Hakkapeliitta the "STRONGEST" in tactic, new Nr1

[Frank Quisinsky]

Hi there,

at first ...
Have a look here ... end of the table ... won and lost up to mate and move number 50 ...

Code: Select all

    Programs                      Score       1-0  =  0-1         EloS  Baye  Ordo  +   -  Draws  MoveØ won<50 lost<50
01. Komodo 9 x64                  785.5/1000  602 367  31  78.5%  3129  3096  3142  18 18  36.7%   87     14       2
02. Stockfish 26Apr2015 BMI2 x64  783.5/1000  593 381  26  78.3%  3127  3092  3140  18 17  38.1%   81     20       0
03. Fire 4 x64                    673.5/1000  450 447 103  67.3%  3034  3014  3039  16 16  44.7%   87     14       0
04. GullChess 3.0 BMI2 x64        663.5/1000  429 469 102  66.3%  3026  3005  3031  16 16  46.9%   89      9       1
05. Equinox 3.30 x64              627.5/1000  398 459 143  62.7%  3000  2982  3002  16 16  45.9%   91      1       2
06. Critter 1.6a x64              587.0/1000  323 528 149  58.7%  2972  2954  2971  15 15  52.8%   91      7       2
07. Sting SF 4.8.4 x64            537.0/1000  278 518 204  53.7%  2939  2923  2933  15 15  51.8%   91      7       0
08. Protector 1.8.0 x64           527.0/1000  259 536 205  52.7%  2932  2919  2926  15 15  53.6%   90      6       1
09. Nirvanachess 2.1c POP x64     492.5/1000  237 511 252  49.2%  2909  2896  2901  15 15  51.1%   89      5       4
10. Hannibal 1.5 x64              490.0/1000  256 468 276  49.0%  2908  2894  2899  16 16  46.8%   92      1      10
11. Chiron 2.0 x64                480.0/1000  256 448 296  48.0%  2901  2891  2891  16 16  44.8%   88      5       6
12. Texel 1.05 x64                477.0/1000  248 458 294  47.7%  2899  2887  2889  16 16  45.8%   92      8       2
13. Naum 4.6 x64                  464.5/1000  218 493 289  46.5%  2891  2880  2880  15 15  49.3%   89      6       2
13. Andscacs 0.81 POP x64         464.5/1000  244 441 315  46.5%  2891  2878  2880  16 16  44.1%   88      9       8
15. Andscacs 0.72 POP x64         382.5/1000  171 423 406  38.2%  2833  2833  2833  16 17  42.3%   88      4       9
--------------------------------------------------------------------------------------------------------------------
15. Senpai 1.0 SSE42 x64          382.0/1000  166 432 402  38.2%  2835  2827  2818  16 16  43.2%   87      1      20
16. Hiarcs 14 WCSC w32            368.5/1000  148 441 411  36.9%  2825  2817  2808  16 16  44.1%   89      3      12
17. Junior 13.3.00 x64            353.0/1000  154 398 448  35.3%  2814  2807  2796  17 17  39.8%   84      9       7
18. Shredder 12 x64               352.5/1000  135 435 430  35.2%  2814  2808  2795  16 16  43.5%   87      1      24
19. Sjeng c't 2010 w32            333.0/1000  132 402 466  33.3%  2799  2794  2780  17 17  40.2%   87      2       8
20. Vajolet2 2.0 POP x64          332.0/1000  136 392 472  33.2%  2798  2792  2779  17 17  39.2%   88      3       9
21. Spike 1.4 Leiden w32          326.0/1000  125 402 473  32.6%  2794  2789  2774  17 17  40.2%   87      3      14

And now Hakkapeliitta ... around 325 to Komodo and Stockfish comes with the following stat (tomorrow available on my site, test-run is still running, latest 125 games):

No fast draw games ... very good!
Won up to move number 50 = 19x after 860 games!
Lost up to move number 50 = 1x (vs. Vajolet).

