The strongest engines ... for opening analyzes ... statistic!

Frank Quisinsky · Post by **Frank Quisinsky** » Thu Jun 02, 2022 8:31 am

Hi there,

at the moment I am looking a bit in stats from the own 80min games on 4.4Ghz Hardware.
Stats end of this positings are currently very clear!

For me most important to build an engine-team of 10 engines for the next FEOBOS project and the analyzes of A00-E99 positions.

For this purpose the following combination is necessary:
- strong in king-attacks with many pieces on board
- strong in king-safty with many pieces on board

In releation to Elo here my TOP-10!

01. Stockfish
02. Dragon (Komodo)
03. Revenge
04. Ethereal ... king safety is TOP, for 13.07 without NN also !!
05. SlowChess Blitz
06. Fire
07. Arasan ... a little surprise for me !
08. Wasp ... have a look in king safety !!!
09. Booot
10. Velvet ... have a look in attacking chess and the Elo the program have !!!

In around 6-8 weeks I will looking again if my FCP Tourney-KI is finnished!
Maybe interesting for the others?

The games can be found under:

Miniatures:
https://www.amateurschach.de/main/_miniatures.htm

Best
Frank

Code: Select all

Trier, Germany
Mai 30th, 2022 (19:45)

Games in database (after run-4)      : 98.310
Short checkmates below 50 moves      :  1.429 = 1,453%
80min games on 4.4Ghz hardware!


Rank                           won                                                  lost                       Rank   Rank+Rank            ELO AVG   Vers.

