Amoeba 2.1 64-bit Gauntlet for CCRL 40/40

[Graham Banks]

PGN - http://kirill-kryukov.com/chess/discuss ... p?id=39150

[Graham Banks]

Code: Select all

CCRL 40/40 Rating List - Custom engine selection
693335 games played by 1834 programs, run by 19 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 February 25, 2017 with Bayeselo based on 693'335 games
Tested by CCRL team, 2005-2017, http://computerchess.org.uk/ccrl/4040/

Rank              Engine                Elo   +    -   Score  AvOp  Games
1 Amoeba 2.0 64-bit                 2729  +29  -29  45.0%  +30.2   372
  Amoeba 2.1 64-bit                 2707  +33  -33  46.3%  +25.1   294
  Amoeba 1.4 64-bit                 2706  +27  -27  46.1%  +27.6   451
  Amoeba 1.3 64-bit                 2692  +33  -33  48.5%   +9.4   298
  Amoeba 1.2 64-bit                 2611  +31  -31  56.1%  -41.2   335

[abulmo2]

Code: Select all

Rank              Engine                Elo   +    -   Score  AvOp  Games
1 Amoeba 2.0 64-bit                 2729  +29  -29  45.0%  +30.2   372
  Amoeba 2.1 64-bit                 2707  +33  -33  46.3%  +25.1   294
  Amoeba 1.4 64-bit                 2706  +27  -27  46.1%  +27.6   451

Thank you for running the tests.
On my own "long" time tests (with times similar to your 40/5, not 40/40), I do not observe a regression for the 2.1 version. It is possible that amoeba 2.1 does not scale well. Whatever, I shall try to release next versions with a higher elo gap, so that everybody can observe a progression. As until now I only corrected a few bugs with little elo impact, not sure this will happens in version 2.2 though.