivisto go ... but at first ... Go Wasp go!

[Angle]

6. ChessBrainVB (TOP 30-40 engine, to many lose on time games with ponder = off and without resign-mode).
7. Beef (TOP 30-40 engine, to many lose on time games with ponder = off and without resign-mode).
8. Hannibal (maybe TOP 30-40 engine, enough games played, to passive style).
9. Senpai (maybe TOP-30-40 engine, enough games played, to many fast loses).

I hope the Stefan Meyer-Kahlen fix sooner or later the GUI problem and that ChessBrainVB and Beef programmers will be solvd the "lose on time" problems. So I can add the engines for the next ...

I have no problems with time losses when testing Beef and ChessBrainVB with ponder = off. I am going to include ChessBrain, Wasp, Hannibal, and Senpai to my next tournament.

I am looking most of time Wasp, Koivisto, Booot, Andscacs, Fizbo and SlowChess Blitz games here in LIVE mode on a big TV ... crime time in Trier, Germany. Koivisto can be the new Fizbo ... very speculative chess, what humans like me like a lot. Go Koivisto go ... but at first ... Go Wasp go!

Koivisto is playnig very well in my current tourney. Concerning Andscacs... you included ver. 095123. According to my tests, two strongest versions are 095092 and 095105. Andscacs 095123 is nearly equal to the official 0.95 while 095092 and 095105 are approx. 40 elo ahead.

[Frank Quisinsky]

Hi Angle,

most people will not have problems with the engines and time loses.
Not the people are playing games "without resign-mode".

Both engines lost on time 4-8 moves before game ended with mate!

In case of ChessBrainVB = 8 times after 320 FCP Qualify Tourney-2021 games!!
If 8 times after 320 games = 50 times after 2.000 FCP Tourney-2021 games ... clearly to much!

In case of Beef 0.36 = 4 times after the first 50 games I try it out.

That's the problem, most people and programmers test always with resign = on and can not see such problems!
For such problems I wrote in the last 20 years around 150 mails.

I can do nothing with games ended with -6 or higher (resign = on).
For strong statistics I need games goes to mate.

5 others engines from TOP-100 have the same problem, can be see in my new "engine-configuration" site. Latest Craftry rarely, or TheBaron rarely, other.

Highest quote on short won games comes from 0.95!!
But 0.95-123 is around 16 Elo stronger (after 2.000 test games).
I tested two other beta versions, inside the Andscacs collection.

All what I have added in the tourney (the Texel beta version, or the GullChess 3.0 Sy version) ... are the strongest available versions I found out. Many test-games played!

Best
Frank

I have an Andscacs collection on my site with the permission from the programmer of Andscacs. The versions you are tested aren't in the collection. It would be great if you can send me the versions! Many people helps for around 6 months to make the collection better. So, if you got beta versions from Andscacs please send the versions to myself ... all of the Andscacs fans should be able to use it. So the programmer of Andscacs and his fans are happy.

Hannibal, Senpai: I have more as 6.000 games with longer time controls played, most are available on my site.
Hannibal and Senpai are in the group of FCP Tourney-2020 engines. Same for Fire, Houdini and Fritz 16 (last Rybka).
All these engine works fine without any problems. Most interesting is Rybka 4.1 with special settings for white / black pieces I tested for SWCR-Rating List. The style is much more interesting as the boring and buggy mid-game from the last Chessbase edition. The special settings comes from Rybka Forum. Some persons found out and indeed ... playing strength goes high.

[Frank Quisinsky]

[ENGINE]
Name=Rybka 4 x64 Exp. 42
Author=Vasik Rajlich
Filename=C:\Chess\Shredder 12\Engines\Rybka\Rybka_4-x64.exe
[OPTIONS]
Max CPUs=1
TC Normal Move Time=100
TC Max Move Time=150
White Pawn cp=3
Black Pawn cp=2
White Knight cp=10
Black Knight cp=8
White King Bishop cp=10
Black King Bishop cp=8
White Queen Bishop cp=10
Black Queen Bishop cp=8
White Rook cp=17
Black Rook cp=13
White Queen cp=39
Black Queen cp=33

Here is the most interesting settings, not v4.1 used for 4.0!
Elo from v4.0 goes higher as v4.1.
And that is the once Rybka version I like a bit (but after my opinion, Rybka is cloned).
v1 from Fruit. And today all the problems Fruit / Senpai produced in mid-games, all the same problems prodcued Rybka and later most of IPPs.