Last Update: 20.09.2022
Dear Chess Friends,
I am pleased to announce that,I managed to organize another NNUE competition!Meanwhile, periodically I run similar tours: Results
Well, let Me say please some words about this tour,1st of all, no doubt that all of them are extremely strong!Franky, especially Top 8 ones are quite close in strengthAnd once more, many thanks to all engine programmers!
But reality is reality..without to not mention this I can't,What a pity that some of the engines are buggy, e.g:SF Polyglot 130922 is out, simply due to not capable to Play 4 Cores/Game (2 Cores per side), really strange...Same Goi's engine error appears under Cutechess too
On other hand, this is not all...Unfortunately Nebula-Vortex 1.0 engine is buggy too...E.g I've noticed many Nebula's games as lost on time...Btw, those buggy games are already edited (fixed...)Otherwise, Nebula would be ranked ,rated much lower...To be more clear (without editing those wrong results).The current standings would be not accurate for sure!And it seems these time losses are happened due toNebula-Vortex 1.0 has no any Move Overhead option...Where the other rest engines are played with M.O 200And theirs all played games are ended successfully...I mean the rest players include no games lost on time
And once more:Please check more carefully your work (before online...)..The eng stability should be as most important part...right?.And I hope all these bugs will be fixed in the next releases...Thanks in advance..by the way, I have to admit also that,Mostly latest SF based engines became more stable...e.gIn one of older competitions, some lost on time: Results
As participants (note that one engine per author is allowed):I've picked Top 8 engines of recent organized tour: ResultsPlus as guest NNUE engine, Raubfish SC is participated tooThe reason, to see Blitz Elo improvement (since Jan 2022)As we see, now is more clear..the difference is about 25 EloThat means, minus plus 25 Elo improvement (in 9 months)Btw, via fast Bullet TC: the Elo improvement is about 35 EloAnd if comparing Raubfish SC's Blitz draw ratio is just: 87%Unbelievable stats..approx. 11% less than current engines!Yes.., no any other engine to be so strong with such draws!Plus Raubfish SC playing itself...for anyone missed: Results
And here are the final standings:The Winner: Polyfish - Congratulations to Khalid Omar, SF team!
# PLAYER RATING POINTS GAMES (%) 1 Polyfish 220826 : 3754.6 220.5 432 51.0%2 CorChess 170922 : 3753.2 219.5 432 50.8%3 Nebula-Vortex 1.0 : 3753.2 219.5 432 50.8%4 SF PB 130722 : 3751.0 218.0 432 50.5%5 Stockfish 170922 : 3749.6 217.0 432 50.2%6 Kayra 1.5 : 3749.6 217.0 432 50.2%7 Swordfish 15.3a : 3748.9 216.5 432 50.1%8 ProteusSF-P 040922 : 3748.1 216.0 432 50.0%9 Raubfish X48c3 SC : 3725.0 200.0 432 46.3%
• Calculated by Ordo chess program / Raubfish engine is fixed to 3725 Elo
More Details:• All engines are played with Perferct2022 (Book Learning ON) So for each player is installed separate copy of opening book • The overall draw-ratio (based on 1944 games) very high 95% Btw without Raubfish SC: the overall-draws are close to 98% And can we say that chess in danger (under these cond.) ?• Download all played tournament games (in PGN): CS XVIII
Best,Sedat
Dear Chess Friends,
I am pleased to announce that,I managed to organize another NNUE competition!Meanwhile, periodically I run similar tours: Results
Well, let Me say please some words about this tour,1st of all, no doubt that all of them are extremely strong!Franky, especially Top 8 ones are quite close in strengthAnd once more, many thanks to all engine programmers!
But reality is reality..without to not mention this I can't,What a pity that some of the engines are buggy, e.g:SF Polyglot 130922 is out, simply due to not capable to Play 4 Cores/Game (2 Cores per side), really strange...Same Goi's engine error appears under Cutechess too
On other hand, this is not all...Unfortunately Nebula-Vortex 1.0 engine is buggy too...E.g I've noticed many Nebula's games as lost on time...Btw, those buggy games are already edited (fixed...)Otherwise, Nebula would be ranked ,rated much lower...To be more clear (without editing those wrong results).The current standings would be not accurate for sure!And it seems these time losses are happened due toNebula-Vortex 1.0 has no any Move Overhead option...Where the other rest engines are played with M.O 200And theirs all played games are ended successfully...I mean the rest players include no games lost on time
And once more:Please check more carefully your work (before online...)..The eng stability should be as most important part...right?.And I hope all these bugs will be fixed in the next releases...Thanks in advance..by the way, I have to admit also that,Mostly latest SF based engines became more stable...e.gIn one of older competitions, some lost on time: Results
As participants (note that one engine per author is allowed):I've picked Top 8 engines of recent organized tour: ResultsPlus as guest NNUE engine, Raubfish SC is participated tooThe reason, to see Blitz Elo improvement (since Jan 2022)As we see, now is more clear..the difference is about 25 EloThat means, minus plus 25 Elo improvement (in 9 months)Btw, via fast Bullet TC: the Elo improvement is about 35 EloAnd if comparing Raubfish SC's Blitz draw ratio is just: 87%Unbelievable stats..approx. 11% less than current engines!Yes.., no any other engine to be so strong with such draws!Plus Raubfish SC playing itself...for anyone missed: Results
And here are the final standings:The Winner: Polyfish - Congratulations to Khalid Omar, SF team!
# PLAYER RATING POINTS GAMES (%) 1 Polyfish 220826 : 3754.6 220.5 432 51.0%2 CorChess 170922 : 3753.2 219.5 432 50.8%3 Nebula-Vortex 1.0 : 3753.2 219.5 432 50.8%4 SF PB 130722 : 3751.0 218.0 432 50.5%5 Stockfish 170922 : 3749.6 217.0 432 50.2%6 Kayra 1.5 : 3749.6 217.0 432 50.2%7 Swordfish 15.3a : 3748.9 216.5 432 50.1%8 ProteusSF-P 040922 : 3748.1 216.0 432 50.0%9 Raubfish X48c3 SC : 3725.0 200.0 432 46.3%
• Calculated by Ordo chess program / Raubfish engine is fixed to 3725 Elo
More Details:• All engines are played with Perferct2022 (Book Learning ON) So for each player is installed separate copy of opening book • The overall draw-ratio (based on 1944 games) very high 95% Btw without Raubfish SC: the overall-draws are close to 98% And can we say that chess in danger (under these cond.) ?• Download all played tournament games (in PGN): CS XVIII
Best,Sedat