Hi there,
Stockfish not won round 14, 15, 16 but the first 13 rounds.
In the last round Winter have fantastic results.
Now all results from round 01-16 of 50 are online.
http://www.amateurschach.de
Have fun with my small tourney.
Best
Frank
FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
Moderators: hgm, Rebel, chrisw
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
Hi there,
please have a look in the stats from Klaus Wlotzka.
After the tourney we have interest to create much more stats in Excel.
For the moment lesser stats by Klaus are available (to many work to added more details from round to round).
But the work Klaus do during the tournament is great!
You will like it, sure here!
Click on
Tournament detail page ...
Here you can found three links to the helps from Klaus during the FCP Tourney-2020 is still running.
Best
Frank
please have a look in the stats from Klaus Wlotzka.
After the tourney we have interest to create much more stats in Excel.
For the moment lesser stats by Klaus are available (to many work to added more details from round to round).
But the work Klaus do during the tournament is great!
You will like it, sure here!
Click on
Tournament detail page ...
Here you can found three links to the helps from Klaus during the FCP Tourney-2020 is still running.
Best
Frank
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
Hi there,
now the results from round 17 of 50 are online.
Best
Frank
now the results from round 17 of 50 are online.
Best
Frank
-
jonkr
- Posts: 178
- Joined: Wed Nov 13, 2019 1:36 am
- Full name: Jonathan Kreuzer
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
Just wanted to say I think it's interesting to see more stats about the games than just pure game results.
I've tried to make sure Slow Chess can be aggressive, and these results back it up so far. ( I did find that even as it got more aggressive against the same weaker opponents, as the opponents got stronger quick wins became rare, and partially dependent on opening, which is one reason I like bullet and FRC.)
It looks like results so far mostly confirm that the top engines almost never blunder so badly they get mated in under 60 moves, so lack of quick games between top engines is mostly from opponent strength rather than lack of aggression. (I've seen games occasionally decided early on by an attack, but even a piece advantage can take a long time to close out.)
It's also interesting to look at variation in number of quick losses from the lower rated engines, at that level it seems there can be quite a variation in King Safety from similar strength engines. At higher level it looks like mostly more wins = more quick wins.
I've tried to make sure Slow Chess can be aggressive, and these results back it up so far. ( I did find that even as it got more aggressive against the same weaker opponents, as the opponents got stronger quick wins became rare, and partially dependent on opening, which is one reason I like bullet and FRC.)
It looks like results so far mostly confirm that the top engines almost never blunder so badly they get mated in under 60 moves, so lack of quick games between top engines is mostly from opponent strength rather than lack of aggression. (I've seen games occasionally decided early on by an attack, but even a piece advantage can take a long time to close out.)
It's also interesting to look at variation in number of quick losses from the lower rated engines, at that level it seems there can be quite a variation in King Safety from similar strength engines. At higher level it looks like mostly more wins = more quick wins.
-
jdart
- Posts: 4367
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
As I mentioned, I think a lot of the short losses are due to bad book lines or a bad decision shortly out of book. Most engines out of the very top ranks still play openings rather badly.
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
Hi Jon,
33.309 Contempt 3 FEOBOS lines are inside the FEOBOS book I am using for FCP Tourney-2020.
End positions checked by 10 engines with 1 minute and 6 cores on Intel i7-5630 with 6 cores per analyzes.
The complete Excel file with around 16 millions of formulas are on my website, billions nps engines produced, the complete FEOBOS development files are online.
10 engines sorted out in teamwork all the bad lines, that what we do here.
An gigantic project, runs around 4 years.
The opening book with lines to 500 ECO codes are bugfree and FEOBOS try to simulate the GM theory from the last 10 years. FEOBOS try to avoid fast draw games (I am using here Contempt=3) too and the PGN with the final result is sorted with a ranking system. We developed a ranking system for opening book lines. With the final FEOBOS results you can create every test-set you like.
Examples:
20 Best B01 lines ...
Or ... What is the best A00, A01, A02 line for create a ECO-500 test-set.
