Engines playing Musketeer Chess, good price

[JohnWoe]

Thanks!

My proposal only affects to "analysing / position fen moves ..."
In normal games it doesn't matter.
I figured out that you have to go stage 1 -> stage 2 -> stage 3 -> stage 4 -> stage 5 -> stage 6.
So that naturally lead to this simplified fen-format.

Was thinking about :
How about a new "Random Musketeer Chess" variant? Where gated pieces are randomized so there's no "gating" stage ?
GUI would send this to engines: D H D@a0 D@b9 ...

[hgm]

I think that would very much subvert the intentions of the inventor. The idea was that the choice of gating squares would be an interesting and important tactical dilemma for the players. If that is true then assigning those at random might seriously disadvantage one player.

Personally I don't like the idea much, and consider the gating method in Seirawan Chess cleaner and as least as interesting. There one doesn't have the extra 'moves' for selecting the gating squares, but the branching ratio during the opening phase is much higher, due to the initial development moves existing in 3 flavors (gate Elephant, gate Hawk or gate nothing). So the opening variety is just as large. And there is the dilemma of postponing the gating to keep your options open, versus the risk you would lose the opportunity to gate altogether.

But one way an engine could treat Musketeer Chess in human-engine games is by simply proposing a random post-prelude position, where it picked piece types and gating squares itself. The user can always hit 'new game' if he doesn't like it. Or edit the position into one he does like, and play from there. You could also play a hybrid game, where you do select pieces, but where these are gated in Seirawan style.

I also consider it a very bad idea to allow a player to refuse his opponent's choice of piece type, and revert to a default choice in that case. I did not implement that at all in the KingSlayer-Aramis prelude.

In engine-engine tournaments the prelude choices are likely to always come from the opening book, the negine having no say in it.

[musketeerchess]

Thanks!

My proposal only affects to "analysing / position fen moves ..."
In normal games it doesn't matter.
I figured out that you have to go stage 1 -> stage 2 -> stage 3 -> stage 4 -> stage 5 -> stage 6.
So that naturally lead to this simplified fen-format.

Was thinking about :
How about a new "Random Musketeer Chess" variant? Where gated pieces are randomized so there's no "gating" stage ?
GUI would send this to engines: D H D@a0 D@b9 ...

Hi
The random gating of pieces could be interesting to test engines.
I tried a manual random gating to show the "importance" of this particular stage of the game, and i was right on my initial assumptions ! In fact, depending on the choices (and also it dépends on piece combination), the Gating Selection influences the final result significantly.

[musketeerchess]

Hi HG
When programming a CECP engine to play under winboard using 60 different pieces, is it possible to name them using 2 Letters instead of one?

This will ease programming.

[hgm]

Unfortunately 2-letter piece IDs are not compatible with the FEN standard, which WinBoard uses for communicating positions to the engine (and for recording initial positions in PGN).

I did provide an alternative, though": you can used 'dressed letters' as piece ID. These are a letter plus a punctuation sign. Currently single quote and exclamation point are allowed as suffixes, so that L, L' and L! are all different pieces. This would allow 3 x 26 different piece IDs. Which seems enough for most games. It is really only the large Shogi variants where I have ever run into trouble regarding the assignment of piece IDs.

[musketeerchess]

Unfortunately 2-letter piece IDs are not compatible with the FEN standard, which WinBoard uses for communicating positions to the engine (and for recording initial positions in PGN).

I did provide an alternative, though": you can used 'dressed letters' as piece ID. These are a letter plus a punctuation sign. Currently single quote and exclamation point are allowed as suffixes, so that L, L' and L! are all different pieces. This would allow 3 x 26 different piece IDs. Which seems enough for most games. It is really only the large Shogi variants where I have ever run into trouble regarding the assignment of piece IDs.

Hi HG
That's a good alternative.

How to make the FEN standard evolve to take into consideration the growing Chess Variant games and to allow easily introducing new pieces. I know that it's hard to think about chess variants involving more than 26 different pieces, that's not the point.

The point is that if the chess variant community wants unity and to become uniform to further pormote chess variants, we need a "universal lettering of pieces".

[hgm]

Actually I think universal lettering of pieces is a "wild goose chase". It doesn't even exist for orthodox Chess; each language uses its own piece names and IDs.

The solution is to have flexible software that can handle the differences.

[musketeerchess]

Actually I think universal lettering of pieces is a "wild goose chase". It doesn't even exist for orthodox Chess; each language uses its own piece names and IDs.

The solution is to have flexible software that can handle the differences.

I Agree

Is it possible to make optimization of the engine parameters using SPSA methodology under winboard ?

[hgm]

I am not really familiar with the SPSA methodology. Isn't that just a a particular strategy for deciding which test matches to do with what versions of the engine? WinBoard can be used to play the matches. But it couldn't change any parameters of the engine; it would not know what to change, or in what way.

[musketeerchess]

I am not really familiar with the SPSA methodology. Isn't that just a a particular strategy for deciding which test matches to do with what versions of the engine? WinBoard can be used to play the matches. But it couldn't change any parameters of the engine; it would not know what to change, or in what way.

It s a methodology used to optimise parameters like piece values etc

A default setting is challenged by different values of the tested parameters and the tested settings evolve according to the mini match result.

The newly tested engine has his parameters tweakened automatically and tested under the same conditions.

For example lets Have an engine with Queen valued as 800 cemtipawns. We will start a match with a clone that has queen value set to 500 centipawns

After 4 games the result will be for example a loss +3 to default engine. Then the queen value of the test engine will be changed and increased according to a precise formula and the new test clone is set with the new parameter and a mini match starts de novo against the default engine and so on etc.