Chess2u
Would you like to react to this message? Create an account in a few clicks or log in to continue.

Chess2uLog in

Proteus Chess SF - developing my Stockfish 14.1 derivative.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
In my Android Proteus V006 is one of the fastest kn/s along with BM 14.9.. it ran arround 250k kn/s on my Drpidfish, using 8 cores and performance mode at my phone..

They both leaving all the SF dev builds in behind for the speed issues..

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
How to use strongest opening book. ? Again SF 007 and blue marlin 14.9 ; swordfish chess engine..and other chess engine.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
AlexChess wrote:
Started final tests with Windows 11 ARM64 using official SF PB 220401

since SF PB doesn't add nothing to the engine strength and I prefer Stockfish Team official builds.

Lol, where exactly you saw SF-PB Official? The only official is Stockfish on Github (and Stockfish Polyglot by Masimiliano, again on Github)..
Regarding SF PB - who said that the engine is stronger than normal SF, the only difference is that SF PB is able to work with .bin opening books, engine is for book testers. In my opinion, you should have known the difference a long time ago.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
PROTEUS SF JBE 007 against KAYRA 1.3 - RAPID CHESS 10 mins + 2 secs / ALL  
Android Snapdragon 626 & 660 8 CPUs 400-500 Kn/s LEARNING ENABLED
100 Games 8moves_v3.bin & Perfect2021.bin books swapping colors
Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 KMkpDTgA

Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 XpETxiRa

ALL PGN GAMES WITH 8moves_v3.pgn : https://pixeldrain.com/u/M1tVc39S

Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 4CMBdzgZ

ALL PGN GAMES WITH Perfect2021.pgn : https://pixeldrain.com/u/HAMZu2M3


I think that Kayra 1.3 of my friend Mehmet is stronger than SF PB 220401 !! I fear next Kayra release  smile

Android build tests completed. I will release Proteus SF JBE 007 for Windows, Android and Mac M1 for Easter

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Proteus SF JBE 007 against SF PB 220401 on Windows 11 ARM64 under
Parallels Desktop 17.1 for Mac M1 4 CPUs 2 Mn/s

TCEC 21 PGN book balanced (not Jeroem Noomen selection)
Tablebases  Syzygy 6 MAN  TC: 3 mins + 2 secs / ALL

Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 8HXpMERt

Due to the learning code, Proteus SF JBE 007 is a little slower than SF PB 010422. I'll try to compile it also for native Windows ARM64
to verify speed differences and performances repeating this match. For now, I start compiling final builds for all Windows x64 flavours.

Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 FqWsVrqR

ALL PGN GAMES: https://pixeldrain.com/u/h57ChDfE

NOTE for @homayoun and my other friends here: : Instead of re-compiling SF PB 010422 like I've done for Mac M1 on Windows, to be sure that the match is fair and reliable I have used the build SF-PB-010422_x64_SSE41_popcnt.exe found on the official .zip[/b] from the author and Proteus-SF-JBE-007-sse41-popcnt.exe .

Balanced opening books are very boring, I prefer dorsz's DOTS4 PGN test suite or Stefan Pohl anti-draw books to verify my improvements, but to feed experience.bin (instead of usual trolls  smile) I'm using also most common books
used on other matches by @Oliprog @Ghppn , Canbaz and other testers and I will release also updates for the experience.bin files for Android, Mac and Windows, based on them.

On your tests, please report also opening book used and Mn/s of your hardware! smile

Kind regards, Alex

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
As always, here I cannot edit the post rolleyes but the opening book was 8moves_v3.pgn BALANCED ...I'm using TCEC 21 in another test.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Hi Alex
I can’t understand what is the SFPb official? I haven’t seen any official release of this engine. Some versions have been compiled by chessman and some of them specially newer updates have been compiled by chess fan1. The bin book support has been provided by MZ. And the engine was resulted from the cooperation of these three persons . But I haven’t seen official release of this engine.
Best regards

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
AlexChess wrote:
As always, here I cannot edit the post rolleyes  but the opening book was 8moves_v3.pgn BALANCED  ...I'm using TCEC 21 in another test.

Well, i've said that hundreds of times. 100-1000 games between two engines identical in strength means nothing.. It's like a drop in the ocean. The best opening book (which SF team is using) is 8moves_v3.pgn which contains more than 5000+ lines, so all your test (100-200 games) with accuracy of 0% are strictly for fun and nothing else.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
AlexChess wrote:
As always, here I cannot edit the post rolleyes  but the opening book was 8moves_v3.pgn BALANCED  ...I'm using TCEC 21 in another test.


