Last Update: 20.06.2023
Hello Chess Friends,
1st of all, This time, I've realized to test the Top engines, which are publicly available... And only those ones which are capable to learn (sure during game-play...) Otherwise, the test conditions will be not so fair..if we will allow all engines..
As current players, I picked the engines which performed not so bad in my recent competitions! Btw, instead of Brainlearn 25, I picked older vers, the reason: Brainlearn 25 Generated no any exp file, I mean via Seld-Q-Learning enabled, where with Same settings, Brainlearn 24 is worked flawlessly (experience is created...) Plus as older version, I used LittleBeast 2, due to it seems slightly stronger...
On other hand, Eman is managed to be Number One under Blitz conditions..actually no BIG Surprise... e.g Mr. Khalid is one of the best programmers in engine learning!
What I can say more (for anyone missed), Unfortunately RaptoRa 2.2 is a private chess engine... so it is not allowed... But the good news is that, Eman, LittleBeast, BrainLearn etc.are available! Note also that current Sugar ver is available too, but after purchasing in Outskirts..
NNUE XXXIV's Winner: EMAN - Congratulations to Khalid Omar, SF team!
Rank Name Elo + - games score oppo. draws 1 Eman 8.91 3776 7 7 800 51% 3770 87% 2 LittleBeast 2 sl 3773 7 7 800 50% 3771 88% 3 BrainLearn 24 SQL 3770 7 7 800 50% 3772 89% 4 SugaR XPrO 240523 3770 7 7 800 50% 3772 88% 5 DarkSisTer 5 3768 7 7 800 49% 3772 87%
• Calculated by Bayeselo / Sugar is fixed to 3770 Elo / SQL = Self-Q-Learning ON
More Details: • All of the NNUE engines are played with no any pre-learned exp files.. But note: during game-play, their eng learning functions were enabled In same time that means, current engines are generated own exp files • And all of the engines are AVX2, tested with theirs default Evalfiles • As book, all used to play Unique suite (exactly same 100 openings) Plus for fair conditions: each of the opening line is repeated as twice • And as usual, all of the engines are played via move overhead: 100 • The overall draw percentage (based on 2000 games) is not high: 88% • All the played tour games are available (in PGN format): NNUE XXXIV
Hello Chess Friends,
1st of all, This time, I've realized to test the Top engines, which are publicly available... And only those ones which are capable to learn (sure during game-play...) Otherwise, the test conditions will be not so fair..if we will allow all engines..
As current players, I picked the engines which performed not so bad in my recent competitions! Btw, instead of Brainlearn 25, I picked older vers, the reason: Brainlearn 25 Generated no any exp file, I mean via Seld-Q-Learning enabled, where with Same settings, Brainlearn 24 is worked flawlessly (experience is created...) Plus as older version, I used LittleBeast 2, due to it seems slightly stronger...
On other hand, Eman is managed to be Number One under Blitz conditions..actually no BIG Surprise... e.g Mr. Khalid is one of the best programmers in engine learning!
What I can say more (for anyone missed), Unfortunately RaptoRa 2.2 is a private chess engine... so it is not allowed... But the good news is that, Eman, LittleBeast, BrainLearn etc.are available! Note also that current Sugar ver is available too, but after purchasing in Outskirts..
NNUE XXXIV's Winner: EMAN - Congratulations to Khalid Omar, SF team!
Rank Name Elo + - games score oppo. draws 1 Eman 8.91 3776 7 7 800 51% 3770 87% 2 LittleBeast 2 sl 3773 7 7 800 50% 3771 88% 3 BrainLearn 24 SQL 3770 7 7 800 50% 3772 89% 4 SugaR XPrO 240523 3770 7 7 800 50% 3772 88% 5 DarkSisTer 5 3768 7 7 800 49% 3772 87%
• Calculated by Bayeselo / Sugar is fixed to 3770 Elo / SQL = Self-Q-Learning ON
More Details: • All of the NNUE engines are played with no any pre-learned exp files.. But note: during game-play, their eng learning functions were enabled In same time that means, current engines are generated own exp files • And all of the engines are AVX2, tested with theirs default Evalfiles • As book, all used to play Unique suite (exactly same 100 openings) Plus for fair conditions: each of the opening line is repeated as twice • And as usual, all of the engines are played via move overhead: 100 • The overall draw percentage (based on 2000 games) is not high: 88% • All the played tour games are available (in PGN format): NNUE XXXIV