Last Update: 21.11.2022
Dear Chess Friends,
As usual, I am pleased to announce that,I organized another new NNUE competition!
1st of all,All NNUE engines are tested via new 50 openings!I named it as 'Unique' because is different than all..It's based mainly on Balsa and Perfect book series Also it includes some openings from SCCT Book ToursNote also that this is as 1st ever NNUE championship,Which is organized under these new Blitz conditions!Because as far as I am aware, no similar test is seen..I mean about such tour with similar optimized openings!In case of interest, Unique suite is available DownloadHave fun and happy testings !)
Continuing...and to be more clear:
Unique suite is tuned for both sides (White and Black),Which have more equal chances to win... By the way, Chess is a game, which is playing by two players, right? In short, would not be better if the openings to be strong For both sides? E.g why we have to force with unusual Weak openings, which favorite just for White or Black?And based on my older testings (via Unique openings):Many NNUE engines (3700+) managed already to win By both sides...in short, I can say over as 'golden' key!I mean no any other suite is created in same method..E.g exactly with same opening (3700+) to win as WhiteOr Black...yes...this is a new idea and I'm pleased toShare with all of you and feel the difference in SCCT!
And a few more words about Unique openings,The overall draw percentage is less than Balsa..But I admit also that Unique is more critical, weaker.. Anyhow the most important, if strength measurements:Unique opening suite seems to be not so bad... andSure instead of comments, I prefer to relay on ResultsYes.. in this way, now you have chances to compare E.g Unique openings are far away from reality or not...In other words, no any other test suite to be so strong..And in same time to produce low draw percentages!
You may know too, Many of opening test suites are based mainly on HumanOpenings..that's sounds as good and as usual less draws..But be aware of that too, just in case of 3500+ eng testingsThen you need thousands of games (per player). otherwise,The error margin can be too high..and frankly even though,No any guarantee...for this I suggest different test suites...But via strong openings: 500-1000 games is enough data,Regarding to determine any chess engine strength
And without to not mention this I can't and once more: Since NNUE on scene, better quality games for sure!Anyhow, situation has been changed.. (not as before)E.g overall draw percentage is dramatically increased ..That means also (in same time), it's too boring chess!
On other hand, it will be a BIG mistake, if we will start toBlame the strongest opening theory...why I say like this,Because the high draw percentage is related mainly to Engine strength, sure the short openings play influence too...But main reason of high draws are engines (not openings)!E.g to be sure, simply you can check LXIV / Human-LikeWhere perfectly clear that the draw ratio is much different..Be aware, for both:I used exactly same books, exc. enginesAnd these both testings are played as NON - NNUE e.g via Strong Evalfiles are expecting 90-95% draws, this should beKnown too, yes. 'papers' and 'reality' are two different things!In other words, I can summarize it (to appear many wins):- No much chances by strong openings!- BIG chances by weak chess engines!
Btw, if still not so clear or for anyone missed, interested..I suggest to check my played opening testings: ResultsMeanwhile, since many years (I think that since 2014)Balsa plus Perfect book series are based on only TopEngine games....and almost all other well-tuned deep books are based on engine games as well..and why?The answer is very simple:We test mainly Top engines, actually it's a long story..But in short, we should not mix apples with oranges!)Hopes helps...
After all,It's up to you, anyone is free to prefer any conditions,But just don't blame/criticize the strongest openings!And don't try to fool us that the weak lines are best!Because SCCT is born for this as well, more Details
And here are the final standings, The Winner FisherovMonk - Congratulations to Fisherov, Stockfish team!
