ShashChess

[amchess]

ShashChess 5.0 released
https://github.com/amchess/ShashChess/releases/tag/5.0

[amchess]

I wait, like usual, for the MAC and Android version.
Many thanks in advance.
Andrea

[Eduard]

Thank You very much! I saw that there is an Android Version of ShashChess 2 in a message before that. Would it also be possible to compile this version of ShashChess 5.0 for Android?

I'm very surprised about the strength.

On my Laptop runs currently a match against Stockfish 220119.

Each engine on 1 core, ponder off, hash 64 Mb, 5 move book, syzygy 5 men, timecontrol blitz 7m +3s. ShashChess 5.0 plays with style (mix) Tal+Capablanca+Petrosian.

After 20 games ShashChess leads with +1 (+1 =19 -0). ShashChess 5.0 is really a little bit stronger than this last Stockfish dev. version.

[Guenther]

After 20 games ShashChess leads with +1 (+1 =19 -0). ShashChess 5.0 is really a little bit stronger than this last Stockfish dev. version.

You can't be serious? 20 games with 19 draws allow no conclusion at all.
(Actually latest SF will always be stronger, as it is 99.9% of the code base and is always ahead in time.)

[Eduard]

I am testing against Stockfish 220119. I do not know why this latest Stockfish losed? One everyone can test it self. After 30 games ShashChess 5.0 losed not one game! Currently standing is +2 =28 -0.

[RubiChess]

I am testing against Stockfish 220119. I do not know why this latest Stockfish losed? One everyone can test it self. After 30 games ShashChess 5.0 losed not one game! Currently standing is +2 =28 -0.

As Guenther told you: Even 30 games don't say anything. ShashChess is probably (almost) latest Stockfish with some few changes so it is expected to play as strong as Stockfish. So you have to play at least a few thousand games to get a conclusion which one is really better.

./Andreas

[amchess]

My goal was simply to demostrate Shashin theory can improve the base Stockfish and overall render it more suitable at long times, for example, in correspondence play.
In my humble opinion, this theory is very fecund, for example, to
-improve the engine strength at long time controls
-broaden its tactical vision (even now, Houdini tactical is better than Stockfish!)
-implement new ideas like mcts, etc
I'm working in with 6 version almost ready.
In a private message, a great contributor of Stockfish community told me this:
" A great work: hundred or thousand lines of code respect to Stockfish!"
Anyway, the code is GPL, not a mistery. So, it's not too difficult for anyone, even with not a lot of programming skill, to compare the difference with respect to the base engine in terms of lines of code, but overall of the weight of my work/idea.
Finally, I'm only interested in long times.
So, it's normal, at short times, Stockfish is better and it's crazy to evaluate the engine with a lot of short RANDOM time control games!
In my opinion, the best, surely hard in terms of time, but ONLY WAY to seriously evaluate it and similar engines is illustrated on my wiki: https://github.com/amchess/ShashChess/wiki/Tests
Surely, there are a lot of improvements/corrections to do: nobody is perfect!
So, any general advice or critics BASED ON FACTS is welcome.
Andrea

[RubiChess]

So, any general advice or critics BASED ON FACTS is welcome.

It wasn't my intention to offend you or your engine. My "...latest Stockfish with some few changes..." was just a rough guess and I must admit that I didn't looked at your source code. My post was only meant to argue about the "A vs B +2 =28 -0 => A better than B" statement.

Keep up your probably good work.

. Andreas

[amchess]

So, any general advice or critics BASED ON FACTS is welcome.

It wasn't my intention to offend you or your engine. My "...latest Stockfish with some few changes..." was just a rough guess and I must admit that I didn't looked at your source code. My post was only meant to argue about the "A vs B +2 =28 -0 => A better than B" statement.

Keep up your probably good work.

. Andreas

No problem: it was just a clarification.
All the best,
Andrea

[Eduard]

The goal of my test was not to find out who is better. I wanted to see the style of ShashChess when it plays. And it seemed to me most interesting to combine all three styles. Of course I read that ShashChess is not optimized for short time controls. Personally, I'm more interested in analysis.
And here I really liked ShashChess. The evaluation of ShashChess is really good! I'm honest: if I had not liked the engine in analysis, then I would have stopped testing.

Thanks for this wonderful engine. It will be very useful to me in analyzing.