ShashChess

[amchess]

https://github.com/amchess/ShashChess/w ... te-results

Seriously, I don't believe these results - test results or the...think it was 15 games with a lopsided result towards ShashChess played between the two.
There must be something I am missing... (?)

These are not matches, but hard positions selected from various forums and ranked according to Shashin's theory.

[Cornfed]

https://github.com/amchess/ShashChess/w ... te-results

Seriously, I don't believe these results - test results or the...think it was 15 games with a lopsided result towards ShashChess played between the two.
There must be something I am missing... (?)

These are not matches, but hard positions selected from various forums and ranked according to Shashin's theory.

Right, chosen positions. But I am also referring to the 15 games mentioned above. Were they hand picked openings for the same purpose? I did not play thru them (yet).

[amchess]

They are 20 games for LTC.

[Cornfed]

They are 20 games for LTC.

Okay, there are 20.
Specifically designed to highlight Shanshin theory implementation?

I was just wondering why the results were SO lopsided in favor of ShashChess.

[connor_mcmonigle]

They are 20 games for LTC.

Okay, there are 20.
Specifically designed to highlight Shanshin theory implementation?

I was just wondering why the results were SO lopsided in favor of ShashChess.

With only 20 games, these results are almost certainly going to be impossible to reproduce anyways. Meaningless

[Cornfed]

They are 20 games for LTC.

Okay, there are 20.
Specifically designed to highlight Shanshin theory implementation?

I was just wondering why the results were SO lopsided in favor of ShashChess.

With only 20 games, these results are almost certainly going to be impossible to reproduce anyways. Meaningless

I disagree.
If indeed the idea of the games and test positions is to 'showcase' somehow the Shashin theory, then all you have to do is find another 20 openings leading to positions would allow the same.
There is nothing wrong with that of course. It simply is what it is.

[connor_mcmonigle]

They are 20 games for LTC.

Okay, there are 20.
Specifically designed to highlight Shanshin theory implementation?

I was just wondering why the results were SO lopsided in favor of ShashChess.

With only 20 games, these results are almost certainly going to be impossible to reproduce anyways. Meaningless

I disagree.
If indeed the idea of the games and test positions is to 'showcase' somehow the Shashin theory, then all you have to do is find another 20 openings leading to positions would allow the same.
There is nothing wrong with that of course. It simply is what it is.

My point was that, were you to run this same match again, you could get drastically different results. You're looking at what amounts to a coin toss.

[Cornfed]

They are 20 games for LTC.

Okay, there are 20.
Specifically designed to highlight Shanshin theory implementation?

I was just wondering why the results were SO lopsided in favor of ShashChess.

With only 20 games, these results are almost certainly going to be impossible to reproduce anyways. Meaningless

I disagree.
If indeed the idea of the games and test positions is to 'showcase' somehow the Shashin theory, then all you have to do is find another 20 openings leading to positions would allow the same.
There is nothing wrong with that of course. It simply is what it is.

My point was that, were you to run this same match again, you could get drastically different results. You're looking at what amounts to a coin toss.

The same time control, exact same openings - leaving book at the position they did again and... instead of 4 wins, 1 loss and 15 draws...it could just as well end 1 win, 4 losses and 15 draws or some other 'drastically different' result? No, I don't think so. Some of the factors would need to change.

I believe they simply chose specific opening positions to play just as they say they did for the test positions...just to demonstrate Shashin theory at work in ShashChess.

[peter]

These are not matches, but hard positions selected from various forums and ranked according to Shashin's theory.

I put together HTC114 from Vincent Lejeune, ERET from Walter Eigenmann, Arasan- testsuite from Jon Dart, remove the doubles and add 77 more out of my personal collection of difficult positions to have 500 all in all.

With these I run A-B-engines with 30 threads of an 16x3,5GHz avx2- CPU (or GPU- based ones with an NVIDIA 3070ti) 10"/positions, 1G hash, default settings (ShashChess without experienc- file and all GM- options unchecked), Syzygy Probe Limit 5.
Here both had the latest green net from SF- dev- site:

Code: Select all

    Program                                    Elo   +/-  Matches  Score   Av.Op.   S.Pos.   MST1    MST2   RIndex

  1 ShashChess18.2-nn13406b1dcbe0.nnue       : 3513    9    405    53.6 %   3487   399/487    2.0s    3.4s   0.94
  2 Stockfish170921                          : 3487    9    405    46.4 %   3513   354/487    1.8s    4.0s   0.86



MST1  : Mean solution time (solved positions only)
MST2  : Mean solution time (solved and unsolved positions)
RIndex: Score according to solution time ranking for each position

Evaluation is done with ELOSatTS (TS=Test Suite) from Frank Schubert, counting not only the solved positions, but also the time to solution of "matched" positions, those are all, that are solved by two engines that are matched against each other that way.
Postions, that aren't solved by any opponent, aren't counted, so here 487 instead of the full 500 are rated.
You have to define a starting value, here I gave 3500 Elo as an estimation to start with regards

[connor_mcmonigle]

They are 20 games for LTC.

Okay, there are 20.
Specifically designed to highlight Shanshin theory implementation?

I was just wondering why the results were SO lopsided in favor of ShashChess.

With only 20 games, these results are almost certainly going to be impossible to reproduce anyways. Meaningless

I disagree.
If indeed the idea of the games and test positions is to 'showcase' somehow the Shashin theory, then all you have to do is find another 20 openings leading to positions would allow the same.
There is nothing wrong with that of course. It simply is what it is.

My point was that, were you to run this same match again, you could get drastically different results. You're looking at what amounts to a coin toss.

The same time control, exact same openings - leaving book at the position they did again and... instead of 4 wins, 1 loss and 15 draws...it could just as well end 1 win, 4 losses and 15 draws or some other 'drastically different' result? No, I don't think so. Some of the factors would need to change.

I believe they simply chose specific opening positions to play just as they say they did for the test positions...just to demonstrate Shashin theory at work in ShashChess.

Modern chess engines behave like chaotic systems as slight perturbations to initial conditions can evolve into very different behavior. Many factors can lead to these perturbations: OS using time for some background process, thermals affecting clock speed and nps, etc.