Hello Chess Friends,


I am pleased to announce that, I organized a new Duel Match II: Books vs NNUE Duel 2nd name can be called 'testing NNUE strength' The current match conditions are almost same as: CS L Meanwhile, a similar Duel match Books vs NNUE Where the Winner of previous one was 'NNUE' too In previous duel, mostly were old-dated opening books But now, many of the players are new opening books... My used EvalFile is new too: nn-62ef826d1a6d.nnue Just I'd like to point out, (about determining the real NN strength) Many engine testers prefer short books: NON-NNUE vs NNUE Where depending on conditions, usually the Evalfile is stronger But running under similar conditions: is it fair ? and I do not think so... E.g NON-NNUE engine benefits up to 5-8 moves (from the short book) While the engine via NN file gains in deep openings, plus in middle game... With fast time controls, the Elo difference can be minus plus 50-100 Elo But giving more time, cores etc. then the difference is less than 50 Elo One thing more, usually a short book vs well-optimized deep book The Elo points are in favor for deep book too..so this is nothing new... For example, sure depending on books, but 50 Elo can be easily gain... Btw, what about in case of using a well-optimized deep book ? E.g for NON-NNUE engine...can we see 100 Elo difference...? I hardly doubt...even there is a possibility the engine+nn file to suffer... Sure, my goal/plan is not to create a anti-NNUE style opening book... But if I will start to work...Believe Me...the situation will be changed... Note: I am referring in case of using latest Stockfish, Cfish etc.. In other words, I strongly believe in that (via current Duel testings...), This is the best way to check the 'real' strength of NNUE (EvalFile) But not via short books, otherwise it's disadvantage for NON-NNUE Anyhow, there is no ideal methods (including the current testing too) But we have to do our best for more fair and better conditions...! And now let's see the performance:of the current Duel II results: As before, in overall, NN file is stronger...but against a few suffers.... And as I stated earlier, it all depends....who is the opponent !).
 The Winner is again EvalFile - Congratulations to SFisGOD/SF Team! Really impressive...nn-62ef826d1a6d.nnue file seems to be very strong... # PLAYER RATING POINTS GAMES (%) 1 EvalFile : 3725.9 287.5 500 57.5% 2 Alchemist 110121 : 3715.8 306.0 540 56.7% 3 M10.5 : 3690.7 340.5 634 53.7% 4 2600-Plus-Club : 3689.1 339.0 634 53.5% 5 RadaR 050121 : 3688.8 169.0 322 52.5% 6 Trend Moves 2020+ : 3687.4 337.5 634 53.2% 7 Perfect2 by Cevdet : 3686.0 202.5 394 51.4% 8 HurnaTal v2 : 3685.7 308.5 584 52.8% 9 Rufus2020 : 3684.7 335.0 634 52.8% 10 1337chess 260720 : 3682.5 260.0 504 51.6% 11 TR Power : 3682.5 333.0 634 52.5% 12 Chemist : 3681.9 332.5 634 52.4% 13 RadaR 011220 : 3681.4 332.0 634 52.4% 14 Cevdet-BIN : 3680.8 331.5 634 52.3% 15 DON Lite 3 : 3680.8 314.5 604 52.1% 16 Lizard 171220 : 3680.3 331.0 634 52.2% 17 Mr Noname 4.5 : 3679.2 330.0 634 52.1% 18 Cerebellum 3M Dual : 3679.2 330.0 634 52.1% 19 GX Power : 3678.6 329.5 634 52.0% 20 Katzenmaier3012 : 3678.1 329.0 634 51.9%

