Kayra 1.1

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

amanjpro
Posts: 883
Joined: Sat Mar 13, 2021 1:47 am
Full name: Amanj Sherwany

Re: Kayra 1.1

Post by amanjpro »

AlexChess wrote: Fri Dec 31, 2021 4:26 pm
Sopel wrote: Fri Dec 31, 2021 1:22 pm
AlexChess wrote: Fri Dec 31, 2021 7:48 am
connor_mcmonigle wrote: Thu Dec 30, 2021 3:31 pm
You literally said you value style over Elo in your previous message. However, you'd almost certainly be unable to tell apart Kayra from SF were the names hidden as they play effectively identically... what exactly do you think you're measuring? There are so many manipulated variables and your sample size is so small that your testing is meaningless. We have rating lists to perform third party, "real world" tests already.

I have nothing against what you're doing. Have fun, but don't pretend your playing with chess engines has any value for engine developers.
Even trusted rating lists aren't so reliable also playing thousands of games (but only between few engines, not the totality available)

1. Mixed results obtained using 1,2,4,8 CPUs
2. The same engine (eg Stockfish 14.1) is evaluated from 3500 up to 3900 ELO!!
3. Different time controls
4. Different hardwares (somebody states that every time you double the hardware speed, you gain 80 ELO)

Best regards, Alex

1. Usually attributed to small sample size. I'm not aware of any significant results with more than one thread (apart from fishtest), as they are quite costly. Also, some engines just scale worse with amount of threads.
2. This starts making sense as soon as you learn that Elo is a relative metric
3. It's very rare that an engine's strength changes drastically compared to others depending on time control. Usually explained by small sample size.
4. What? Apart from testing Lc0 I don't know anyone who would use different hardware for different engines being tested in a single match. The relative performance across hardware is usually pretty comparable between engines.
Thank you for you comment Sopel.

4. I mean how can I obtain a trustable ELO using a Ryzen 9 5850 and my Android Snap 626 Smartphone :wink:

Note: I'm trying to understand how to make my tests more meaningful.

Best regards and happy New year 2022!
You cannot have an elo system across different platforms. If you have a rating list for M1 devices (ARM architecture), then it stays that. NNUE performs terrible on ARM for example. If it is for Android, then your rating list is only for Android. For Windows on 64-bit arch you get a separate list. To be fair, even on Windows, different instruction sets produce different numbers. NNUE engines are almost twice as fast on AVX2 CPUs, while HCE engines don't benefit that much from AVX.

What do you want from an engine on Android? do you want it to be the best? that is something nobody cares about it. If you want the best engine (for whatever the best means), you go for a massive data center, using SF latest dev version or LC0 or Dragon. If you want to enjoy, then toss elo and rating lists away and try to enjoy the chess itself
User avatar
AlexChess
Posts: 1562
Joined: Sat Feb 06, 2021 8:06 am
Full name: Alex Morales

Re: Kayra 1.1

Post by AlexChess »

amanjpro wrote: Fri Dec 31, 2021 4:34 pm
AlexChess wrote: Fri Dec 31, 2021 4:26 pm
Sopel wrote: Fri Dec 31, 2021 1:22 pm
AlexChess wrote: Fri Dec 31, 2021 7:48 am
connor_mcmonigle wrote: Thu Dec 30, 2021 3:31 pm
You literally said you value style over Elo in your previous message. However, you'd almost certainly be unable to tell apart Kayra from SF were the names hidden as they play effectively identically... what exactly do you think you're measuring? There are so many manipulated variables and your sample size is so small that your testing is meaningless. We have rating lists to perform third party, "real world" tests already.

I have nothing against what you're doing. Have fun, but don't pretend your playing with chess engines has any value for engine developers.
Even trusted rating lists aren't so reliable also playing thousands of games (but only between few engines, not the totality available)

1. Mixed results obtained using 1,2,4,8 CPUs
2. The same engine (eg Stockfish 14.1) is evaluated from 3500 up to 3900 ELO!!
3. Different time controls
4. Different hardwares (somebody states that every time you double the hardware speed, you gain 80 ELO)

Best regards, Alex

1. Usually attributed to small sample size. I'm not aware of any significant results with more than one thread (apart from fishtest), as they are quite costly. Also, some engines just scale worse with amount of threads.
2. This starts making sense as soon as you learn that Elo is a relative metric
3. It's very rare that an engine's strength changes drastically compared to others depending on time control. Usually explained by small sample size.
4. What? Apart from testing Lc0 I don't know anyone who would use different hardware for different engines being tested in a single match. The relative performance across hardware is usually pretty comparable between engines.
Thank you for you comment Sopel.

4. I mean how can I obtain a trustable ELO using a Ryzen 9 5850 and my Android Snap 626 Smartphone :wink:

Note: I'm trying to understand how to make my tests more meaningful.