01. Stockfish             :    136 games out of  4.320 games =      3,148%            0       =      0%         01.       2       *op-a     3477      3
02. Lc0 GPU               :     37 games out of  1.200 games =      3,083%            0       =      0%         01.       3                 3483      1
03. Dragon (Komodo)       :    127 games out of  4.950 games =      2,565%            0       =      0%         01.       4       *op-a     3460      3
04. Revenge               :     65 games out of  3.600 games =      1,805%            1 game  =      0,027%     06.      10       *op-a     3335      2
05. SlowChess Blitz       :     48 games out of  3.600 games =      1,333%            1 game  =      0,027%     06.      11       *op-a     3365      2
06. Arasan                :     60 games out of  4.860 games =      1,234%            2 games =      0,041%     08.      14       *op-a     3258      3
07. Fire                  :     50 games out of  4.080 games =      1,222%            1 game  =      0,024%     05.      12       *op-a     3350      2
08. Igel                  :     38 games out of  3.390 games =      1,120%            6 games =      0,176%     15.      23                 3274      1
09. Ethereal              :     36 games out of  3.750 games =      0,960%            0       =      0%         01.      10       *op-a     3301      2
10. Booot                 :     38 games out of  4.110 games =      0,924%            6 games =      0,145%     13.      23       *op-a     3221      2
-----------------------------------------------------------------------------------------------------------------------------------------------------------
11. Velvet                :     24 games out of  3.120 games =      0,769%           10 games =      0,320%     22.      33       *op-a     3140      2
12. Laser                 :     26 games out of  3.390 games =      0,766%            6 games =      0,176%     15.      27                 3130      1
13. Wasp                  :     67 games out of  8.790 games =      0,762%            4 games =      0,045%     09.      22       *op-a     3176      7
14. Berserk               :     32 games out of  4.470 games =      0,715%            4 games =      0,089%     10.      24                 3357      3
15. Tucano                :     17 games out of  2.400 games =      0,708%            4 games =      0,166%     14.      29                 3228      1
15. rofChade              :     44 games out of  6.210 games =      0,708%           11 games =      0,177%     17.      32                 3284      5
17. Schooner              :     24 games out of  3.390 games =      0,707%            4 games =      0,117%     12.      29                 3138      1
17. Defenchess            :     24 games out of  3.390 games =      0,707%           11 games =      0,324%     23.      40                 3140      1
19. RubiChess             :     30 games out of  4.320 games =      0,694%            4 games =      0,092%     11.      30                 3322      3
20. Seer                  :     37 games out of  5.520 games =      0,670%           12 games =      0,217%     18.      38                 3295      3
-----------------------------------------------------------------------------------------------------------------------------------------------------------
21. Vajolet2              :      8 games out of  1.200 games =      0,666%            3 games =      0,250%     19.      40                 3017      1
22. Weiss                 :     22 games out of  3.390 games =      0,648%           11 games =      0,324%     23.      45                 3157      1
23. Nemorino              :     21 games out of  3.390 games =      0,619%           17 games =      0,501%     31.      54                 3254      1
24. Xiphos                :     19 games out of  3.390 games =      0,560%            9 games =      0,265%     20.      44                 3175      1
24. Lc0 CPU               :     20 games out of  3.570 games =      0,560%           59 games =      1,652%     46.      70                 3238      1
26. Koivisto              :     27 games out of  5.010 games =      0,538%           15 games =      0,299%     21.      47                 3357      4
27. Expositor             :      6 games out of  1.200 games =      0,500%           26 games =      2,166%     49.      76                 3017      1
28. Fizbo                 :     16 games out of  3.390 games =      0,471%           15 games =      0,442%     27.      55                 3110      1
29. DanaSah               :     13 games out of  2.880 games =      0,451%           36 games =      1,250%     38.      67                 3106      1
30. Fritz (Ginkgo)        :     18 games out of  4.140 games =      0,434%           36 games =      0,869%     36.      66                 3151      2
-----------------------------------------------------------------------------------------------------------------------------------------------------------
31. Rebel                 :     17 games out of  3.930 games =      0,432%           17 games =      0,432%     26.      57                 3162      3
32. Combusken             :     15 games out of  3.600 games =      0,416%           51 games =      1,416%     41.      73                 3100      2
32. Demolito              :      5 games out of  1.200 games =      0,416%           20 games =      1,666%     47.      79                 3027      1
34. Shredder              :     13 games out of  3.390 games =      0,383%           20 games =      0,383%     25.      59                 3125      1
35. LG Evolution          :      5 games out of  1.380 games =      0,362%           20 games =      1,449%     43.      77                 3105      1
36. Marvin                :     12 games out of  3.390 games =      0,353%           16 games =      0,471%     30.      66                 3114      1
37. Texel                 :      4 games out of  1.200 games =      0,333%           15 games =      1,250%     38.      75                 2983      1
38. Andscacs              :     12 games out of  3.630 games =      0,330%           17 games =      0,468%     29.      67                 3097      2
39. Winter                :      6 games out of  1.890 games =      0,317%           29 games =      1,534%     45.      84                 3079      1
40. Minic                 :     12 games out of  4.110 games =      0,291%           24 games =      0,583%     33.      73                 3203      2
-----------------------------------------------------------------------------------------------------------------------------------------------------------
41. Clover                :     14 games out of  4.830 games =      0,289%           46 games =      0,952%     37.      78                 3173      2
42. Halogen               :     12 games out of  4.290 games =      0,279%           55 games =      2,331%     51.      93                 3183      3
43. Chiron                :      9 games out of  3.390 games =      0,265%           18 games =      0,530%     32.      75                 3109      1
44. GullChess             :      5 games out of  1.890 games =      0,264%           39 games =      2,063%     48.      92                 3074      1
45. Hiarcs                :      6 games out of  2.400 games =      0,250%           11 games =      0,458%     28.      73                 3107      1
45. Counter               :      3 games out of  1.200 games =      0,250%            8 games =      0,666%     35.      80                 3009      1
45. Drofa                 :      3 games out of  1.200 games =      0,250%           26 games =      2,166%     49.      94                 3032      1
45. Fritz (Rybka)         :      3 games out of  1.200 games =      0,250%           39 games =      3,250%     54.      99                 3015      1
49. Dark Toga             :      4 games out of  1.710 games =      0,233%           10 games =      0,584%     34.      83                 3107      1
50. Nirvanachess          :      2 games out of  1.200 games =      0,166%           17 games =      1,416%     41.      91                 3037      1
-----------------------------------------------------------------------------------------------------------------------------------------------------------
51. Stash                 :      6 games out of  3.810 games =      0,157%           50 games =      1,312%     40.      91                 3085      3
52. Zahak                 :      7 games out of  5.460 games =      0,128%          134 games =      2,454%     52.     104                 3140      3
53. Orion                 :      2 games out of  1.710 games =      0,116%           26 games =      1,520%     44.      97                 3074      1
54. Mantissa              :      1 game  out of  1.200 games =      0,083%           38 games =      3,166%     53.     107                 3030      1
55. Black Marlin          :      2 games out of  3.120 games =      0,064%          130 games =      4,166%     57.     112                 3032      2
56. Mr Bob                :      0 games out of  1.200 games =      0,000%           64 games =      5,333%     56.     112                 2989      1
56. Bit-Genie             :      0 games out of  1.380 games =      0,000%           55 games =      3,985%     55.     111                 3119      1
-----------------------------------------------------------------------------------------------------------------------------------------------------------