Why are you using these stupid openings from TCEC ?
Do you think it's okay ?
Lots and lots of garbage and junk lines in TCEC...

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Homayoun wrote:
Hi Alex
I can’t understand what is the SFPb official? I haven’t seen any official release of this engine. Some versions have been compiled by chessman and some of them specially newer updates have been compiled by chess fan1. The bin book support has been provided by MZ. And the engine was resulted from the cooperation of these three persons . But I haven’t seen official release of this engine.
Best regards


I agree with you, I was referring to the fact that I have used a not modded (official) binary that it could be ruined by me  smile  

...but then, why tributing honors to chessfan1| skynet | 135657 | Aleksey | Alex (he has a great fantasy in changing his name  rolleyes  )
if he hasn't changed nothing to improve the engine strength ??? eek

Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 4STXWuvE

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
ZamChess wrote:
AlexChess wrote:
As always, here I cannot edit the post rolleyes  but the opening book was 8moves_v3.pgn BALANCED  ...I'm using TCEC 21 in another test.


Why are you using these stupid openings from TCEC ?
Do you think it's okay ?
Lots and lots of garbage and junk lines in TCEC...


Why you doesn't like TCEC? I think instead that they are very skilled. Jeroen Noomen and the others give a free and reliable service to the whole chess community since years, using powerful hardware and a statistical approach SUPER PARTES. TCEC pgn balanced for qualifications and Noomen select UNBALANCED positions for superfinals that are really funny to watch and analyse in live chat, not a boring collection of games full of blunders due to the lack of time using weak hardware and concurrent games only to speed up the process.

https://tcec-chess.com/

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Sometimes the goal of a new engine is not strength. For example we have crystal engine (I like it very much) and the goal of engine is to play more tactical than stockfish itself and fortress detecting in endgames although it is not stronger than stockfish but many chess fans like it just for nice playing style. The goal of SFpb is to add ability of use bin books to stockfish and nothing else. It was a free alternative for stockfish polyglot which wasn’t free for a period of time. And the authors never claim that it is stronger than stockfish official.
Best regards

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
ZamChess wrote:
AlexChess wrote:
As always, here I cannot edit the post rolleyes  but the opening book was 8moves_v3.pgn BALANCED  ...I'm using TCEC 21 in another test.


Why are you using these stupid openings from TCEC ?
Do you think it's okay ?
Lots and lots of garbage and junk lines in TCEC...


What I tend to do is edit the Sufi lines to get rid of the junk and combine them into one customized TCEC suite. Sort of like what Stefan Pohl does with his Noomen SuFi LowDraw openings - except, in my case, I strive to have a mix (but never ~2-3 ply openings). I believe unbalanced openings have a role in engine development, but they should be used together with balanced lines and other tools such as test suites, etc. Mehmet made an excellent observation over at Talkchess where he said

"We only know that Stockfish is better than Dragon in unbalanced openings and it has a more aggressive playstyle. I don't think Stockfish is actually more powerful than Dragon more than 10 - 20 elo at long time controls with balanced openings."

I totally agree. Reliance on unbalanced openings distorts results.

And ChessFan1 is right here, too, IMHO.

". . . 100-1000 games between two engines identical in strength means nothing.. It's like a drop in the ocean. . . "

This is also the case with building experience. Except in that case, you need to focus on methods deeds has proven to be effective. These things take time. I know because I have used them. These tools are even being used to revitalize the old Houdini 6.03 engine incorporating a type of neural network.

Don't rush your work, Alex. Focus on what you want to achieve and then release new versions. This is supposed to be fun, my friend! smile

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Homayoun wrote:
Sometimes the goal of a new engine is not strength. For example we have crystal engine (I like it very much) and the goal of engine is to play more tactical than stockfish itself and fortress detecting in endgames although it is not stronger than stockfish but many chess fans like it just for nice playing style. The goal of SFpb is to add ability of use bin books to stockfish and nothing else. It was a free alternative for stockfish polyglot which  wasn’t free for a period of time. And the authors never claim that it is stronger than stockfish official.
Best regards


I like Crystal, too, and Sting (which may no longer be in development). Joe Ellis's fortress detection code is part of Phaedra. Some of Mike Byrne's ideas, too. These types of engines are more for analysis or problem solving. Not for engine v engine play. That was never my goal. Creating a derivative with a unique style is an idea I have tried to encourage Alex to explore.

