Another Stockfish Clone/ Genko

[dkappe]

Have you tried submitting any of your changes to fishtest?

[mehmet123]

Have you tried submitting any of your changes to fishtest?

Genko is a new project and I don't think it is as strong as Kayra. If I talk about Kayra, ı have stated many times that Kayra got good results against Stockfish Dev at 5 minutes or longer time control matches and the results of Kayra is unstable at bullet tests in my tests. This is largely true for Genko too. The time controls at Fishtest isn't suitable for Kayra/Genko (10 sec + 0.1/1 min + 0.6 sec).

I exclude all my own tests for objectivity. At indepence tests latest versions of Kayra won all their one to one matches vs Stocfish Dev and vs some strong Stockfish Derivatives (min: 300 games/ 5 minutes/ 4 cores). At Inwoba rating list (4 Cores/ 4 min + 2 sec TC) the performances of Kayra 1.5 and Kayra 1.4 aren't bad too. Only the result of Kayra 1.6 can be said to be a little bad.
(Kayra 1.6 was released at 2022/09, Kayra 1.5 was released at 2022/07 and Kayra 1.4 was released at 2022/05)

Inwoba Rating List:

Rank Name Elo + - games score oppo. draws
1 Kayra 1.5 avx2 3502 12 12 2300 70% 3376 59%
2 Kayra 1.6 avx2 3494 11 10 2500 62% 3422 73%
3 Kayra 1.4 avx2 3489 12 12 2000 63% 3410 71%
4 CorChess 3 171022 3487 15 15 1400 72% 3353 55%
5 StockfishMZ 220622 3484 13 13 1500 52% 3476 94%
6 Flaming Titz 230822 AVX2 3484 12 12 1600 52% 3475 87%
7 StockfishMZ 230522 avx2 3481 47 47 100 54% 3464 91%
8 CorChess 3 230422 3480 12 12 2000 62% 3410 74%
9 Swordfish 15.3a-avx2 3479 10 10 2300 51% 3475 93%
10 BrainLearn 16 3478 10 10 2600 52% 3470 95%
11 Blue Marlin 15.3a-avx2 3477 12 12 1500 50% 3476 92%
12 Stockfish 130322 3477 21 21 500 50% 3475 98%
13 Kayra 1.3 avx2 3476 21 21 500 50% 3475 98%
14 ProteusSF RBE 008a 3475 11 11 2500 61% 3392 76%
15 EMAN 7.93 64-bit AVX2 3475 11 11 1900 51% 3471 97%
16 BrainLearn 17 3473 9 9 3400 57% 3434 84%
17 Swordfish 15.1-avx2 3472 11 11 1900 50% 3470 99%
18 EMAN 8.10 64-bit AVX2 3470 9 8 3800 59% 3420 80%
19 Blue Marlin 15.1-avx2 3470 19 19 600 50% 3471 98%
20 Raubfisch X48d2_nn_sl-avx2 3468 9 9 3300 60% 3413 79%
21 EMAN 7.80 64-bit AVX2 3467 14 13 1500 65% 3381 69%
22 Fat Titz 110122 64 AVX2 NUMA 3467 12 12 1800 62% 3398 75%
23 Kayra 1.1 avx2 3465 24 24 400 52% 3456 93%
24 ShashChess 24 3464 13 13 1500 49% 3467 86%
25 Swordfish 14.7-avx2 3463 13 13 1600 63% 3387 72%
26 Fisherov v0.98c NNUE popcnt 3463 18 18 700 49% 3469 96%
.
.
Ryzen 7 3700X 3600 MHz, 32 GB Ram, 4 Cores/ 4 min + 2 sec TC, Ponder Off, 1024MB Hash, Salvo Opening Suite (8moves), KLO-150, KLO-250, TBS: Nalimov, Gaviota, Robbobases, Syzygy
https://www.mediafire.com/folder/6b585g ... /Documents

[Eduard]

You talk about tournaments, but only imagine a private tournament that nobody cares about.