Games


 More Details: Duel Match II is played by same chess engine: CF EXT 291120 C40 EvalFile is played via nn-62ef826d1a6d.nnue and without any book Where all of the opening books are played without using any EvalFile The overall draw percentage (based on 500 games) is not so high 79% The Duel Match II's Individual Statistics (Higher is Better) Note: Only 4 opening books managed to perform better than EvalFile! DieHard, Perfect, Rufus and Vitamin: Congratulations and well-done! 10 Books scored equal, really not bad ....unfortunately rest lost...(
 1 EvalFile 3725 500 (+ 91,=393,- 16), 57.5 % Rufus2020 : 10 (+ 0,= 9,- 1), 45.0 % Perfect2 by Cevdet : 10 (+ 0,= 9,- 1), 45.0 % Vitamin 220420 : 10 (+ 0,= 9,- 1), 45.0 % DieHard : 10 (+ 0,= 9,- 1), 45.0 % Mr Noname 4.5 : 10 (+ 1,= 8,- 1), 50.0 % Trend Moves 2020+ : 10 (+ 1,= 8,- 1), 50.0 % Hyena : 10 (+ 1,= 8,- 1), 50.0 % Katzenmaier3012 : 10 (+ 0,= 10,- 0), 50.0 % ProDeo 3 : 10 (+ 0,= 10,- 0), 50.0 % 1337chess 260720 : 10 (+ 1,= 8,- 1), 50.0 % RadaR 011220 : 10 (+ 0,= 10,- 0), 50.0 % Bigshot 271220 : 10 (+ 1,= 8,- 1), 50.0 % Feta 2 : 10 (+ 0,= 10,- 0), 50.0 % Rouge 190620 : 10 (+ 0,= 10,- 0), 50.0 % Alchemist 110121 : 10 (+ 1,= 9,- 0), 55.0 % DON Lite 3 : 10 (+ 1,= 9,- 0), 55.0 % 2600-Plus-Club : 10 (+ 1,= 9,- 0), 55.0 % Cevdet-BIN : 10 (+ 1,= 9,- 0), 55.0 % Lion 5 : 10 (+ 1,= 9,- 0), 55.0 % GX Power : 10 (+ 1,= 9,- 0), 55.0 % Cerebellum 3M Dual : 10 (+ 1,= 9,- 0), 55.0 % Blue 060121 : 10 (+ 1,= 9,- 0), 55.0 % ChazzChess : 10 (+ 1,= 9,- 0), 55.0 % HurnaTal v2 : 10 (+ 2,= 7,- 1), 55.0 % Tokugawa-Beta-7 : 10 (+ 2,= 7,- 1), 55.0 % Ranomi 150318 : 10 (+ 1,= 9,- 0), 55.0 % Lockenfrosch : 10 (+ 2,= 8,- 0), 60.0 % Lizard 171220 : 10 (+ 3,= 6,- 1), 60.0 % Chemist : 10 (+ 2,= 8,- 0), 60.0 % SolistaBig : 10 (+ 2,= 8,- 0), 60.0 % ICCF : 10 (+ 2,= 8,- 0), 60.0 % SSDF : 10 (+ 2,= 8,- 0), 60.0 % engPowerbook2021 : 10 (+ 3,= 6,- 1), 60.0 % ArmoR : 10 (+ 3,= 6,- 1), 60.0 % M10.5 : 10 (+ 2,= 8,- 0), 60.0 % M10.2 : 10 (+ 2,= 8,- 0), 60.0 % TR Power : 10 (+ 3,= 6,- 1), 60.0 % RadaR 050121 : 10 (+ 2,= 8,- 0), 60.0 % Zodiac Reborn : 10 (+ 2,= 8,- 0), 60.0 % Perfect by Cevdet : 10 (+ 3,= 7,- 0), 65.0 % OLiprog : 10 (+ 3,= 7,- 0), 65.0 % Everest 071020 : 10 (+ 3,= 7,- 0), 65.0 % DualBrain AZ : 10 (+ 3,= 7,- 0), 65.0 % Buffalo v2 : 10 (+ 3,= 7,- 0), 65.0 % DON Lite 2 : 10 (+ 4,= 6,- 0), 70.0 % CF1977 Rebirth : 10 (+ 4,= 6,- 0), 70.0 % HIPPO v3 : 10 (+ 4,= 6,- 0), 70.0 % Rouge 080121 : 10 (+ 5,= 4,- 1), 70.0 % Appolo : 10 (+ 4,= 6,- 0), 70.0 % NNUE : 10 (+ 6,= 4,- 0), 80.0 %

Best Wishes,Sedat Canbaz