We should to compare all with Elo to the best engines today are available and find out ...

The correspondence players, opening book experts have a new and great analyze engine. I think after all what I saw ... Hakkapeliitta is most aggressive engine today is available.

A great moment for all of us ...

We have to 90% a new "Tactic Number 1" and the programmer should try that in the next versions this playing style will not lost for endgame Elo. We have enough strong engines for this playing phase.

In computer chess perhaps 325 to Komodo and Stockfish ... vs. Humans, I am sure Komodo and Stockfish are maybe 150-175 Elo stronger only, because in the playing phase humans have the advantage vs. comps Hakkapeliitta comes with the info ... grandmasters, international master ... not vs. me.

Best
Frank

Games can be found tomorrow on my webpage.
Hakkapelitta destroyed all others in the group of TOP-21 (not Stockfish and Komodo) up to move number 50 with great combinations.

[Frank Quisinsky]

Correction ... in tactic with many pieces on board.
Most interesting after the openings and most interesting in visiting eng-eng games from my point of view.

We are intelligent enough for looking today more on the different playing phases as on 1:0, 0:1 and 1/2:1/2 results we produced Elo.

Engines are more and more stronger and in this time it's much more interesting to looking on the different playing phases as on Elo only. Will give us more possibilites for analysing games or other things we like to do with engines.

Very big compliment to the programmer from Finland.
I am waiting no so many years of a Spark / Junior successor and have it with Alfil and Hakkapeliitta. Cheng is also on the way, Texel too ... and other modern engines, like Nirvana and Andscasc ... the style of this Group of engines is much more interesting as the 90er style Shredder have. You can see in the table I posted before ... what modern engine in the first playing phase do with Shredder!

Could be interesting to compare tactic test Suite results with this two engines with many pieces on the board.

Alfil (but the programmer have to work on the all the Errors) and Hakkapeliitta are clearly the engine developments of the year because the programmers gave us what we missed.

Best
Frank

[Frank Quisinsky]

Hi there,

the stats to middlegame / Transposition into endgame and endgame will be follow in the next hour.

But here is what I mean ...
Have a look on won up to move number 50 and lost up to move number 50. This is sensational aggressive for an engine 313 Ordo Elo weaker as the Number 1.

