Release : Chess System Tal EXTREME

[Peter Berger]

I don't have a Junior engine, but if you have I can give you a setup to play a few thousand Junior games yourself.

Thanks. My Junior engines are on DVDs though that I don't know how to access easily. I'll get back to you.

[BrendanJNorman]

Back in the days we used to romanticize about Junior's attacking ability/aggressiveness, and I indeed remember the 2002 (?) match with Deep Fritz fondly.

Thinker was another engine with super aggressive play.

I would love to see an NNUE version of either of them. Or even old classics like Phalanx, Comet, Crafty, WChess, Baron etc would be interesting as NNUEs just to see what comes of the playing style.

Anyway, congrats to Chris and Ed, I haven't been here for a while, but look forward to testing your latest chess monsters soon.

Thanks for all the hard work.

[pohl4711]

Not only Rebel EAS 3.0 scored 150.000 style points more than 2.0, surprisingly it also is considerable stronger than the Patricia's. And thus I changed my mind.

What does "And thus I changed my mind." mean here?

Personally I would be very much interested in engines that find a good compromise between EAS score and playing strength.

My personal very first impression with Chess System Tal extreme ( still way too little games so far here) is a bit similar to that of Patricia 5.0 extreme. It feels a little bit overtuned when looking at the games. At times they both seem to make pretty random sacs that are just good for scoring some additional EAS points but don't feel too natural. ELO might be useful for balance here.

If you pit them against older engines, their main strength is that they can outsearch them, even if they took the saccing too far.

I would be interested to see some EAS values for the later Junior engines. If style programming becomes a thing, they should be remembered as being at the top strength-wise while regularly doing real positional sacs at the same time.

This can easily be measured, if there are enough games. Because the EAS-Tool adjusts its short wins bonusses automatically to the input gamebase, it also works fine with much weaker engines (the weaker the engines, the shorter the average length of the games).
So, if the old Junior HCE plays enough games in a test, an EAS-scoring can easily be done. So, take Junior HCE, and make it play some thousand games vs. 3-5 opponents, roughly at the the same Celo-level...
Here (as an example) the EAS-list of a lot of TheKing-Settings, where I added the superweak OpenTal engine just for fun:

Code: Select all

                         early           bad  avg.win 
Rank  EAS-Score  sacs    sacs   shorts  draws  moves  Engine/player 
---------------------------------------------------------------------------
   1    865262  66.67%  16.48%  49.37%  04.20%   44   Open Tal 1.2  
   2    452369  46.75%  13.88%  39.05%  05.91%   47   TK Aggressive  
   3    349958  40.39%  13.63%  39.55%  17.27%   46   TK TrS Aggressive  
   4    271731  31.83%  11.96%  31.83%  14.84%   50   TK Native AkAg  
   5    256183  31.50%  13.46%  27.43%  10.84%   52   TK Active  
   6    251699  34.92%  09.88%  31.72%  16.81%   50   TK TrS AkAg  
   7    204223  22.61%  11.15%  28.61%  19.87%   51   TK Nat AkAg_Fig120  
   8    185966  22.95%  10.71%  24.59%  14.88%   54   TK TrS Active  
   9    184108  22.58%  08.56%  27.77%  19.41%   52   TK Nativ AkAg 20sub  
  10    149308  20.24%  08.30%  19.10%  13.20%   56   TK TrS Normal  
  11    146322  18.34%  10.89%  21.45%  14.99%   56   TK Beth Harmon  
  12    145222  18.68%  07.00%  19.52%  13.37%   55   TK sel20 Researcher  
  13    143132  17.23%  03.93%  19.49%  13.11%   55   TK Researcher  
  14    141526  17.17%  08.73%  18.89%  13.91%   57   TK Prometheus  
  15    135332  14.86%  10.40%  21.12%  16.65%   56   TK se 20 Razorback  
  16    133701  14.84%  07.18%  18.87%  13.36%   57   TK Razorback  
  17    132437  15.72%  07.94%  19.13%  15.64%   57   TK Normal  
  18    128602  13.11%  05.30%  18.99%  17.47%   57   Gandalf 7  
  19    117354  15.26%  08.07%  14.79%  15.17%   59   TK Solid  
  20    114303  12.65%  07.45%  19.15%  19.25%   56   K2 0.95  
  21    110830  12.65%  08.96%  14.45%  14.93%   62   TK TrS Solid  
  22    110344  14.13%  08.33%  22.46%  25.51%   55   TK SPCC Active  
  23     96414  08.57%  12.50%  16.49%  19.96%   57   TK SPCC Normal  
  24     95859  09.05%  02.04%  16.70%  17.74%   55   RedQueen 1.1.98  
  25     94059  08.19%  08.05%  17.14%  21.51%   57   TK SchroderAktiv  
  26     92944  08.88%  05.34%  18.10%  21.40%   55   Delfi 5.4  
  27     92296  07.26%  17.57%  14.33%  19.99%   59   TK SPCC Solid  
  28     80627  08.73%  08.05%  13.94%  20.86%   58   TK Schroeder  
  29     79060  06.53%  01.78%  17.19%  25.67%   56   Rebel 13  
  30     76751  09.43%  02.07%  16.25%  28.76%   59   Orion 0.6  
  31     63221  04.94%  03.54%  14.39%  29.54%   58   Ruffian Leiden  
-------------------------------------------------------------------
*** Average length of all won games:             55 moves
*** Movelimit for early sac bonus  :             27 moves

