Kayra 1.1

[AlexChess]

I've been working on the Kayra engine for only 1.5 months, mostly in a limited time. The negative effect of making more than a hundred lines code changes in the search part is the inability to adequately test which code changes are good and which code changes have a negative effect. I still hope to see Kayra in better places when I find and replace the codes that negatively affect the performance and add more powerful codes in the future.

According to some test results, the situation is not hopeless for Kayra.
https://sites.google.com/site/computers ... ayra-vs-sf
https://sites.google.com/site/computers ... ct-nn-cs-8

Also my SuperBlitz confirms these incredible results of Kayra 1.0. Please add polyglot support so I can test Kayra 1.1 with Eduard's Kayra.bin book

[mehmet123]

I've been working on the Kayra engine for only 1.5 months, mostly in a limited time. The negative effect of making more than a hundred lines code changes in the search part is the inability to adequately test which code changes are good and which code changes have a negative effect. I still hope to see Kayra in better places when I find and replace the codes that negatively affect the performance and add more powerful codes in the future.

According to some test results, the situation is not hopeless for Kayra.
https://sites.google.com/site/computers ... ayra-vs-sf
https://sites.google.com/site/computers ... ct-nn-cs-8

Also my SuperBlitz confirms these incredible results of Kayra 1.0. Please add polyglot support so I can test Kayra 1.1 with Eduard's Kayra.bin book

The important thing for me is to be able to develop Kayra for now and that's why I act like Kayra is a very weak engine. So far, I have never had an assertive speech about Kayra because it's too early for that. There are so many things to be developed and changed in Kayra. The only problem for me is the time problem.

[Guenther]

I've been working on the Kayra engine for only 1.5 months, mostly in a limited time. The negative effect of making more than a hundred lines code changes in the search part is the inability to adequately test which code changes are good and which code changes have a negative effect. I still hope to see Kayra in better places when I find and replace the codes that negatively affect the performance and add more powerful codes in the future.

According to some test results, the situation is not hopeless for Kayra.
https://sites.google.com/site/computers ... ayra-vs-sf
https://sites.google.com/site/computers ... ct-nn-cs-8

Oh my god, you are no programmer, why pretending so?
You have not changed/written any line of code, just changed some numbers you found, first in the eval, this lost around 800 elo IIRC (now reverted) and now a little bit the search, still nnue makes by far most of its strength.

If I change nothing in the source of SF it is as strong as it is. If I change 1-X random numbers (in the right place - better said, not too important
places for now, with nnue) it will get along with 95-99% of the strength of SF dev.
Why should someone care for those useless 'changes'? (except those who are blinded by whatever drivel and are no programmers)

And the quoted link from SSCT 'tests' now against much older SF 14.1. Is this a conspiracy plot aka 'make Kayra (nearly) as great as SF (not again)'
from a sect of believers? Ofc your sect is much worse than you, but without those handful of crazy people those threads would have found
their deserved end since long... It can even be supposed that some just prolong this Kayra testing for trolling in their xmas boredom.

[Eduard]

I've been working on the Kayra engine for only 1.5 months, mostly in a limited time. The negative effect of making more than a hundred lines code changes in the search part is the inability to adequately test which code changes are good and which code changes have a negative effect. I still hope to see Kayra in better places when I find and replace the codes that negatively affect the performance and add more powerful codes in the future.

According to some test results, the situation is not hopeless for Kayra.
https://sites.google.com/site/computers ... ayra-vs-sf
https://sites.google.com/site/computers ... ct-nn-cs-8

Also my SuperBlitz confirms these incredible results of Kayra 1.0. Please add polyglot support so I can test Kayra 1.1 with Eduard's Kayra.bin book

And be sure, I can then eliminate all errors in the book and make the book better.

[mehmet123]

You have not changed/written any line of code, just changed some numbers you found, first in the eval, this lost around 800 elo IIRC (now reverted) and now a little bit the search, still nnue makes by far most of its strength.

You are still talk about Kayra 1.0. Yes in Kayra 1.0 code changes is very limited. But I talk about Kayra 1.1.
I explained before why I weakened Kayra 1.0 classic version so much with parameter changes. I don't want to talk about the same things all the time.

[Guenther]

You have not changed/written any line of code, just changed some numbers you found, first in the eval, this lost around 800 elo IIRC (now reverted) and now a little bit the search, still nnue makes by far most of its strength.

You are still talk about Kayra 1.0. Yes in Kayra 1.0 code changes is very limited. But I talk about Kayra 1.1.
I explained before why I weakened Kayra 1.0 classic version so much with parameter changes. I don't want to talk about the same things all the time.

I speak about Kayra 1.1 I have the sources and compiled it. Are you confused?
Do you know what it means to write additional own code (not engaging or mixing older SF or SF forks code) at all?

[RubiChess]

I wonder if anyone can setup some "SF derivation" automata that generates 2-3 SF clones per day by shuffling some parameters.
Enough to occupy testing capacity of our SSS-testing friends Alex, Eduard (who should take care of his Vitamin D) etc. so that they have fewer to no time to post anymore.

[mehmet123]

I speak about Kayra 1.1 I have the sources and compiled it. Are you confused?
Do you know what it means to write additional own code (not engaging or mixing older SF or SF forks code) at all?

The only thing left to do is wait for Stockfish team to make a change table for Kayra 1.1 as they did with Kayra 1.0. Otherwise, it is pointless to continue this discussion.

[mehmet123]

And the quoted link from SSCT 'tests' now against much older SF 14.1. Is this a conspiracy plot aka 'make Kayra (nearly) as great as SF (not again)'
from a sect of believers? Ofc your sect is much worse than you, but without those handful of crazy people those threads would have found
their deserved end since long... It can even be supposed that some just prolong this Kayra testing for trolling in their xmas boredom.

Is the newest version is always the strongest version?
Is the difference between the two Stockfish versions in the real world is as high as Fishtest?

http://www.computerchess.org.uk/ccrl/40 ... t_all.html
https://ssdf.bosjo.net/list.htm

[amanjpro]

Guether, you and Sopel should ask yourselves what Mehmet did wrong in this thread? Then nobody needs to try something new anymore. We just want to have fun and enjoy developing and testing. Unfortunately, you are also one of the biggest critics here. I don't know why we should still do anything at all.

I don't understand your point. I asked what additional features this engine has (and received no response yet), because it provides no actual when compared to Stockfish as it is now. What's the point of cloning Stockfish and just making it worse? What did he do wrong you want to know? I want to know what did he do right. There is nothing in this engine that's useful, considering the alternatives.

Btw. that test someone requested is on the way, 300+10 with uho book, partial result:

Code: Select all

Score of sf_20211227 vs kayra_1.1: 352 - 335 - 565  [0.507] 1252

Which other features has Fat Titz? Exactly like Kayra, despite few changes it seems a very strong SF derivative. Most people is saying the same about it and I have always defended you because I like derivatives that are perfecty allowed by GPLv3 and MIT and offer a collaborative way to improve all the engines and Stockfish itself to skilled developers (not all derivatives are in TOP 10)

Little exercise for you: How can an engine violate the MIT license?