--. Andscacs NNSf         :     41 games out of  1.200 games =      3,416%            0       =      0%
--. ShashChess (SF)       :     65 games out of  2.580 games =      2,519%            0       =      0%
--. Rodent NNSf           :      9 games out of  1.740 games =      0,517%            2 games =      0,114%
--. Coiled NNSf           :      4 games out of  1.200 games =      0,333%            3 games =      0,250%
--. Rodent NNRe           :      3 games out of  1.410 games =      0,212%           25 games =      1,773%
--. Caissa NNSf           :      4 games out of  2.580 games =      0,155%           35 games =      1,356%
--. Toga IV NNRe          :      0 games out of  1.380 games =      0,000%           15 games =      1,086%
--. Gogobello NNSf        :      0 games out of  1.230 games =      0,000%           31 games =      2,520%

---------------------------------------------------------------------------------------------------------------------------------------------------------
*BA = TOP-10 CPU-Engines for chess openings analyzes!
---------------------------------------------------------------------------------------------------------------------------------------------------------

Note:
- engine version's I tested since around Nov. 2021.

In this time Halogen dev. is cleary improved. Many games played with v10, problems with king safety.
It seems that Koivisto in newer version are clearly improved in attacking chess.
Different engines have problems with mating (not the fastest way to set mate).

But all in all ...
The "TOP-10" are very clear in my opinion!

Frank Quisinsky · Post by **Frank Quisinsky** » Thu Jun 02, 2022 9:14 am

Hi there,

I forgot, no engine have a lesser move-average as Velvet!
This is TOP !!!

Can be see here ... 75 moves (games without resign-mode):

KI-Ratings:
https://www.amateurschach.de/t-ki/common/ki-ratings.jpg

Best
Frank

Ferdy · Post by **Ferdy** » Sat Jun 04, 2022 6:46 pm

Publish a web page for the fcp-2022 top 41 chess engines tournament. Just some tables, couple of stats and game replay for decisive games. Most table headers have sort and search capabilities.

Frank Quisinsky · Post by **Frank Quisinsky** » Sat Jun 04, 2022 7:16 pm

Ferdy,

wow ...
Have some more ... "run-1 - run-5" ... 24600er tourneys!
"run-6 is still running since yesterday" ... without NNRe, NNSf ...

A dream to see that!!
Great work ...

We can do so many interesting things with material "each one against each others".

I saw your work and thinking ...
Ferdy, honestly ...

I am 55 years (not very old) but after so many things I do for computer chess I am thinking often ... I am more on the end of the own energy (I made to many mistakes in the last months). You are young and full power and energy, many others young programmers also.

Need no worry, the power and energy of the newer generation is great!
A good example is the Koivisto team, programmers like to help others ... the Ethereal guy is a good example.
I am not a best friend of Andrew ... but I like his work a lot!

Feel free to use my material for your own ideas.
Sure, more interesting as allways the same "QuisiQuasi" stuff.

Really great ...
Thank you, you made me really happy!

Best
Frank

Much better as my old website ...
I need a website chief ... all the work with old *.html is really very out.

Frank Quisinsky · Post by **Frank Quisinsky** » Sat Jun 04, 2022 7:22 pm

Just great, can see 50-moves rules ... so many other things!

Ferdy · Post by **Ferdy** » Sun Jun 05, 2022 9:48 am

Thanks for letting me use the pgn file. There are some games on 3-fold repetition, should be fine if ply count is not low. I added some items in the stats, one of those is 3-fold repetition interactive where we can adjust some player ratings and option to include opening to see the histogram of ply count. Like what is the mean of ply count in sicilian when white is weaker than black? or when both players are strong.

Ferdy · Post by **Ferdy** » Mon Jun 06, 2022 6:11 am

Added player and opening interactive interface one thing I notice is that Stockfish has more pirc as white. Perhaps pairing software can be improved.

Overall the white pirc leads.

Frank Quisinsky · Post by **Frank Quisinsky** » Mon Jun 06, 2022 7:50 am

Hi Ferdinand,

words fail me!
To be at the throat of someone ... my FEOBOS!
Remember, in FEOBOS are different of your own ideas and programmings inside!

I am working in my tourney's with the own FEOBOS ... very selfish!
Means, like more to see 1. e4 as 1. d4 openings.
OK, all 500 ECO codes are inside but around 5% more for openings I like to see.
= ~5% more 1. e4 as 1. d4 comparing to Grandmasters currently praxis.

Sure that you analyze different things with your great work.
End of your work it can be interesting to get your recommandation?!

in questions of:

- how many participating engines
- how many games per pairing
- max. Elo differences between the the strongest / weakest engine in the group
- or to others topics you are thinking ... needs to be improved?

Best
Frank

Modern Times · Post by **Modern Times** » Mon Jun 06, 2022 8:08 am

Ferdy wrote: ↑Sat Jun 04, 2022 6:46 pm Publish a web page for the fcp-2022 top 41 chess engines tournament. Just some tables, couple of stats and game replay for decisive games. Most table headers have sort and search capabilities.