What do you need Genko for? For analysis, the engine is weaker than Kayra. For people who like to analyze their own chess games I wouldn't recommend Genko, sorry.

There are 3 servers on the internet where you can play with engines: Lichess, Playchess and InfinityChess.

You don't play there. You present an engine that you only test private. Why don't you play on the servers yourself? Lichess is free, InfinityChess is free. There is a new Banksia GUI version to play live online. I recently played on Lichess.

I also test my engine on the other servers, my friends and I test thousands of games live. What about an engine that only plays well in the quiet at some private tournament. I only saw Genko once on PlayChess, but no one plays with it there anymore. It's no wonder. If the author doesn't play it live himself, why should others do it?

InfinityChess GUI is free and can be downloaded here after you have registered. There is a tournament every day at 4pm server time (Abu Dhabi). Time controls are from 5m+3s to 16m+2s:

http://www.infinitychess.com/

[Eduard]

I'm developing an engine because I want to play it live! I don't care what other people think! Deutronius is currently at the top of the PlayChess rankings. I particularly enjoy games like this. Live from Server PlayChess:

[Event "5 min, rated"]
[Site "Engine Room"]
[Date "2022.11.24"]
[Round "?"]
[White "Amonfritz, Stockfish 051122"]
[Black "Detlef Uter, Deutronius 211122-a"]
[Result "0-1"]
[ECO "E04"]
[WhiteElo "2910"]
[BlackElo "2951"]
[PlyCount "100"]
[EventDate "2022.11.24"]
[EventType "blitz"]

1. d4 d5 2. c4 e6 3. Nf3 Nf6 4. g3 dxc4 5. Bg2 a6 6. O-O Nc6 7. e3 Bd7 8. Qe2 Bd6 9. Qxc4 O-O 10. Nbd2 Rc8 11. Qc2 Nb4 12. Qb3 Bb5 13. a3 Bxf1 14. Bxf1 Nc6 15. Qxb7 Nb8 16. b4 c6 17. e4 Re8 18. e5 Re7 19. Qa8 Bxe5 20. dxe5 Nd5 21. Nc4 Rec7 22. Bg5 Qf8 23. Nd4 h6 24. Bd2 Rd8 25. a4 Rdd7 26. Nxc6 Rxc6 27. a5 Rd8 28. Rb1 Rc7 29. Be3 Nxe3 30. Nxe3 Rd4 31. b5 axb5 32. Bxb5 Nd7 33. Qxf8+ Kxf8 34. a6 Nxe5 35. f4 Nf3+ 36. Kf2 Nd2 37. Ra1 Rd8 38. Be2 Ra8 39. f5 Nb3 40. Ra3 Nd4 41. Bd3 Rca7 42. g4 Rb8 43. Ra4 Rb2+ 44. Kg3 Nb5 45. Kf3 Nc7 46. fxe6 fxe6 47. g5 hxg5 48. Ra3 Rb6 49. Kg4 Nxa6 50. Kh5 Ra8 0-1

Yes, my engine is playing on 124 threads, so what? The games are fantastic. I then analyze such positions:

Amonfritz,Stockfish 051122 - Detlef Uter,Deutronius 211122-a 0-1 0.0, 5 min, rated blitz 2022
[fen]1nrq1rk1/1Q3ppp/p1pbpn2/8/1P1PP3/P4NP1/3N1P1P/R1B2BK1 b - - 0 1[/fen]

Best move is Re8!

Analysis by Genko 1.0 avx2:

17...Be7 18.Rb1
= (0.18 --) Depth: 42/49 00:01:01 1157MN, tb=1397

Analysis by Deutronius 241122-avx2:

17...Re8
=/+ (-0.37) Depth: 39/58 00:00:43 853MN, tb=95

This analysis is good, but not good enough for me. I will try to improve the engine. Live games are the icing on the cake, that's why I do everything I do! If you want to improve something on your engine, then analyze such practical positions.

