Latest Update: 26.05.2024
Hello Chess Friends,
1st of all, let Me start please with this new saying:The more I learn about Top NNUE engines,The more I see Top similarity each other!
To be more clear,Similarity tools are not so dead, but even rather I would sayThe ORIGINALITY is almost dead when it comes to NNUE!
But I admit too we have many programmers who are honestAnd do not hide from where come theirs eng origins...right?What about all others eng programmers? if still not so clear:All these strong ideas are came directly from MARS or ?) andHow many of the created engines are started from scratch ?)For these reasons, I like the authors of SF-based engines too!If nothing else..these authors are fair...and give Me one reason To not test Raubfish or RapTora or Fat Fritz..since these enginesSuit to the current used rules (up to 70% in similarity is allowed)
Btw, you can download the latest SIMEX2's Test ResultsWhere I assume you'll see/know about what is going on...
And here are the latest played results (main target 3600+):
The Champion is again Stockfish 16.1- Congratulations to Stockfish Team! And strange that latest SF dev. performed worse, normally newer should be better..
Rank Name Elo + - games score oppo. draws 1 Stockfish 16.1 3772 5 5 2000 65% 3693 70% 2 Stockfish 210524 3769 5 5 2000 64% 3694 71% 3 RapTora 2.2 3761 6 7 1200 58% 3715 81% 4 Raubfish X48c3 3746 7 7 1200 55% 3717 81% 5 K.Dragon 3.3 3740 6 6 1500 54% 3713 84% 6 Berserk 13 3736 6 6 1500 54% 3713 86% 7 Obsidian 12 3726 6 7 1200 55% 3703 79% 8 Fat Fritz 2 3703 5 5 1700 54% 3682 77% 9 Caissa 1.18 3703 6 6 1400 48% 3717 84% 10 Obsidian 11 3699 7 7 1100 48% 3709 84% 11 RubiChess 0124 3692 7 6 1300 46% 3717 84% 12 Alexandria 6.1.0 3688 7 7 1200 47% 3703 82% 13 CSTal 2.00 3684 7 7 1200 57% 3643 79% 14 Titan 1.0.0 3683 6 6 1200 49% 3687 75% 15 PlentyChess 1.0 3683 7 7 1200 46% 3703 81% 16 Koivisto 090223 3678 7 7 1200 56% 3643 77% 17 Clover 6.1.19 3678 7 7 1200 45% 3707 81% 18 Rebel 16.3 3675 7 6 1300 45% 3703 82% 19 Uralochka 3.41a 3670 6 6 1200 47% 3684 71% 20 Seer v2.8 3667 7 7 1200 43% 3705 78% 21 rofChade 3.1 3662 7 7 1200 53% 3640 74% 22 Revenge 3 3659 7 7 1300 52% 3643 78% 23 Igel 3.5.0 3655 7 7 1200 50% 3654 81% 24 Velvet v7.3.0 3645 6 6 1300 42% 3685 66% 25 Viridithas 12 3643 7 7 1200 50% 3641 77% 26 Fire 8NN MC3 3631 7 7 1200 50% 3630 79% 27 BlackMarlin 9 3631 7 7 1200 49% 3635 75% 28 Arasan 24.2.2 3624 6 6 1300 39% 3683 69% 29 SlowChess 2.9 3622 7 7 1300 46% 3647 76% 30 Lizard 10.3 3619 7 7 1200 46% 3646 75% 31 Stormphrax 4.1.0 3617 7 7 1200 45% 3649 73% 32 Altair 7.0.0 3612 7 7 1200 44% 3646 74% 33 Booot 7.3 3608 7 7 1200 44% 3647 72% 34 Motor 0.40 3582 6 6 1200 32% 3676 60%
• Calculated by Bayeselo / K. Dragon fixed to 3740 / Blue Name = Last Update
More Details:• All NNUE engines are mainly AVX2 and tested via default Evalfiles• The Overall Draw-ratio (based on 22100 games) is not so high 77% • All of played games are available (in PGN) SCCT CS L / Fire 8NN
And a few more words about original Top chess engines,Before NNUE, the similarity was mainly up to 40% - 50%,Where sounds perfectly fine to Me, at least not so high!But nowadays these numbers are increased av. 60% - 70%Sure I refer about using 100 ms and 1 core but just if using Slower TC and more cores, then similarity is much higher!Besides, I refer also about such Top engines, which are Considering as ORIGINAL (by vast majority of people)!
Meanwhile, if I am not going to use the Similarities tools,Then I wonder: who are best experts in recognizing the Originality of the engines? Because I see many testers,Who refuse to test many of the available Top engines,Very likely such as in their eyes: they are non-original,Where theirs tested engines are as original work, sureThis is a long story, but as last words,I wish to say over: Trust, but verify!
Because for less double standards:I'm doing my best to prefer that way and what about you?
Ok .. hope helps and have a nice weekend )
Best,Sedat