Very nice !

Ferdy · Post by **Ferdy** » Tue Jun 07, 2022 6:35 am

Frank Quisinsky wrote: ↑Mon Jun 06, 2022 7:50 am Remember, in FEOBOS are different of your own ideas and programmings inside!

The opening names and ECO stats that I used were based on eco.pgn from pgn-extract.

I will be using a revised eco.pgn based from your eco.pgn.

Example:

Code: Select all

[Event "ECO"]
[Site "Trier"]
[Date "2020.12.15"]
[Round "1"]
[White "ECO"]
[Black "ECO"]
[Result "1/2-1/2"]
[WhiteElo "2800"]
[BlackElo "2800"]
[ECO "A04/11"]
[PlyCount "21"]
[Source "unknown"]
[Opening "Reti: King's Indian attack, Neo-closed Sicilian"]

1. Nf3 c5 2. g3 Nc6 3. Bg2 g6 4. O-O Bg7 5. d3 d6 6. e4 e5 7. c3 Nge7 8. a3 O-O
9. b4 b6 10. Bb2 h6 11. Nbd2 1/2-1/2

Will be revised to:

Code: Select all

[Opening "Reti"]
[Variation "King's Indian attack, Neo-closed Sicilian"]

So that it is easier to generate opening name stats and variation stats if needed.

According to wiki on list of openings at https://en.wikipedia.org/wiki/List_of_chess_openings:

Code: Select all

A04:
A04 Réti Opening: ...

Code: Select all

A11:
A11 English, Caro–Kann defensive system, 1...c6 (without: 2.e4, 2.d4)

And according to the line:

Code: Select all

1. Nf3 c5 2. g3 Nc6 3. Bg2 g6 4. O-O Bg7 5. d3 d6 6. e4 e5 7. c3 Nge7 8. a3 O-O
9. b4 b6 10. Bb2 h6 11. Nbd2 1/2-1/2

There is no c4 move, so perhaps this cannot be transposed into english opening as english opening starts with c4 with ECO starting at A10. We can break the A04/11 code into:

Code: Select all

[ECO "A04"]
[ECO1 "A11"]

where:
ECO is the main classification name
ECO1 is the secondary classification name

Stats can use the main ECO code and ECO1 if needed.

Perhaps the eco code in FEOBOS eco.pgn needs to be reviewed.

The eco.pgn in pgn-extract has also its drawback. This is the one I used currently.

Example:

Code: Select all

[ECO "B07"]
[Opening "Pirc defence"]

1. e4 d6 2. d4 Nf6 3. Nc3 *

[ECO "B07"]
[Opening "Pirc"]
[Variation "Ufimtsev-Pytel variation"]

1. e4 d6 2. d4 Nf6 3. Nc3 c6 *

Notice the Opening value one is "Pirc defence" and the other is just "Pirc". So during generation of opening name stats I can have entries for both but clearly this is just Pirc.

Scid has its own eco too at https://sourceforge.net/p/scid/code/ci/ ... e/scid.eco.

Example:

Code: Select all

B07a "Pirc"  1.e4 d6 *
B07a "Pirc"  1.e4 d6 2.g3 *
B07a "Pirc"  1.e4 d6 2.c4 *
B07a "Pirc: 2.Nc3"  1.e4 d6 2.Nc3 *
B07a "Pirc: 2.Nc3"  1.e4 d6 2.Nc3 Nf6 *
B07b "Pirc: 2.Nf3"  1.e4 d6 2.Nf3 *
B07b "Pirc/Reti: Wade Defence"  1.e4 d6 2.Nf3 Bg4 *
B07b "Pirc: 2.Nf3 Nf6"  1.e4 d6 2.Nf3 Nf6 *
B07b "Pirc: 2.Nf3 Nf6 3.Nc3"  1.e4 d6 2.Nf3 Nf6 3.Nc3 *
B07c "Pirc: 2.f4"  1.e4 d6 2.f4 *
B07c "Pirc: 2.f4 Nf6"  1.e4 d6 2.f4 Nf6 *
B07c "Pirc: 2.f4 Nf6 3.Nc3"  1.e4 d6 2.f4 Nf6 3.Nc3 *

Code: Select all

B07c "Pirc: 2.f4 Nf6 3.Nc3"  1.e4 d6 2.f4 Nf6 3.Nc3 *

Can be reformatted into:

Code: Select all

code: B07c, name: Pirc, variation: 2.f4 Nf6 3.Nc3, line: 1.e4 d6 2.f4 Nf6 3.Nc3 *

Converting this to pgn is fun.

The strongest engines ... for opening analyzes ... statistic!

The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!

Re: The strongest engines ... for opening analyzes ... statistic!