Rank Name Elo + - Games Score Oppo. Draws 1 FisherovMonk 1.2 3762 5 5 1700 51% 3753 90% 2 Zeus 26.5 3761 5 5 1700 51% 3754 90% 3 Kookaburra 1.01 3761 5 5 1700 51% 3754 91% 4 SolistaAttack v2 3760 5 5 1700 51% 3751 91% 5 Polyfish 091022 3760 5 5 1700 51% 3754 89% 6 Solidor 121122 3759 5 5 1700 52% 3748 89% 7 Kayra 1.7 3758 5 5 1700 51% 3752 91% 8 StockfishPB 290922 3758 5 5 1700 51% 3754 89% 9 Stockfish 071122 3756 5 5 1700 51% 3751 91% 10 BrainLearn 20 3754 5 5 1600 50% 3753 91% 11 Cool Iris 5.2 3753 5 5 1600 51% 3750 90% 12 YuliGM PrO 4 3753 5 5 1700 51% 3748 85% 13 ProteusSF-P 040922 3753 5 5 1600 50% 3753 89% 14 Swordfish 15.3a Goi 3752 5 5 1700 51% 3750 87% 15 Stockfish 15 3750 5 5 1600 50% 3749 87% 16 TealOracle 3749 5 5 1600 51% 3743 86% 17 DeepBlue 180422 3742 5 5 1500 50% 3742 85% 18 Big Ballz 010922 3736 6 6 1200 50% 3737 85% 19 Fat Titz 2 3734 6 6 1200 49% 3740 84% 20 Honey 14.1.01 3733 6 6 1200 49% 3737 86% 21 Crystal 5 KWK 3720 6 5 1200 47% 3739 84% 22 Cfish 130821 3706 5 5 1200 44% 3741 83% 23 Spider 080621 3688 5 5 1200 40% 3741 74%
• Calculated by Bayeselo / SF15 is fixed to 3750 Elo / Blue Name = Last Update
More Details: • Mostly of the engines BMI2, tested via theirs default Evalfiles Exception, Zeus v25.6 is played, the reason v27 is not BMI2 And Zeus is used to play via nn.zeus (based on13.07.2022) • Plus some engines are out, simply due to they are very close In similarities to others.. e.g the target is between 68%-74% • All NNUE engines played with a new 'Unique' opening suite Note also that, each of the opening line is playing as 'twice' • The overall draw percentage (based on 17700 games) 88% • Some older (Stockfish 15 etc.) are tested just for comparing And it's clear that no much Elo progress since many months • All played tournament games are available: NNUE CS XXIII
Best, Sedat
Dear Chess Friends,
As usual, I am pleased to announce that,I organized another new NNUE competition!
1st of all,All NNUE engines are tested via new 50 openings!I named it as 'Unique' because is different than all..It's based mainly on Balsa and Perfect book series Also it includes some openings from SCCT Book ToursNote also that this is as 1st ever NNUE championship,Which is organized under these new Blitz conditions!Because as far as I am aware, no similar test is seen..I mean about such tour with similar optimized openings!In case of interest, Unique suite is available DownloadHave fun and happy testings !)
Continuing...and to be more clear:
Unique suite is tuned for both sides (White and Black),Which have more equal chances to win... By the way, Chess is a game, which is playing by two players, right? In short, would not be better if the openings to be strong For both sides? E.g why we have to force with unusual Weak openings, which favorite just for White or Black?And based on my older testings (via Unique openings):Many NNUE engines (3700+) managed already to win By both sides...in short, I can say over as 'golden' key!I mean no any other suite is created in same method..E.g exactly with same opening (3700+) to win as WhiteOr Black...yes...this is a new idea and I'm pleased toShare with all of you and feel the difference in SCCT!
And a few more words about Unique openings,The overall draw percentage is less than Balsa..But I admit also that Unique is more critical, weaker.. Anyhow the most important, if strength measurements:Unique opening suite seems to be not so bad... andSure instead of comments, I prefer to relay on ResultsYes.. in this way, now you have chances to compare E.g Unique openings are far away from reality or not...In other words, no any other test suite to be so strong..And in same time to produce low draw percentages!
You may know too, Many of opening test suites are based mainly on HumanOpenings..that's sounds as good and as usual less draws..But be aware of that too, just in case of 3500+ eng testingsThen you need thousands of games (per player). otherwise,The error margin can be too high..and frankly even though,No any guarantee...for this I suggest different test suites...But via strong openings: 500-1000 games is enough data,Regarding to determine any chess engine strength
And without to not mention this I can't and once more: Since NNUE on scene, better quality games for sure!Anyhow, situation has been changed.. (not as before)E.g overall draw percentage is dramatically increased ..That means also (in same time), it's too boring chess!