Best
Frank

Code: Select all

    Programs                      Score       1-0  =  0-1         EloS  Baye  Ordo  +   -  Draws  MoveØ won<50 lost<50
01. Komodo 9 x64                  783.5/1000  599 369  32  78.3%  3120  3096  3142  18 18  36.9%   87     13       2
02. Stockfish 26Apr2015 BMI2 x64  783.0/1000  592 382  26  78.3%  3119  3094  3142  17 17  38.2%   81     17       0
03. Fire 4 x64                    664.0/1000  434 460 106  66.4%  3020  3009  3035  16 16  46.0%   88     12       0
04. GullChess 3.0 BMI2 x64        658.0/1000  421 474 105  65.8%  3015  3004  3030  16 16  47.4%   89      8       1
05. Equinox 3.30 x64              621.0/1000  391 460 149  62.1%  2989  2981  3000  16 16  46.0%   90      1       4
06. Critter 1.6a x64              583.5/1000  320 527 153  58.4%  2963  2954  2972  15 15  52.7%   91      7       2
07. Sting SF 4.8.4 x64            534.0/1000  275 518 207  53.4%  2930  2924  2935  15 15  51.8%   91      7       1
08. Protector 1.8.0 x64           531.0/1000  268 526 206  53.1%  2928  2925  2933  15 15  52.6%   90      6       3
09. Nirvanachess 2.1c POP x64     487.5/1000  233 509 258  48.8%  2899  2896  2901  15 15  50.9%   89      4       5
10. Hannibal 1.5 x64              486.5/1000  253 467 280  48.6%  2898  2895  2900  16 16  46.7%   93      1      11
11. Chiron 2.0 x64                480.5/1000  256 449 295  48.0%  2894  2894  2896  16 16  44.9%   88      4       7
12. Texel 1.05 x64                466.5/1000  233 467 300  46.7%  2885  2883  2886  16 16  46.7%   92      7       3
13. Andscacs 0.81 POP x64         461.0/1000  245 432 323  46.1%  2881  2880  2881  16 16  43.2%   88      9       9
14. Naum 4.6 x64                  459.0/1000  213 492 295  45.9%  2880  2879  2880  15 15  49.2%   89      5       3
15. Hakkapeliitta 3.0 x64         390.0/1000  190 400 410  39.0%  2833  2834  2829  17 17  40.0%   88     23       2
16. Senpai 1.0 SSE42 x64          385.0/1000  172 426 402  38.5%  2830  2832  2825  16 16  42.6%   87      1      22
17. Hiarcs 14 WCSC w32            371.5/1000  151 441 408  37.1%  2820  2822  2814  16 16  44.1%   89      2      12
18. Junior 13.3.00 x64            354.0/1000  154 400 446  35.4%  2808  2811  2801  17 17  40.0%   84      8       7
19. Shredder 12 x64               346.5/1000  129 435 436  34.6%  2802  2808  2795  16 16  43.5%   86      1      24
20. Sjeng c't 2010 w32            336.0/1000  133 406 461  33.6%  2794  2799  2787  17 17  40.6%   87      2       9
21. Vajolet2 2.0 POP x64          318.0/1000  127 382 491  31.8%  2781  2786  2772  17 17  38.2%   87      4      15

[Frank Quisinsky]

I posted before the stats mate up to move number 50.
Now in my middlegame stat up to move number 60 ...

Now Hakkapeliitta isn't the number 1 ... here the number 3. So with more moves and lesser pieces on the board the engine lost the tactic power if we compare to the others.

But most interesting is "Opening Book analyses", tactic analyses with many pieces on the board and here I am sure ... we have a new number 1.

Best
Frank

And this is what I mean too ...
The different from 313 Elo to Komodo 9 is in reality 71 up to move number 60. Vs. very strong humans Hakkapeliitta will be a monster Opponent ... the new Shirov / Tal in computer chess.

Code: Select all

FCT League 2, 3.05.L1-03, games with move number 020-059
Elostat 1.3, EloØ = 2.909, Category = 26

    Programs                         Elo   +   - Games Score Av.Op. Draws   Won Draw Lost
01. Komodo 9 x64                   : 3049  39 37  145  71.4%  2890  51.7%    66   75    4
02. Stockfish 26Apr2015 BMI2 x64   : 3039  30 27  191  68.8%  2901  61.3%    73  117    1
03. Hakkapeliitta 3.0 x64          : 2978  37 35  127  61.4%  2897  63.0%    38   80    9
04. Fire 4 x64                     : 2971  30 28  167  59.9%  2901  68.3%    43  114   10
05. GullChess 3.0 BMI2 x64         : 2954  29 27  142  56.7%  2907  75.4%    27  107    8
06. Critter 1.6a x64               : 2953  26 24  148  55.7%  2913  79.1%    24  117    7
07. Sting SF 4.8.4 x64             : 2950  27 25  151  57.0%  2901  76.8%    28  116    7
08. Protector 1.8.0 x64            : 2949  26 24  162  55.9%  2908  77.2%    28  125    9
09. Texel 1.05 x64                 : 2920  34 34  129  51.6%  2909  67.4%    23   87   19
10. Chiron 2.0 x64                 : 2918  28 28  161  51.2%  2909  72.7%    24  117   20
11. Nirvanachess 2.1c POP x64      : 2912  26 26  166  50.9%  2906  76.5%    21  127   18
12. Equinox 3.30 x64               : 2911  30 30  135  48.9%  2918  74.1%    16  100   19
13. Naum 4.6 x64                   : 2887  28 29  153  45.8%  2916  73.2%    14  112   27
14. Junior 13.3.00 x64             : 2879  33 33  154  45.1%  2913  64.3%    20   99   35
15. Andscacs 0.81 POP x64          : 2872  33 33  154  45.1%  2905  64.3%    20   99   35
16. Sjeng c't 2010 w32             : 2852  31 33  131  40.5%  2919  70.2%     7   92   32
17. Hannibal 1.5 x64               : 2851  26 29  133  41.7%  2909  77.4%     4  103   26
18. Vajolet2 2.0 POP x64           : 2827  35 36  146  37.3%  2917  59.6%    11   87   48
19. Hiarcs 14 WCSC w32             : 2824  37 38  134  37.3%  2914  58.2%    11   78   45
20. Senpai 1.0 SSE42 x64           : 2794  33 35  161  34.2%  2908  58.4%     8   94   59
21. Shredder 12 x64                : 2788  32 33  186  31.7%  2921  54.8%     8  102   76