Thats important, so grandmaster have to all popular theory moves the information what engines like to play.
But absolutely right is, that a lot of engines don't understand different opening systems.
So FEOBOS opening system will help the programmers to find out the problems very fast.
Best
Frank
http://www.amateurschach.de/main/_new-opening-book.htm
Bad openings lines for engine testing ... that's water under the bridge!!
The tourney runs with a perfect opening book, created in special for such things!!
PS:
Give me one bad position and I will give you directly 10 engines analyzes and the place of the position in FEOBOS ranking systems.
All such things are available in our Excel file.
An example can be see on my start page:
https://www.amateurschach.de (Excel grafic to the Ethereal - Fizbo game).
33.309 Contempt 3 FEOBOS lines are inside the FEOBOS book I am using for FCP Tourney-2020.
End positions checked by 10 engines with 1 minute and 6 cores on Intel i7-5630 with 6 cores per analyzes.
The complete Excel file with around 16 millions of formulas are on my website, billions nps engines produced, the complete FEOBOS development files are online.
10 engines sorted out in teamwork all the bad lines, that what we do here.
An gigantic project, runs around 4 years.
The opening book with lines to 500 ECO codes are bugfree and FEOBOS try to simulate the GM theory from the last 10 years. FEOBOS try to avoid fast draw games (I am using here Contempt=3) too and the PGN with the final result is sorted with a ranking system. We developed a ranking system for opening book lines. With the final FEOBOS results you can create every test-set you like.
Examples:
20 Best B01 lines ...
Or ... What is the best A00, A01, A02 line for create a ECO-500 test-set.
Thats important, so grandmaster have to all popular theory moves the information what engines like to play.
But absolutely right is, that a lot of engines don't understand different opening systems.
So FEOBOS opening system will help the programmers to find out the problems very fast.
Best
Frank
http://www.amateurschach.de/main/_new-opening-book.htm
Bad openings lines for engine testing ... that's water under the bridge!!
The tourney runs with a perfect opening book, created in special for such things!!
PS:
Give me one bad position and I will give you directly 10 engines analyzes and the place of the position in FEOBOS ranking systems.
All such things are available in our Excel file.
An example can be see on my start page:
https://www.amateurschach.de (Excel grafic to the Ethereal - Fizbo game).
Last edited by Frank Quisinsky on Sun Aug 16, 2020 10:13 pm, edited 1 time in total.
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
What I like is ...
FCP Tourney-2020, is the biggest tourney ever played in computer chess history.
FCP Tourney-2020, with the fastest hardware with 1 Core (Intel i9-10900, with 4.9Ghz).
FCP Tourney-2020, simulated the complete current GM-Theory, bugfree ... all 500 ECO codes.
FCP Tourney-2020, is the tourney with the hightest category XXXII every played.
FCP Tourney-2020, 41.000 games!
FCP Tourney-2020, highest time control in combination with fastest hardware.
FCP Tourney-2020, a beta test for the best 41 engines to the same time!
And I have an Excel expert for statistics ... (the files Klaus Wlotzka generated during the tourney).
After the tounrey we can more do, not during the tourney.
To update each of the statisics we have, round by round, is a lot of work.
I think we have for the moment enough stats in the tournament package, means during the tournament is still running.
More isn't possible, or we are sitting the complete day on the PC.
Of course, more is possible ... but after the tourney!
FCP Tourney-2020, is the biggest tourney ever played in computer chess history.
FCP Tourney-2020, with the fastest hardware with 1 Core (Intel i9-10900, with 4.9Ghz).
FCP Tourney-2020, simulated the complete current GM-Theory, bugfree ... all 500 ECO codes.
FCP Tourney-2020, is the tourney with the hightest category XXXII every played.
FCP Tourney-2020, 41.000 games!
FCP Tourney-2020, highest time control in combination with fastest hardware.
FCP Tourney-2020, a beta test for the best 41 engines to the same time!
And I have an Excel expert for statistics ... (the files Klaus Wlotzka generated during the tourney).
After the tounrey we can more do, not during the tourney.