Best regards and happy New year 2022!
You cannot have an elo system across different platforms. If you have a rating list for M1 devices (ARM architecture), then it stays that. NNUE performs terrible on ARM for example. If it is for Android, then your rating list is only for Android. For Windows on 64-bit arch you get a separate list. To be fair, even on Windows, different instruction sets produce different numbers. NNUE engines are almost twice as fast on AVX2 CPUs, while HCE engines don't benefit that much from AVX.

What do you want from an engine on Android? do you want it to be the best? that is something nobody cares about it. If you want the best engine (for whatever the best means), you go for a massive data center, using SF latest dev version or LC0 or Dragon. If you want to enjoy, then toss elo and rating lists away and try to enjoy the chess itself
Thank you for your explanation, too :)
Chess engines and dedicated chess computers fan since 1981 :D macOS Sequoia 16GB-512GB, Windows 11 & Ubuntu ARM64.
ProteusSF Dev Forum
JohnWoe
Posts: 529
Joined: Sat Mar 02, 2013 11:31 pm

Re: Kayra 1.1

Post by JohnWoe »

Elo is relative. It depends which number you anchor it to. You could start from 10,000Elo (EngineX). Then EngineY beats it by 200 Elo. It will have 10,200 Elo. EngineX(3800 Elo) beats EngineY (3000Elo) 100-0. 100-0 doesn't tell anything about ratings of those engines. Even if EngineY is putting up a fight in some games. The chess scoring system is too binary.

HCE is worth nothing in SF. You could remove it totally and lose 0 Elo. Why is it still there. I don't know.
I don't know what's inside Kayra. Not downloaded anything. Put if you do modifications to source code. Then it's good to have a name for it. Then you are the author of those modifications. Credit where credit is due!

For example I added some new features to SF. That might be useful to others too. Not Elo gainers. Just more features. How I should publish it? With a new name or some strange git diff?
User avatar
AlexChess
Posts: 1562
Joined: Sat Feb 06, 2021 8:06 am
Full name: Alex Morales

Re: Kayra 1.1

Post by AlexChess »

JohnWoe wrote: Fri Dec 31, 2021 10:05 pm Elo is relative. It depends which number you anchor it to. You could start from 10,000Elo (EngineX). Then EngineY beats it by 200 Elo. It will have 10,200 Elo. EngineX(3800 Elo) beats EngineY (3000Elo) 100-0. 100-0 doesn't tell anything about ratings of those engines. Even if EngineY is putting up a fight in some games. The chess scoring system is too binary.
I have reduced ELO by 150 points to all engines of my SuperBlitz, trying to syncronize computer with human ELO according to GM Larry Kaufman suggestion. Now very top engines have 3500 ELO. You are right, ELO is relative ...the important info is the ranking.
Chess engines and dedicated chess computers fan since 1981 :D macOS Sequoia 16GB-512GB, Windows 11 & Ubuntu ARM64.
ProteusSF Dev Forum
mehmet123
Posts: 690
Joined: Sun Jan 26, 2020 10:38 pm
Location: Turkey
Full name: Mehmet Karaman

Re: Kayra 1.1

Post by mehmet123 »

AlexChess wrote: Thu Dec 30, 2021 7:48 am In 2022 finally I'll begin to develop my own GPLv3 SF derivative that also in Ed's Code of Honor it's allowed... His RebelFish is a must :wink:
I have no doubt that you will develop more powerful chess engine than Kayra.
However, Kayra will not easily lose the title of being the most discussed and criticized chess engine. Also Kayra (private version) is the first chess engine whose tests were banned on a forum :)
User avatar
AlexChess
Posts: 1562
Joined: Sat Feb 06, 2021 8:06 am
Full name: Alex Morales

Re: Kayra 1.1

Post by AlexChess »

:oops:
mehmet123 wrote: Sun Jan 02, 2022 5:56 pm
AlexChess wrote: Thu Dec 30, 2021 7:48 am In 2022 finally I'll begin to develop my own GPLv3 SF derivative that also in Ed's Code of Honor it's allowed... His RebelFish is a must :wink:
I have no doubt that you will develop more powerful chess engine than Kayra.
However, Kayra will not easily lose the title of being the most discussed and criticized chess engine. Also Kayra (private version) is the first chess engine whose tests were banned on a forum :)
Absolutely not :D On chess development I'm really a beginner and I'll keep ProteusChessSF private until It will play completely different from all derivatives. I want to experiment positional sacrifices, no matter the strength (it has already lost 90 ELO :roll:) . I start from Stockfish source only because it is well commented and easy to understand. If 20 years ago there would be a so huge documentation available, it will be much easier for me than starting at 57 :oops:
Chess engines and dedicated chess computers fan since 1981 :D macOS Sequoia 16GB-512GB, Windows 11 & Ubuntu ARM64.
ProteusSF Dev Forum
DrEinstein
Posts: 75
Joined: Wed Sep 15, 2021 8:50 pm
Full name: Albert Einstein