[mehmet123]

You talk about tournaments, but only imagine a private tournament that nobody cares about.

What do you need Genko for? For analysis, the engine is weaker than Kayra. For people who like to analyze their own chess games I wouldn't recommend Genko, sorry.

There are 3 servers on the internet where you can play with engines: Lichess, Playchess and InfinityChess.

You don't play there. You present an engine that you only test private. Why don't you play on the servers yourself? Lichess is free, InfinityChess is free. There is a new Banksia GUI version to play live online. I recently played on Lichess.

I also test my engine on the other servers, my friends and I test thousands of games live. What about an engine that only plays well in the quiet at some private tournament. I only saw Genko once on PlayChess, but no one plays with it there anymore. It's no wonder. If the author doesn't play it live himself, why should others do it?

InfinityChess GUI is free and can be downloaded here after you have registered. There is a tournament every day at 4pm server time (Abu Dhabi). Time controls are from 5m+3s to 16m+2s:

http://www.infinitychess.com/

At a chess platform where opening books are so dominant, it is very difficult to understand the true power of the chess engine. If it was mandatory to use the same standard opening book for the engines in these platforms, then it would be possible to test the real power of the engines. I think that with Stockfish 12 (first official version with NNUE net) that plays with a strong and innnonative opening book , it can be very successful in these platforms. I had a little experience at Playchess two years ago. With a not so strong chess engine and a not so strong hardware (4-6 cores) i got goot results with my Hubble experimental book. For a while my chess engine had played at 1 core but the opponents chess engines had played at 8- 32 cores but they couldnt beat me. Elo difference with Stockfish X at 32 cores and Stockfish X at 1 core is 120 -150 elo. If there is no victory in a long game series despite this serious power difference, it is the opening books, not the chess engine, that are really tested on these platforms.

[Eduard]

You want people to use your engine, but you're not moving forward. At least Andrea Manzo is there on PlayChess watching. MZ played and tested a lot on PlayChess (even with 2 accounts at the same time). The Eman author plays the tournaments on InfinityChess every day. He even plays with 3 or 4 accounts at the same time. The Raubfisch author also plays on PlayChess, and also on InfinityChess. I played against him in the freestyle tournament. The Fire author is also on InfinityChess. We all test live. Only you don't do that and only post unimportant private results.

[mehmet123]

You want people to use your engine, but you're not moving forward. At least Andrea Manzo is there on PlayChess watching. MZ played and tested a lot on PlayChess (even with 2 accounts at the same time). The Eman author plays the tournaments on InfinityChess every day. He even plays with 3 or 4 accounts at the same time. The Raubfisch author also plays on PlayChess, and also on InfinityChess. I played against him in the freestyle tournament. The Fire author is also on InfinityChess. We all test live. Only you don't do that and only post unimportant private results.

I didn't have a curiosity about whether people playing on platforms like Playchess use Kayra. I haven't logged into my Playchess account for a long time anyway. The number of download numbers doesn't matter to me either but Kayra is already one of the most downloaded chess engine after Stockfish. I think that especially Turkish chess fans are effective for this.

A few years ago on a not active Turkish chess forum, I started sharing links to new chess programs (not commercial) and news about computer chess.
I took action because it was enough for me to inform even one person about chess programs. In time, interest in this topic has increased, especially after the release of Kayra chess engines and it has exceeded as a record number like 816,000 as of now.

My main purpose of writing in this topic is to increase the culture of young people about Computer Chess World. Young chess fans not only should know popular programs like Stockfish, Komodo, Houdini, but they should know today's engines like Ethereal, Koivisto, Wasp etc. and they should know important nostalgic chess programs like MacHack, Chess, Kaissa, Bell, Cray Blitz, Hitech, Deep Thought, Zappa etc. Also, chess fans should not forget the names of people who contributed to Computer Chess World such as Georgy Adelson-Velsky, Hans Berliner, Dan/Kathleen Spracklen, Tony/Linda Scherzer, Ken Thompson, Robert Hyatt, Don Dailey etc.