To update each of the statisics we have, round by round, is a lot of work.
I think we have for the moment enough stats in the tournament package, means during the tournament is still running.
More isn't possible, or we are sitting the complete day on the PC.
Of course, more is possible ... but after the tourney!
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
At first to SlowChess:Just wanted to say I think it's interesting to see more stats about the games than just pure game results.
I've tried to make sure Slow Chess can be aggressive, and these results back it up so far. ( I did find that even as it got more aggressive against the same weaker opponents, as the opponents got stronger quick wins became rare, and partially dependent on opening, which is one reason I like bullet and FRC.)
It looks like results so far mostly confirm that the top engines almost never blunder so badly they get mated in under 60 moves, so lack of quick games between top engines is mostly from opponent strength rather than lack of aggression. (I've seen games occasionally decided early on by an attack, but even a piece advantage can take a long time to close out.)
It's also interesting to look at variation in number of quick losses from the lower rated engines, at that level it seems there can be quite a variation in King Safety from similar strength engines. At higher level it looks like mostly more wins = more quick wins.
- SlowChess lost 52 games after 680, sure good material for looking.
- Move average = good
- Draw quote without contempt = to many fast draw games if we compare it with the others.
- Short game stats = good
- results vs. Schooner aren't good (can be random after only 17 rounds).
Maybe most feared opponent?
SlowChess is very strong, TOP-5 for the moment and the stats for the Elo SlowChess have seems to be normaly, I think better as normaly. I saw a lot of nice games with many pieces on board. I can't see any opening system SlowChess have problems. Interesting in questions of SlowChess are the E90-E99 theory. I remember me on 4 games I saw in live mode. One game vs. Arasan (E97 or E99) just great what SlowChess do after opening books moves because this opening system is not easy to understand for computer chess engines). Not easy to find out problems from engines with such a high Elo performance, SlowChess have. But maybe the 51 lost games will helps here to search a bit.
All in all ...
SlowChess, without to create more statistics in detail ... short comments:
SlowChess produced to many fast draw games but a lot of very nice aggressive games with many pieces on board (what I like most).
I missed for many engines more speculative moves Fizbo like to play and more speculative attacking moves with black pieces. Here SlowChess have strong results (white / black quantity of fast won games).
Again, not easy to find bad things in SlowChess. For the moment we are looking only in the middlegame, not in transposition into endgame and endgame. This we will do later, after the tourney with interesting statistics in Excel.
Great engine ...
Good statistics ...
High Elo level ... no weaknesses I can see for the moment
Nice style
But to many fast draw games ...
For myself the result for SlowChess for the moment!
And for Wasp testing ...
I try to find out the group of best engines for Wasp testing (main reason for the tournament).
No Wasp feature test-run without SlowChess ... that's quit clear.
SlowChess can help for Wasp development, for find mistakes in Wasp.
Best
Frank
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
I think it's not easy to avoid draw without Contempt.
With FEOBOS book contempt is a bit possible but engine should have more knowledge in source code.
FEOBOS opening book produced 0.35% short draw games below 20 moves only (if a book can play all 500 ECO codes is this book statistic sensational). Also the fast draws below 30 moves are with FEOBOS sensational.
Looking in Hannibal and Demolito stats.
High move average with rarely short draw games.
It's a way, OK but ...
The secret is:
Lesser moves average with rarely short draw games.
Wasp is here on a good way and we have other engines with perfect statistics to that point (Fritz 17 for an example).
Looking on Wasp:
Wasp avoid often draw in late middlegames and vs. stronger opponents many of that games Wasp lost later in transposition into endgame.
I think if 40-50 moves played and a draw is possible ... if the pawn structure isn't perfect ... engine should give draw and not avoid draw. I think, that many programmers looking here more in detail (example I gave with Fritz 17).
Booot can be much more improved here ...
Lesser move average (nice) but higher draw quote (the problem) ...
Booot should try to avoid draw if not more as 40-50 moves played or better ... if more as 16 pieces on the board and I am sure the performance for Booot will be better.