Interesting: The old HCE engines all have a much worse early-sac stat, compared to the new aggressive nnue-engines.

[Rebel]

I don't have a Junior engine, but if you have I can give you a setup to play a few thousand Junior games yourself.

Thanks. My Junior engines are on DVDs though that I don't know how to access easily. I'll get back to you.

Actually it is more simple, downloaded the Junior games from CCRL. Also the games of Fritz 10, evaluation wise the best Fritz IMO.

Code: Select all

 STYLE   King          Short           
 Total  Attack   %     Games   %     Sacs    %    Engine
145879   71800 17.0%    5408 02.8%   68671 29.3%  Junior 10  
133198   82707 19.6%    5881 03.5%   44610 19.6%  Junior 9  
120947   66150 15.4%    4278 02.6%   50519 23.7%  Junior 10.1  
109212   63197 15.6%    1082 01.8%   44933 21.6%  Fritz 10  

One game

[pgn][Event "Best of Chess"]
[Site "King-Attacks, Sacrifices, Short Games"]
[Date "2006.02.05"]
[Round "?"]
[White "Junior 9"]
[Black "Chess Tiger 15"]
[Result "1-0"]
[PlyCount "79"]
[King "18024"]
[Short "500"]
[Sac "0"]
[Total "18524"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Be3 e6 7. f3 b5 8.
Qd2 Nbd7 9. g4 h6 10. O-O-O b4 11. Nce2 d5 12. exd5 Nxd5 13. Nf4 Bb7 14.
Bc4 N7b6 15. Ndxe6 fxe6 16. Bxb6 Nxb6 17. Qe3 Qf6 18. Bxe6 Bc5 19. Qxc5
Qxf4+ 20. Kb1 Bxf3 21. Bf5 Kf7 22. Qxb6 Rhe8 23. Rd7+ Kf8 24. Qg6 Re7 25.
Rxe7 Kxe7 26. Qxg7+ Kd6 27. Re1 Kc5 28. Re6 a5 29. Qe7+ Kd4 30. c3+ bxc3
31. Qg7+ Kc5 32. Qxc3+ Qc4 33. Qa3+ Qb4 34. Qxf3 Qe1+ 35. Rxe1 Re8 36. Rc1+
Kd6 37. Qc6+ Ke5 38. Qxe8+ Kf4 39. Rf1+ Kg5 40. Qe7# 1-0[/pgn]

[Rowen]

Not only Rebel EAS 3.0 scored 150.000 style points more than 2.0, surprisingly it also is considerable stronger than the Patricia's. And thus I changed my mind.

What does "And thus I changed my mind." mean here?

Personally I would be very much interested in engines that find a good compromise between EAS score and playing strength.

My personal very first impression with Chess System Tal extreme ( still way too little games so far here) is a bit similar to that of Patricia 5.0 extreme. It feels a little bit overtuned when looking at the games. At times they both seem to make pretty random sacs that are just good for scoring some additional EAS points but don't feel too natural. ELO might be useful for balance here.

If you pit them against older engines, their main strength is that they can outsearch them, even if they took the saccing too far.

I would be interested to see some EAS values for the later Junior engines. If style programming becomes a thing, they should be remembered as being at the top strength-wise while regularly doing real positional sacs at the same time.

