Last Update: 13.10.2023

Hello Chess Friends,


As usually, I am very pleased to announce that,

New series of NNUE strength testing organized!

Btw, (for anyone missed) previous results: III / IV


1st of all, 

Many strength tests organized mainly to see the real power of books!Yes...depending on cond. NN, TC, openings, opponents etc. we may See various results in performances, at least  latest tests indicate this..Other fact is that just in case of using weak openings...for examples:NON-NNUE+strong openings may overcome NNUE+weak openings!Exc. that recently Everest NON-NNUE book is proved that: ResultsDON Hercules 2's produced new data... can be as other reference..And just in that case...even doubling time controls...does not help !!
To be more clear about latest new matches,I picked especially both players, because according to previousGames (in recent tours) the results are ended in favor for DON...Now is clear that even giving NNUE assistance for weaker lines,Besides even giving more time...DON NON-NN is just unstoppable!E.g DON Hercules 2 is played even via Cfish 030820 NON-NNUEAnd this is not all.. I set 2m+1s for DON Hercules's opening book,Where Lion 110423 book is played via Raubfish X48c3 SC NNUEPlus, I set as double TC: 4m+2s for Lion 110423's opening book
And playing even under these disadvantage test conditions,DON Hercules NON-NNUE is performed about 56 Elo better: 
1 DON Hercules2 NON-NN 2m+1s +16/-8/=26 58.00%  29.0/502 Lion 110423 NNUE ON 4m+2s  +8/-16/=26 42.00%  21.0/50
What I can say more over the current Duel match,Once more we noticed that DON book fully deserves 'Hercules' name !)But this is also true that Lion 110323 is not weak one, just suffers vs DON...E.g especially current Lion book ver. managed to qualify in recent tours!-------------------------------------------------------------------------------------------------------2nd test: Both players via Balsa, 2m+1s +237 Elo (in favor for NNUE)In same that means, in case of Balsa openings, NNUE Elo diff. is BIG:.
1 Raubfish X48c3 SC NN  +297/-1/=202 79.60%  398.0/5002 Cfish 030820 NON-NN   +1/-297/=202 20.40%  102.0/500-------------------------------------------------------------------------------------------------------3rd test: it's a new rating, where all played by Raubfish SC via 30sec+0.5sec,Exception, only DON NON-NN book is played by Cfish 030820 via 2min+1secAnd plus DON NON-NN played with time controls via 8m+1s and 15m+1s too
Even under these cond..it's clear that NNUE Bullet seems to be so strong !E.g if comparing.NNUE fast Bullet is about 40 Elo over 2m+1s NON-NNUEOr if comparing..NNUE fast Bullet is about 27 Elo over 8m+1s NON-NNUEJust if 15m+1s for NON-NN: their performances are more close each other!(Sure I'm referring...in case of comparing DON Hercules 2's played results)
Rank Name              Elo    +    - games score oppo. draws    1 ORIONPAX3         3778    8    8   700   53%  3758   88%    2 DON NON-NN 15m1s  3772    9    9   550   51%  3762   82%    3 DON Hercules 2    3769    9    9   550   51%  3762   90%    4 Everest 300923    3769    8    8   700   51%  3759   83%    5 CachoSCCT opt     3766    8    8   700   51%  3759   89%    6 ShadRach          3766    8    8   700   51%  3759   87%    7 OlisFish 260923   3763    8    8   700   51%  3760   94%    8 SENTINEL 2        3763    8    8   700   51%  3760   91%    9 OPTIMUS 4         3761    8    8   700   50%  3760   93%   10 LesChats 071023   3760    8    8   700   50%  3760   86%   11 Zaytsev           3755    8    8   700   49%  3760   90%   12 Pimienta          3755    8    8   700   49%  3760   86%   13 Red Dragon        3749    8    8   700   48%  3761   87%   14 DON NON-NN 8m1s   3742    9    9   550   47%  3762   81%   15 DON NON-NN 2m1s   3729    9    8   550   45%  3762   80% 
    Some Notes:    DON NON-NN 2m1s = DON Hercules 2 by Cfish 030820, used 2min+1sec    DON NON-NN 8m1s = DON Hercules 2 by Cfish 030820, used 8min+1sec    DON NON-NN 15m1s = DON Hercules 2  by Cfish 030820, used 15m+1s---------------------------------------------------------------------------------------------------------4th test:  Balsa openings, 30s+0.5s+NN vs 15m+1s NON-NN                E.g according to current test: +57 Elo (for Bullet NN)       1 Raubfish X48c3 SC NNUE  +125/-43/=332 58.20%  291.0/5002 Cfish 030820 NON-NN     +43/-125/=332 41.80%  209.0/500---------------------------------------------------------------------------------------------------------5th test: another new rating, where all played via Raubfish SC via 1min+1secTarget of running current test is to see / compare of recent Ultimate (2m+1s)Actually the other Bullet rating (30sec+0.5sec) is played for comparing as well