Many things were said in the forums of Kayra chess engines that it violated GPL rules( pre test times), clone, weak, codes / parameters are ridiculous, etc. These are not things that I'm angry about, people are free to think as they want. But even today I get angry at the disrespectful people who make fun of Robert Hyatt and insult him because of Crafty engine isn' t one of the top engines in Rybkaforum. Although the chess engines developed by people who have made significant contributions to the Computer Chess World in the past are among the weakest chess engines today, these people don' t lose anything from their dignity.

[Eduard]

You are free to do whatever you want, only: Engines like Genko are unnecessary. I really don't know what you want to achieve with this engine? The engine uses a different net. Nothing has to be bad, but it's not better than the standard network either. Otherwise the developers would have named this network as the master. And the other changes you made are nothing special either. In Kayra you changed so many parameters that it's actually hard to get out of this parameter forest now.

That's why you're not making progress with Kayra, and that's why you tried something new. This is always the case when you move so far away from the master engine (Stockfish dev) that it is difficult to implement the latest master code, combining too many changed parameters with the latest ones is not easy. You've changed it too much and now you can't get any further. MZ also wrote something about Kayra in the Chess2U forums, but you ignore it like everything else. When I think of your sentence "I want to create the best engine in the world", it was pretty arrogant. But keep it up, your fans will surely support you.

[mehmet123]

You are free to do whatever you want, only: Engines like Genko are unnecessary. I really don't know what you want to achieve with this engine? The engine uses a different net. Nothing has to be bad, but it's not better than the standard network either. Otherwise the developers would have named this network as the master. And the other changes you made are nothing special either. In Kayra you changed so many parameters that it's actually hard to get out of this parameter forest now.

That's why you're not making progress with Kayra, and that's why you tried something new. This is always the case when you move so far away from the master engine (Stockfish dev) that it is difficult to implement the latest master code, combining too many changed parameters with the latest ones is not easy. You've changed it too much and now you can't get any further. MZ also wrote something about Kayra in the Chess2U forums, but you ignore it like everything else. When I think of your sentence "I want to create the best engine in the world", it was pretty arrogant. But keep it up, your fans will surely support you.

The NNUE network( NN-154550620475) that now uses Genko isn't better than default NNUE net but also this net isn't worse than default net. In tests NN-154550620475 beat default NNUE net two times (+ 0,45 elo, +0,37 elo ).This NNUE network didn't become a standard NNUE net due to the small difference in elo.
For me on working Genko is much easier than on Kayra. I felt need to publish this chess engine because I realized the potential to strengthening some failed patches with minor codes and parameter changes. For nowadays I was busy with work, I haven't spend more time on Genko, but I think this chess engine will be more powerful in 2023 even without adding new Stockfish patches.

Should I say that I want to develop a chess engine worse than Stockfish? Kayra / Genko are Stockfish derivative chess engines. There are two possibilities, they will either be stronger than Stockfish and become the strongest chess engine in the world, or they will be weaker than Stockfish. Where is the arrogance in that? I have shown my respect for the Stockfish Developers on every platform. The name of the "Stockfish Developers" is mentioned in every version of the Kayra/Genko engine.

[Eduard]

As for the network: Why use another network if it is not better than the master network? That is illogical. The master net is used for fish cooking. All optimizations are coordinated with the master network. And this is exactly where deviations can occur if a different network is used.

But you know everything better than everyone else. Other nets are sometimes tested for fish cooking, but then with other parameters. The latest parameters and patches are tested with the master network. You do it differently, you adopt the parameters, but then you use a different network that was tested sometime earlier with different parameters.

Oh, arguing with you makes no sense to me. You are too smart for me. You are smarter than everyone else. That's why you have to constantly build a new engine. When you get stuck, you start over, like you do with Genko.

I wish you a lot of fun with it, because having fun is the most important thing for me. However, Genko is no better than Stockfish and has no different style. I don't need such an engine.