Sting SF 15 64-bit Gauntlet for CCRL 40/40

[Graham Banks]

PGN - http://kirill-kryukov.com/chess/discuss ... p?id=44743

[Graham Banks]

Code: Select all

CCRL 40/40 Rating List - Custom engine selection
1065281 games played by 2493 programs, run by 22 testers
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 40 minutes on Athlon 64 X2 4600+ (2.4 GHz), about 15 minutes on a modern Intel CPU.
Computed on July 6, 2019 with Bayeselo based on 1'065'281 games
Tested by CCRL team, 2005-2019, http://computerchess.org.uk/ccrl/4040/

Rank               Engine                Elo   +    -   Score  AvOp  Games
1 Sting SF 4.7 64-bit                3065  +22  -22  46.2%  +21.1   599
  Sting SF 4.8.4 64-bit              3064  +27  -27  47.4%  +14.9   430
  Stockfish 2.1.1 64-bit             3053  +15  -15  51.2%   -6.9  1258
  Sting SF 4.6                       3047  +30  -30  45.4%  +29.1   334
  Sting SF 5 64-bit                  3040  +18  -18  44.7%  +33.2   938
  Sting SF 8.5 64-bit                3040  +29  -29  49.1%   +3.9   345
  Sting SF 8.1 64-bit                3039  +28  -28  54.1%  -25.6   391
  Sting SF 3 64-bit                  3038  +16  -16  47.8%  +11.6  1195
  Sting SF 9.99 64-bit               3037  +29  -29  50.3%   -4.8   350
  Sting SF 6.7 64-bit                3035  +28  -28  48.9%   +8.5   367
  Sting SF 15 64-bit                 3031  +26  -26  51.4%   -7.3   432
  Sting SF 4.7                       3028  +22  -22  44.5%  +32.8   632
  Sting SF 4 64-bit                  3026  +18  -18  49.1%   +3.2   915
  Sting SF 2 64-bit                  3022  +18  -18  50.1%   -1.4   943
  Sting SF 6.1 64-bit                3017  +30  -30  47.6%  +15.2   332
  Sting SF 10.1 64-bit               3013  +28  -28  48.3%   +7.8   410
  Sting SF 9.9 64-bit                3009  +17  -17  42.7%  +47.5  1109
  Sting SF 7.3 64-bit                3006  +31  -31  51.3%   -3.6   304
  Sting SF 2 32-bit                  3003  +17  -17  51.3%  -10.1  1062
  Sting SF 9 64-bit                  2995  +17  -17  43.3%  +51.3  1089

[lech]

Thanks Graham and all testers!
It seems to be a good result and I hope the next versions will keep it.
It is really very important to know the ELO difference between newest Sting 15 and old Stockfish 2.1.1 .

[Kotlov]

Sting SF 15 64-bit 3031
Sting SF 2 64-bit 3022

No ideas to improve?

[Dann Corbit]

He does not aim for Elo improvement.
He aims for problem analysis improvement.
They are not the same goal.

[Kotlov]

He does not aim for Elo improvement.
He aims for problem analysis improvement.
They are not the same goal.

What's the problem?

[carldaman]

The "problems" are usually endgame studies with deep solutions, often involving zugzwang. This is where Sting excels, its raison d'etre, if you will.

[Dann Corbit]

He does not aim for Elo improvement.
He aims for problem analysis improvement.
They are not the same goal.

What's the problem?

Sting comes with a database of EPD positions.
Try to solve them with Stockfish, then try to solve them with Sting.
The author has examined certain categories of chess problems where Stockfish suffers.
He wrote a version of Stockfish that can solve these puzzles.
Doing this make Stockfish a bit weaker in the general case, but much stronger for these special kinds of positions.

[Kotlov]

ah... ok