This can easily be measured, if there are enough games. Because the EAS-Tool adjusts its short wins bonusses automatically to the input gamebase, it also works fine with much weaker engines (the weaker the engines, the shorter the average length of the games).
So, if the old Junior HCE plays enough games in a test, an EAS-scoring can easily be done. So, take Junior HCE, and make it play some thousand games vs. 3-5 opponents, roughly at the the same Celo-level...
Here (as an example) the EAS-list of a lot of TheKing-Settings, where I added the superweak OpenTal engine just for fun:

Code: Select all

                         early           bad  avg.win 
Rank  EAS-Score  sacs    sacs   shorts  draws  moves  Engine/player 
---------------------------------------------------------------------------
   1    865262  66.67%  16.48%  49.37%  04.20%   44   Open Tal 1.2  
   2    452369  46.75%  13.88%  39.05%  05.91%   47   TK Aggressive  
   3    349958  40.39%  13.63%  39.55%  17.27%   46   TK TrS Aggressive  
   4    271731  31.83%  11.96%  31.83%  14.84%   50   TK Native AkAg  
   5    256183  31.50%  13.46%  27.43%  10.84%   52   TK Active  
   6    251699  34.92%  09.88%  31.72%  16.81%   50   TK TrS AkAg  
   7    204223  22.61%  11.15%  28.61%  19.87%   51   TK Nat AkAg_Fig120  
   8    185966  22.95%  10.71%  24.59%  14.88%   54   TK TrS Active  
   9    184108  22.58%  08.56%  27.77%  19.41%   52   TK Nativ AkAg 20sub  
  10    149308  20.24%  08.30%  19.10%  13.20%   56   TK TrS Normal  
  11    146322  18.34%  10.89%  21.45%  14.99%   56   TK Beth Harmon  
  12    145222  18.68%  07.00%  19.52%  13.37%   55   TK sel20 Researcher  
  13    143132  17.23%  03.93%  19.49%  13.11%   55   TK Researcher  
  14    141526  17.17%  08.73%  18.89%  13.91%   57   TK Prometheus  
  15    135332  14.86%  10.40%  21.12%  16.65%   56   TK se 20 Razorback  
  16    133701  14.84%  07.18%  18.87%  13.36%   57   TK Razorback  
  17    132437  15.72%  07.94%  19.13%  15.64%   57   TK Normal  
  18    128602  13.11%  05.30%  18.99%  17.47%   57   Gandalf 7  
  19    117354  15.26%  08.07%  14.79%  15.17%   59   TK Solid  
  20    114303  12.65%  07.45%  19.15%  19.25%   56   K2 0.95  
  21    110830  12.65%  08.96%  14.45%  14.93%   62   TK TrS Solid  
  22    110344  14.13%  08.33%  22.46%  25.51%   55   TK SPCC Active  
  23     96414  08.57%  12.50%  16.49%  19.96%   57   TK SPCC Normal  
  24     95859  09.05%  02.04%  16.70%  17.74%   55   RedQueen 1.1.98  
  25     94059  08.19%  08.05%  17.14%  21.51%   57   TK SchroderAktiv  
  26     92944  08.88%  05.34%  18.10%  21.40%   55   Delfi 5.4  
  27     92296  07.26%  17.57%  14.33%  19.99%   59   TK SPCC Solid  
  28     80627  08.73%  08.05%  13.94%  20.86%   58   TK Schroeder  
  29     79060  06.53%  01.78%  17.19%  25.67%   56   Rebel 13  
  30     76751  09.43%  02.07%  16.25%  28.76%   59   Orion 0.6  
  31     63221  04.94%  03.54%  14.39%  29.54%   58   Ruffian Leiden  
-------------------------------------------------------------------
*** Average length of all won games:             55 moves
*** Movelimit for early sac bonus  :             27 moves

Interesting: The old HCE engines all have a much worse early-sac stat, compared to the new aggressive nnue-engines.

Is it possible to get similar EAS data for the skill levels with Patricia 5.0 extremeEAS? I have played quite a few handicapped games with it and other Patricia versions (long games) using skill levels of about 16-18 and it doesn't seem particularly aggressive if anything it seems quite solid? Apart from being weaker would the skill levels affect style such as aggressiveness?

Thanks to the programmers for releasing chess system Tal extreme, it is a fun engine to play against, it can be handicapped using it in the Fritz GUI 'rated games' option. Not sure how, perhaps depth is reduced and maybe not ideal? but at least the engine can be played against.