Fritz 17 is strong here ...
Lesser move average, lesser draw quote.
I think Fritz 17 have special parameters for such things!
Dynamic Contempt can be interesting!
Engine should not avoid draw with more as 16 pieces on board if the pawn structure is perfect or should give directly the draw if the pawn structure isn't perfect. Here is important only ... how strong are the own pawn structure.
Best
Frank
To the readers ...
Sorry for my bad English!!
With FEOBOS book contempt is a bit possible but engine should have more knowledge in source code.
FEOBOS opening book produced 0.35% short draw games below 20 moves only (if a book can play all 500 ECO codes is this book statistic sensational). Also the fast draws below 30 moves are with FEOBOS sensational.
Looking in Hannibal and Demolito stats.
High move average with rarely short draw games.
It's a way, OK but ...
The secret is:
Lesser moves average with rarely short draw games.
Wasp is here on a good way and we have other engines with perfect statistics to that point (Fritz 17 for an example).
Looking on Wasp:
Wasp avoid often draw in late middlegames and vs. stronger opponents many of that games Wasp lost later in transposition into endgame.
I think if 40-50 moves played and a draw is possible ... if the pawn structure isn't perfect ... engine should give draw and not avoid draw. I think, that many programmers looking here more in detail (example I gave with Fritz 17).
Booot can be much more improved here ...
Lesser move average (nice) but higher draw quote (the problem) ...
Booot should try to avoid draw if not more as 40-50 moves played or better ... if more as 16 pieces on the board and I am sure the performance for Booot will be better.
Fritz 17 is strong here ...
Lesser move average, lesser draw quote.
I think Fritz 17 have special parameters for such things!
Dynamic Contempt can be interesting!
Engine should not avoid draw with more as 16 pieces on board if the pawn structure is perfect or should give directly the draw if the pawn structure isn't perfect. Here is important only ... how strong are the own pawn structure.
Best
Frank
To the readers ...
Sorry for my bad English!!
-
jdart
- Posts: 4367
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: FCP Tourney-2020: 41.000 games 20+5 to play on Intel i9-10900k
Here is one example. There are many others:PS:
Give me one bad position and I will give you directly 10 engines analyzes and the place of the position in FEOBOS ranking systems.
All such things are available in our Excel file.
[pgn][Event "20 Minutes/Game + 5 Seconds/Move"]
[Site "fcp-tourney-2020, WASP-1"]
[Date "2020.08.09"]
[Round "16.17"]
[White "Fritz 17 (Ginkgo) x64"]
[Black "Fritz 16 (Rybka) x64"]
[Result "1-0"]
[ECO "D20"]
[PlyCount "87"]
[EventDate "2020.??.??"]
1. d4 d5 2. c4 dxc4 3. e4 Nf6 4. e5 Nd5 5. Bxc4 Nb6 6. Bd3 Nc6 7. Nf3 Be6 $2 (
7... Bg4 $1) 8. Nc3 Qd7 9. O-O O-O-O 10. Be3 a6 11. Rc1 Kb8 12. Ng5 Bg4 13. f3
Bh5 14. g4 f6 15. Bf5 Qe8 16. Ne6 Bg6 17. d5 Bxf5 18. gxf5 Nc4 19. Bf2 N6xe5
20. f4 Rc8 21. fxe5 Nxe5 22. Na4 Qb5 23. Qc2 c6 24. Nb6 g6 25. Nxc8 Kxc8 26.
dxc6 bxc6 27. Be3 Rg8 28. Rfd1 Bg7 29. fxg6 hxg6 30. Qg2 Qb7 31. Nxg7 Rxg7 32.
Qh3+ Kc7 33. Bf4 Qb6+ 34. Kg2 e6 35. Qh8 Qxb2+ 36. Kh1 Rf7 37. Qg8 Qb5 38.
Qxf7+ Kb6 39. Qe7 c5 40. Rb1 f5 41. Qd8+ Kb7 42. Bxe5 Ka7 43. Rd7+ Qb7+ 44.
Rbxb7# 1-0
[/pgn]