ShashChess

[peter]

ShashChess 4.0 released
https://github.com/amchess/ShashChess/releases/tag/4.0

Thanks a lot, Andrea, for continuing your work.
Just one question: UCI- Options now have Analysis Contempt to be switched of and on for each side, but I don't see Contempt itself to be set. Does that mean, Shashin- Options checked work in Analysis Mode only if Analysis Contempt is checked for any or both sides too, or also with AC off?
Does it depend?

[Dariusz]

Below macOS exec.
Thx for new version
https://www.solidfiles.com/v/QdkBPxgr2gXmM

[maxdeg]

Below macOS exec.
Thx for new version
https://www.solidfiles.com/v/QdkBPxgr2gXmM

+1

[amchess]

ShashChess 4.0 released
https://github.com/amchess/ShashChess/releases/tag/4.0

Thanks a lot, Andrea, for continuing your work.
Just one question: UCI- Options now have Analysis Contempt to be switched of and on for each side, but I don't see Contempt itself to be set. Does that mean, Shashin- Options checked work in Analysis Mode only if Analysis Contempt is checked for any or both sides too, or also with AC off?
Does it depend?

The contempt value is implicitly initialized by Shashin's options.
Based on our tests, the best values are the following, as you can see in file types.h:
default, capablanca,tal-petrosian, tal-capablanca-petrosian -> 12
petrosian->-6
capablanca-petrosian->3
tal-capablanca->21
tal->30

In general, this is also, in my opinion the interest of Shashin's theory: replace too technical uci options (like, for example, also null move, lmr, etc), with more meaningful and user friendly and flexible ones:
numerical values should be for developers/testers, not for users!

[amchess]

Below macOS exec.
Thx for new version
https://www.solidfiles.com/v/QdkBPxgr2gXmM

+1

Many thanks!
I added it.
Can you send me a PM with your first and last name?
So, I can add you in the acknowledgement section of the readme file.
Andrea

[peter]

numerical values should be for developers/testers, not for users!

Thanks again, for your prompt answer too.

By myself I'm not as much interested in Elo as in Analysis Mode and in thematic games and, from certain positions of interest, played out lines.

Guess I already found out how to configure ShashChess4 for Analysis Mode and "biased" game-playing postion- dependent, having already some experience with ShashChess3.

Yet greatest fun for me anyhow is always to find best settings of engines for a certain single position.


BTW notice: strength in chess (of engines and of humans) is always postion- dependent anyhow, there isn't something like "overall playing strength" in reality, at least not in an overall reproducibly testable way.

Elo, measured in the way usable for engines today, the better the engines get, is more and more positional testing of strength at bookless game- playing only, even if started from positions seemingly only more or less near to the character of the very first starting position of chess itself.

The more modern engines have positions near to very first starting position already in hash in a few seconds of computing or at once in NN at starting their rating- games, even almost as much as from the one and only real unique Initial Position, the more they get selected Elo- wise by game- playing of that kind only, for their rating against other engines and against themselves.

That's the real dead end street of engine- development by rating measured in Elo only, as we are used to test engines, and it's not a numeric problem, you can choose any other measurement instead of Elo as well, it's a problem of testing- method and testing- positions (starting positions for the rating games).
Another word of principally, even if not fully the same meaning, to me is "the draw problem".

[amchess]

I updated executables because of a problem with Move Overhead uci option.
I added it because the default value (30) is good for Linux, not for windows, especially for slow guis.
So, for windows, based on my tests, it's better to set it at 100, for game play.
The windows and Linux versions now are updated.
I wait for an update of Android and Mac.
Best regards and sorry for the inconvenience: I'm an ubuntu fan...
Andrea

[Marek Soszynski]

It's fun to run 3 separate instances at the same time: Capa, Petr, and Tal.

However, when Analysis Contempt is set to "Both", the Petr score is 1 or 2 pawns higher than the others, even at the start.

[amchess]

The Android executables have been updated.
I wait for mac version.

[amchess]

Now all executables are good