On other hand, it will be a BIG mistake, if we will start toBlame the strongest opening theory...why I say like this,Because the high draw percentage is related mainly to Engine strength, sure the short openings play influence too...But main reason of high draws are engines (not openings)!E.g to be sure, simply you can check LXIV / Human-LikeWhere perfectly clear that the draw ratio is much different..Be aware, for both:I used exactly same books, exc. enginesAnd these both testings are played as NON - NNUE e.g via Strong Evalfiles are expecting 90-95% draws, this should beKnown too, yes. 'papers' and 'reality' are two different things!In other words, I can summarize it (to appear many wins):- No much chances by strong openings!- BIG chances by weak chess engines!
Btw, if still not so clear or for anyone missed, interested..I suggest to check my played opening testings: ResultsMeanwhile, since many years (I think that since 2014)Balsa plus Perfect book series are based on only TopEngine games....and almost all other well-tuned deep books are based on engine games as well..and why?The answer is very simple:We test mainly Top engines, actually it's a long story..But in short, we should not mix apples with oranges!)Hopes helps...
After all,It's up to you, anyone is free to prefer any conditions,But just don't blame/criticize the strongest openings!And don't try to fool us that the weak lines are best!Because SCCT is born for this as well, more Details
And here are the final standings, The Winner FisherovMonk - Congratulations to Fisherov, Stockfish team!
Rank Name Elo + - Games Score Oppo. Draws 1 FisherovMonk 1.2 3762 5 5 1700 51% 3753 90% 2 Zeus 26.5 3761 5 5 1700 51% 3754 90% 3 Kookaburra 1.01 3761 5 5 1700 51% 3754 91% 4 SolistaAttack v2 3760 5 5 1700 51% 3751 91% 5 Polyfish 091022 3760 5 5 1700 51% 3754 89% 6 Solidor 121122 3759 5 5 1700 52% 3748 89% 7 Kayra 1.7 3758 5 5 1700 51% 3752 91% 8 StockfishPB 290922 3758 5 5 1700 51% 3754 89% 9 Stockfish 071122 3756 5 5 1700 51% 3751 91% 10 BrainLearn 20 3754 5 5 1600 50% 3753 91% 11 Cool Iris 5.2 3753 5 5 1600 51% 3750 90% 12 YuliGM PrO 4 3753 5 5 1700 51% 3748 85% 13 ProteusSF-P 040922 3753 5 5 1600 50% 3753 89% 14 Swordfish 15.3a Goi 3752 5 5 1700 51% 3750 87% 15 Stockfish 15 3750 5 5 1600 50% 3749 87% 16 TealOracle 3749 5 5 1600 51% 3743 86% 17 DeepBlue 180422 3742 5 5 1500 50% 3742 85% 18 Big Ballz 010922 3736 6 6 1200 50% 3737 85% 19 Fat Titz 2 3734 6 6 1200 49% 3740 84% 20 Honey 14.1.01 3733 6 6 1200 49% 3737 86% 21 Crystal 5 KWK 3720 6 5 1200 47% 3739 84% 22 Cfish 130821 3706 5 5 1200 44% 3741 83% 23 Spider 080621 3688 5 5 1200 40% 3741 74%
• Calculated by Bayeselo / SF15 is fixed to 3750 Elo / Blue Name = Last Update
More Details: • Mostly of the engines BMI2, tested via theirs default Evalfiles Exception, Zeus v25.6 is played, the reason v27 is not BMI2 And Zeus is used to play via nn.zeus (based on13.07.2022) • Plus some engines are out, simply due to they are very close In similarities to others.. e.g the target is between 68%-74% • All NNUE engines played with a new 'Unique' opening suite Note also that, each of the opening line is playing as 'twice' • The overall draw percentage (based on 17700 games) 88% • Some older (Stockfish 15 etc.) are tested just for comparing And it's clear that no much Elo progress since many months • All played tournament games are available: NNUE CS XXIII
Best, Sedat