I think some people get confused because they see SF PB and Stockfish in the same tournament (no bin books used). SCCT, for example. If you examine the source, only polyglot code has been added to SF PB.  I think what he does is commendable (I, too, miss Brainfish. It was my favorite engine along with Houdini). It lets people use books with Stockfish. He publishes the source and respects the GPLv3. Too bad everybody doesn't do that.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
ZamChess wrote:
AlexChess wrote:
As always, here I cannot edit the post rolleyes  but the opening book was 8moves_v3.pgn BALANCED  ...I'm using TCEC 21 in another test.


Why are you using these stupid openings from TCEC ?
Do you think it's okay ?
Lots and lots of garbage and junk lines in TCEC...

He still does not know this, it takes him a long time to realize (lol). All the TCEC pgn books are pretty much the same (with only some main opening lines), they was designed especially for the TCEC matches, where the engines run on ultra-modern computers with a game duration of more than two hours, respectively, the duration of the tests takes months.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Hi Alex,
What's the strenght of Proteus JB 007 without book against the others, have you done a test.
It's seem that the best way to create a book based on Proteus, it's to create his own book based on his games against the others.
Kayra 1.1 bin was created like this. And if you have don't use it already, i can assure you that's a very very strong book.

Friendly, Sékou

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
AlexChess wrote:
Homayoun wrote:
Hi Alex
I can’t understand what is the SFPb official? I haven’t seen any official release of this engine. Some versions have been compiled by chessman and some of them specially newer updates have been compiled by chess fan1. The bin book support has been provided by MZ. And the engine was resulted from the cooperation of these three persons . But I haven’t seen official release of this engine.
Best regards


I agree with you, I was referring to the fact that I have used a not modded (official) binary that it could be ruined by me  smile  

...but then, why tributing honors to chessfan1
| skynet | 135657 | Aleksey | Alex (he has a great fantasy in changing his name  rolleyes  )
if he hasn't changed nothing to improve the engine strength ???

First of all, i've said it to you already via PM, i'm also known as "Nick" from Britain, "Jason" from Cyprus and many more. lol You still doesn't understand why people are using "nicknames", aren't you? I hope you understand one day, just don't be late.

Secondly Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 Wall, i repeat for the slow-witted (also read my post above). I asked Marco to help me with adding support for opening books in the source code, he did that. The idea was to make an engine in memory of Brainfish, there is not any extra strength / elo gain, can you understand that? Engine is for book testers.. Btw, who is tributing honors to me (lol)? A lot of people don't know how to compile the engine, i'm simple helping them, that's all.. People who just want to thank me for this, they do, end of story. Although you probably won't understand it.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
ChessFan1 wrote:

. . .. All the TCEC pgn books are pretty much the same (with only some main opening lines), they was designed especially for the TCEC matches, where the engines run on ultra-modern computers with a game duration of more than two hours, respectively, the duration of the tests takes months.


If you get rid of the absurd lines, many of the TCEC openings are still very drawish - even on a Threadripper. There is a poster on the forum you help moderate who has very high draw rates using the Sufi 21 test set. Nothing wrong with that. Nor is using unbalanced openings such as DOTS or UHO. Just do not try and make an elo assessment using them as your sole reference. They are tools that can and should be used to evaluate engine strengths and weaknesses. Even the highly respected Peter Grayson would tell you that.

dorsz, the creator of the DOTS suites, clearly says this in a post on OpenChess,

"Dynamic Openings Test Suite (DOTS) is a set of selected positions for running engine matches. The purpose of this set is not to measure Elo differences but to provide dynamic and interesting games."

If you want to create an engine/derivative with a unique style, you need to diversify your methods of testing. If you are just trying to create another clone engine of Stockfish (which is usually weaker in the end  bigsmile ), then carry on with whatever illogical means you prefer. . .

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
A review of one of the members of my FB page about new Proteus 007 JBE : Proteus is the strongest engine for now i think... He has completed skill. He can play tactical, he can play positional, he can make good pawn structures for winning in the endgame againts strong opponent. He have positive evaluation and he confidence he can win in every game, not only play for equalize the position like another engine but he play for winning in every position againts strong opponent.

Nice work!

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
ProteusSF JBE 007 +12 OVER SF PB 220401 in a 100 games Blitz 3 mins + 2 secs / all on Windows 11 4 CPUs 2 Mn/s!

TCEC 2021 Joroen Noomen select book switching color Tablebases Syzygy 6 Man LEARNING ENABLED


Proteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 PhpwDJqPT

ALL 100 games in PGN: https://pixeldrain.com/u/3fbtwG84

April 17, 2022 - ProteusSF JBE 007 RELEASED for Windows x64, Android and Mac Silicon M1!