Rank Name              Elo    +    - games score oppo. draws    1 Everest 300923   3775    5    6   550   54%  3753   87%    2 ORIONPAX3        3767    6    6   550   52%  3754   89%    3 Pimienta         3757    6    7   550   50%  3755   95%    4 ShadRach         3757    6    7   550   50%  3755   92%    5 CachoSCCT opt    3757    6    7   550   50%  3755   91%    6 Zaytsev          3755    6    6   550   50%  3755   94%    7 DON Hercules 2   3753    7    6   550   50%  3755   93%    8 OlisFish 260923  3751    6    6   550   49%  3755   96%    9 OPTIMUS 4        3750    7    6   550   49%  3755   94%   10 SENTINEL 2       3750    6    6   550   49%  3755   96%   11 Red Dragon       3745    6    6   550   48%  3756   89%   12 LesChats 071023  3742    6    6   550   48%  3756   91% ---------------------------------------------------------------------------------------------------------6th test organized just for comparing..e.g played at 30s+0.5s via Raubfish SCSeveral players used, the ones which could not qualify...plus who qualified: LXVI In other words, now we can compare e.g NON-NNUE vs NNUE of recent: LXVI And it seems under these cond. (via Raubfish SC NN) DON is doing very wellPlus LesChats, Raid 3 TR, where LXVI's Winner (SENTINEL) was not so lucky..About Cubail, Arasan. as before..as NON-NNUE or NNUE.. both books suffer !
Rank Name             Elo    +    - games score oppo. draws    1 DON Hercules 2   3785    9   10   300   58%  3737   80%    2 LesChats 071023  3756   11   12   300   53%  3737   88%    3 Raid 3 TR        3755   11   12   300   53%  3737   89%    4 SENTINEL 2       3742   12   12   300   51%  3737   88%    5 Champs2020       3742    8    8   600   49%  3749   84%    6 Champs2022       3740    8    8   600   49%  3749   87%    7 Cubail 120823    3731   12   12   300   49%  3737   84%    8 Champs2021       3730    8    8   600   47%  3749   88%    9 Arasan 24X       3723   12   11   300   48%  3737   88% --------------------------------------------------------------------------------------------------------7th test same NNUE cond. as above, exception slower TC used 60sec+0.6secAs usually, DON with superior performance, this time Champs2022 is so good..Also good performance come by LesChats, SENTINEL and Raid book, whereUnfortunately again both books (Cubail, Arasan) could not gain much points..!
 Rank Name             Elo    +    - games score oppo. draws    1 DON Hercules 2   3785    7    8   300   59%  3739   77%    2 Champs2022       3763    7    7   600   52%  3753   90%    3 SENTINEL 2       3757    9   10   300   53%  3739   91%    4 LesChats 071023  3757    9   10   300   53%  3739   88%    5 Raid 3 TR        3755    9   10   300   53%  3739   87%    6 Arasan 24X       3734   10   10   300   49%  3739   93%    7 Cubail 120823    3733   10   10   300   49%  3739   86%    8 Champs2021       3730    7    6   600   46%  3753   85%    9 Champs2020       3726    6    6   600   45%  3753   87% ---------------------------------------------------------------------------------------------------------8th test: HASH testings: Balsa, 1m+0.6s, Raubfish SC for both        1   32 MB hash   +40/-40/=920 50.00%  500.0/1000  2   128 MB hash  +40/-40/=920 50.00%  500.0/1000             1   64 MB hash   +23/-23/=354 50.00%  200.0/400  2   128 MB hash  +23/-23/=354 50.00%  200.0/400  ------------------------------------------------In this test same conditions as above, exc. 30sec+0.5sec used:              1   32 MB hash  +34/-22/=344 51.50%  206.0/4002   64 MB hash  +22/-34/=344 48.50%  194.0/400---------------------------------------------------------------------------------------------------------More Details,All  of the played games are available (in PGN format): Strength VAlso the latest games (Bullet vs Rapid) are available: NN Vs NON32 MB hash size seems to be stronger in fast Bullet time controlsDeep Books+Bullet+NN is close in strength to Rapid NON-NN, butVia Balsa, 30s+05s+NN seems to be clearly stronger than 15m+1sNON-NN e.g according to recent above testings  approx. +57 EloAnd why via Balsa openings more Elo diff. that DON's openings?Simply because DON's openings are much more deeper than Balsa...In same time that means (going deeper..) NON-NN gains more time!In other words, giving more time means more strength, Elo points!
What I can say more,Just imagine before NNUE era: Rapid is considered as good quality..At present, even fast Bullet NN is close level as in the past - Rapid !After all I think that the real problem is in our brains...if Bullet bad...)It is time for new revolution !)
Ok.. that's all for now...see you later )

Best,Sedat