Games         : 1588
White Perf.   : 54.9%
Black Perf.   : 45.1%
White Wins    :  335 Games (21.1%)
Black Wins    :  179 Games (11.3%)
Draws         : 1074 Games (67.6%)

ECO A00-A99   :  263 Games (16.6%)
ECO B00-B99   :  330 Games (20.8%)
ECO C00-C99   :  334 Games (21.0%)
ECO D00-D99   :  353 Games (22.2%)
ECO E00-E99   :  308 Games (19.4%)

Stats to Transposition into endgame and endgame will be in the late evening available. No more time for chess.

But nice to see that my "early warning system" to my rating list works fine. That's what I like to see, not Elo only (to boring).

[shrapnel]

Your English is really weird.
I can't make head or tail of anything you write.
Do you use Google Translator ?
Just curious.

[Frank Quisinsky]

Sorry for my bad english!

I hope you will understand the easy statistics I posted before. Should be to understand for everyone. If not, I can't help and ... if you like ... you have to read the messages by the others.

Good luck and have a nice day!

Best
Frank

[Bloodbane]

What can I say, I am really thankful for this glowing praise, and also thankful for you running these tests. The middlegame statistics are pretty wonderful, since I've always wanted to make my engine the next Junior. On the other hand, it is no surprise to me that Hakkapeliitta is bad at endgames. I'll list all the endgame knowledge it has:

1. Syzygy tablebases
2. Endgame piece-square tables.
3. That's it.

Should be pretty easy to find something to add there.

[cdani]

What can I say, I am really thankful for this glowing praise, and also thankful for you running these tests. The middlegame statistics are pretty wonderful, since I've always wanted to make my engine the next Junior. On the other hand, it is no surprise to me that Hakkapeliitta is bad at endgames. I'll list all the endgame knowledge it has:

1. Syzygy tablebases
2. Endgame piece-square tables.
3. That's it.

Should be pretty easy to find something to add there.

It will be terrible strong then!!

[Frank Quisinsky]

Hi Daniel, Mikka,

oh yes, so many potential for a higher rating in Hakkapeliitta is. Very important is to hold the playing style in the early middlegame ... my biggest wish for ... I hope ... next available versions.

And to Daniel ...
Your engine do a wonderful job too ...
So many "newer" develops in the TOP-21 today.
Means I am looking many years again and again to the same engines in different versions. Now with Hakkapeliitta, Andscacs, Senpai, NirvanaChess, StingSF, Texel included in TOP-21 and all is much more interesting to follow.

Different others are on the way.
Very interesting in Alfil next version, hope of Fizbo with a Shredder GUI bugfix. Deuterium, Cheng, ICE, Vajolet2, others are on the way ... will be a hard fight!

Best
Frank

[neelbasant]

Really Good in middle game.

But problem is it is SINGLE PROCESSOR.

NEEL