DOWNLOAD IT HERE : bit.ly 3HLexsP

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
sarona wrote:
ChessFan1 wrote:

. . .. All the TCEC pgn books are pretty much the same (with only some main opening lines), they was designed especially for the TCEC matches, where the engines run on ultra-modern computers with a game duration of more than two hours, respectively, the duration of the tests takes months.


If you get rid of the absurd lines, many of the TCEC openings are still very drawish - even on a Threadripper. There is a poster on the forum you help moderate who has very high draw rates using the Sufi 21 test set. Nothing wrong with that. Nor is using unbalanced openings such as DOTS or UHO. Just do not try and make an elo assessment using them as your sole reference. They are tools that can and should be used to evaluate engine strengths and weaknesses. Even the highly respected Peter Grayson would tell you that.

dorsz, the creator of the DOTS suites, clearly says this in a post on OpenChess,

"Dynamic Openings Test Suite (DOTS) is a set of selected positions for running engine matches. The purpose of this set is not to measure Elo differences but to provide dynamic and interesting games."

If you want to create an engine/derivative with a unique style, you need to diversify your methods of testing. If you are just trying to create another clone engine of Stockfish (which is usually weaker in the end  bigsmile ), then carry on with whatever illogical means you prefer. . .


Hi Ron! TCEC books, like DOTS4 are only for testing purposes. Anyway swapping sides and repeating each position make the match fair. People can use every opening book with it, but I'm more interested in how Proteus SF calculates positions than in books. If you report matches, give/tell me also the opening book used, hardware , OS ad Mn/s if you want that I will use them to improve the next version.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
sarona wrote:
ChessFan1 wrote:

. . .. All the TCEC pgn books are pretty much the same (with only some main opening lines), they was designed especially for the TCEC matches, where the engines run on ultra-modern computers with a game duration of more than two hours, respectively, the duration of the tests takes months.


If you get rid of the absurd lines, many of the TCEC openings are still very drawish - even on a Threadripper. There is a poster on the forum you help moderate who has very high draw rates using the Sufi 21 test set. Nothing wrong with that. Nor is using unbalanced openings such as DOTS or UHO. Just do not try and make an elo assessment using them as your sole reference. They are tools that can and should be used to evaluate engine strengths and weaknesses. Even the highly respected Peter Grayson would tell you that.

dorsz, the creator of the DOTS suites, clearly says this in a post on OpenChess,

"Dynamic Openings Test Suite (DOTS) is a set of selected positions for running engine matches. The purpose of this set is not to measure Elo differences but to provide dynamic and interesting games."
----------------------
If you want to create an engine/derivative with a unique style, you need to diversify your methods of testing. If you are just trying to create another clone engine of Stockfish (which is usually weaker in the end  bigsmile ), then carry on with whatever illogical means you prefer. . .

I agree 100%, the only thing I can add is that for testing the engine for strength / elo gain, it is best to use one book for example 8_moves_V3, it is very hard to get elo gain with it, but, if you got elo gain - it will be real for sure, with other books like UHO the elo gain will be even more. Also a chess engine need to be tested with at least 2 different time controls. Yep, this require a lot of time and resources, but it is the only guaranteed way.
----------------------
This is dedicated to Alex Morales, , with his whatever changes in code, and his whatever 100 games per test.

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Hi Alex can u provide another link to download ? I have a difficult ads problem to acces the current link u gave

Thank u

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Just follow the engine...https://www.facebook.com/Motores-de-ajedrez-AC-112015897649644

Regards;

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Sekou82 wrote:
Hi Alex,
What's the strenght of Proteus JB 007 without book against the others, have you done a test.
It's seem that the best way to create a book based on Proteus, it's to create his own book based on his games against the others.
Kayra 1.1 bin was created like this. And if you have don't use it already, i can assure you that's a very very strong book.

Friendly, Sékou


Thank you for the info. I can try, but which book (or no book) for the others? I could add Solista-Eduard's Kayra 1.1.bin for Mehmet's engine.
I think that Kayra 1.3 is stronger than Stockfish at least against ProteusSF and I want to train my engine against it smile

Kind regards, Alex

descriptionProteus Chess SF - developing my Stockfish 14.1 derivative. - Page 5 EmptyRe: Proteus Chess SF - developing my Stockfish 14.1 derivative.

more_horiz
Permissions in this forum:
You cannot reply to topics in this forum