Re: Kayra 1.1

Post by DrEinstein »

AlexChess wrote: Sun Jan 02, 2022 7:23 pm
Absolutely not :D On chess development I'm really a beginner and I'll keep
ProteusChessSF private until It will play completely different from all derivatives. I want to experiment positional sacrifices, no matter the strength (it has already lost 90 ELO :roll:) . I start from Stockfish source only because it is well commented and easy to understand. If 20 years ago there would be a so huge documentation available, it will be much easier for me than starting at 57
But the name of the upcoming completely different playing super engine was already presented to the public. Congrats :mrgreen:
User avatar
AlexChess
Posts: 1562
Joined: Sat Feb 06, 2021 8:06 am
Full name: Alex Morales

Re: Kayra 1.1

Post by AlexChess »

DrEinstein wrote: Sun Jan 02, 2022 8:56 pm
AlexChess wrote: Sun Jan 02, 2022 7:23 pm
Absolutely not :D On chess development I'm really a beginner and I'll keep
ProteusChessSF private until It will play completely different from all derivatives. I want to experiment positional sacrifices, no matter the strength (it has already lost 90 ELO :roll:) . I start from Stockfish source only because it is well commented and easy to understand. If 20 years ago there would be a so huge documentation available, it will be much easier for me than starting at 57
But the name of the upcoming completely different playing super engine was already presented to the public. Congrats :mrgreen:
Thank you. Also your name is a promise. What's your engine? :wink:
Chess engines and dedicated chess computers fan since 1981 :D macOS Sequoia 16GB-512GB, Windows 11 & Ubuntu ARM64.
ProteusSF Dev Forum
DrEinstein
Posts: 75
Joined: Wed Sep 15, 2021 8:50 pm
Full name: Albert Einstein

Re: Kayra 1.1

Post by DrEinstein »

AlexChess wrote: Mon Jan 03, 2022 10:57 am
DrEinstein wrote: Sun Jan 02, 2022 8:56 pm
AlexChess wrote: Sun Jan 02, 2022 7:23 pm
Absolutely not :D On chess development I'm really a beginner and I'll keep
ProteusChessSF private until It will play completely different from all derivatives. I want to experiment positional sacrifices, no matter the strength (it has already lost 90 ELO :roll:) . I start from Stockfish source only because it is well commented and easy to understand. If 20 years ago there would be a so huge documentation available, it will be much easier for me than starting at 57
But the name of the upcoming completely different playing super engine was already presented to the public. Congrats :mrgreen:
Thank you. Also your name is a promise. What's your engine? :wink:
I have just forked SF and modified it a bit. Working name is StockfishXXXXXX and XXXXXX is private. Depending on its additional settings, it plays exactly like SFdev or completely different from all derivatives, with less than 900 Elo if necessary. :lol:

Edit: Here is a game SFxxxxxx vs SFxxxxxx:
Unfortunately, no idea about ELO, but close to random mover.

[pgn]
[Event "?"]
[Site "?"]
[Date "2022.01.03"]
[Round "?"]
[White "SFxxxxxx 211221"]
[Black "SFxxxxxx 211221"]
[Result "0-1"]
[TimeControl "60/15"]

1. e4 a6 2. f4 c6 3. Qf3 d5 4. e5 Bd7 5. a3 Be6 6. d4 a5 7. f5 Bd7 8. Bd3 Nh6
9. c3 c5 10. dxc5 Nxf5 11. c6 Nxc6 12. Bxf5 Bxf5 13. Qxf5 a4 14. Bf4 Qb6 15. b4
axb3 16. e6 Nd8 17. Be5 fxe6 18. Qf3 b2 19. Ra2 Nc6 20. Qh5+ Kd7 21. Ne2 Qb3
22. Rxb2 Qxb2 23. O-O g6 24. Nd4 gxh5 25. Nxc6 Rg8 26. g3 Kxc6 27. a4 Bh6 28.
Kh1 Rxa4 29. Re1 Ra2 30. g4 Qg2# 0-1
[/pgn]
User avatar
AlexChess
Posts: 1562
Joined: Sat Feb 06, 2021 8:06 am
Full name: Alex Morales

Re: Kayra 1.1

Post by AlexChess »

DrEinstein wrote: Mon Jan 03, 2022 12:19 pm
I have just forked SF and modified it a bit. Working name is StockfishXXXXXX and XXXXXX is private. Depending on its additional settings, it plays exactly like SFdev or completely different from all derivatives, with less than 900 Elo if necessary. :lol:
So, despite your nick you are nobody. Don't fill this Mehmet topic with garbage, go to work.
Chess engines and dedicated chess computers fan since 1981 :D macOS Sequoia 16GB-512GB, Windows 11 & Ubuntu ARM64.
ProteusSF Dev Forum