Musketeer Chess Variant pieces available at HouseofStaunton

[musketeerchess]

Hi
Your technique to let fairy-max use a defined combo to select the different armies is imho the good way to deal with the different variants possible in Musketeer Chess.

In fact, each combination of pair of pieces is a different variant (like in FRC) which implies different possibilities and different strategies.

When an engine is to be developed, the simplest way is to include the default combination of pieces (Leopard and Cannon), then for further development, we shall have the choice of various combinations.

For engine vs engine play the combination should be either fixed before the match or be random choice

[Evert]

I think the most sensible approach is to treat them as different variants. The engine would have to know anyway that you are not allowed to promote to piece type Z if X and Y where the only pieces you had in hand at the start of the game, which makes treatment as a single variant containing all possible pieces difficult. A similar problem occurs in Chess with Different Armies. In Fairy-Max I use an engine-defined combo option to select the actual armies.

Ah, good point about the promotions!
I think I have them limited to QRBN at the moment, for essentially that reason.
I have a vague notion to make Sjaak II aware of shuffle variants (and able to generate starting positions); as a part of this I want to allow it to permute certain pieces between the board and holdings (for Seirawan2880) and I was thinking I'd extend that to pick N pieces from a specified set of M pieces. Those pieces in he set that are not picked are then removed from the promotion options and pieceToChar.
It's a lot of hassle though and it's not even on my TODO list (yet)...

However, a part of me hates the idea of having to generate all possible permutations instead of letting the computer handle it...

I guess having options that only have a meaning for one particular variant is also rather unsatisfactory.

Oh, for sure. That's not happening.
If I wanted to do that, I would allow the user to override the startup FEN for any variant. That's a whole other can of worms though (the engine options dialog isn't ideal for that, solutions are either ugly or cumbersome or both).

[Evert]

In fact, each combination of pair of pieces is a different variant (like in FRC) which implies different possibilities and different strategies.

Well, yes and no. Your variant is basically a super-set of Seirawan Chess and I don't think it makes much sense to treat it as anything other than that. Yes, the combination of pieces changes how you play the game, but for the most part this doesn't change much.

When an engine is to be developed, the simplest way is to include the default combination of pieces (Leopard and Cannon), then for further development, we shall have the choice of various combinations.

Well, re-tuning the full evaluation for every possible combination is an untenable goal, so you need some fairly generic terms in there.

However, adding the extra pieces isn't all that difficult, you just need to specify their moves.

[musketeerchess]

Hi
When i take a look at your homepage i find it easy to tune chess variants with sjakk. Very simple instructions.

When i talked about beginning with the default combination concerning programming Musketeer Chess Variant this was to make it the easiest way to program an engine playing musketeer chess. Then implementing new pieces which adds new possible combinations could be done progressively.

The difficult part in Musketeer Chess is the evaluation of the new pieces. I have an idea naturally concerning the relative value of the pieces.

[Evert]

When i take a look at your homepage i find it easy to tune chess variants with sjakk. Very simple instructions.

Well, I certainly hope that implementing new variants is something that isn't too hard - it's part of the design goal.
(By the way, Sjakk is a different engine than Sjaak II and by a different author)

When i talked about beginning with the default combination concerning programming Musketeer Chess Variant this was to make it the easiest way to program an engine playing musketeer chess. Then implementing new pieces which adds new possible combinations could be done progressively.

It can be quite tricky to add more pieces later on; it's generally easier to design the engine up-front with a certain number of pieces in mind (Sjaak II can handle up to 32 piece types).

The difficult part in Musketeer Chess is the evaluation of the new pieces. I have an idea naturally concerning the relative value of the pieces.

The values I've put in are rough guesses on my part, but Sjaak II has an algorithm for making a crude educated guess at the value of a piece. Here's what it thinks:

Code: Select all

#[Musketeerish] 0w>pieceinfo
 0 Knight N (N, n)
   Leaper  10000000 (move)
   Leaper  10000000 (capture)
   Move statistics: 2--8 <5>
   Cannot deliver mate alone
   Can deliver mate with help of B 
   Minor-class piece
   King attack weight 8
   Game-phase weight  8 / 160
   Value             [333 333]
 1 Bishop B (B, b)
   Slider  0000000c (move)
   Slider  0000000c (capture)
   Move statistics: 7--13 <9>
   Pair bonus
   Cannot deliver mate alone
   Can deliver mate with help of N B 
   Minor-class piece
   King attack weight 8
   Game-phase weight  8 / 160
   Value             [333 333]
 2 Rook R (R, r)
   Slider  00000003 (move)
   Slider  00000003 (capture)
   Move statistics: 14--14 <14>
   Major-class piece
   King attack weight 12
   Game-phase weight  12 / 160
   Value             [500 500]
 3 Queen Q (Q, q)
   Slider  0000000f (move)
   Slider  0000000f (capture)
   Move statistics: 21--27 <23>
   Super-class piece
   King attack weight 24
   Game-phase weight  24 / 160
   Value             [1000 1000]
 4 Dragon D (D, d)
   Slider  0000000f (move)
   Leaper  10000000 (move)
   Slider  0000000f (capture)
   Leaper  10000000 (capture)
   Move statistics: 23--35 <28>
   Super-class piece
   King attack weight 32
   Game-phase weight  32 / 160
   Value             [1333 1333]
 5 Chancellor C (C, c)
   Slider  00000003 (move)
   Leaper  10000000 (move)
   Slider  00000003 (capture)
   Leaper  10000000 (capture)
   Move statistics: 16--22 <19>
   Super-class piece
   King attack weight 20
   Game-phase weight  20 / 160
   Value             [833 833]
 6 Archbishop A (A, a)
   Slider  0000000c (move)
   Leaper  10000000 (move)
   Slider  0000000c (capture)
   Leaper  10000000 (capture)
   Move statistics: 9--21 <14>
   Super-class piece
   King attack weight 20
   Game-phase weight  20 / 160
   Value             [833 833]
 7 Elephant E (E, e)
   Leaper  10010000 (move)
   Leaper  10010000 (capture)
   Move statistics: 6--16 <12>
   Super-class piece
   King attack weight 24
   Game-phase weight  24 / 160
   Value             [1000 1000]
 8 Hawk H (H, h)
   Leaper  10020000 (move)
   Leaper  10020000 (capture)
   Move statistics: 6--16 <9>
   Major-class piece
   King attack weight 8
   Game-phase weight  8 / 160
   Value             [333 333]
 9 Leopard L (L, l)
   Leaper  10000000 (move)
   Stepper 00000100 (move)
   Leaper  10000000 (capture)
   Stepper 00000100 (capture)
   Move statistics: 4--16 <11>
   Super-class piece
   King attack weight 20
   Game-phase weight  20 / 160
   Value             [833 833]
 10 Cannon O (O, o)
   Leaper  90000000 (move)
   Leaper  90000000 (capture)
   Move statistics: 6--16 <12>
   Super-class piece
   King attack weight 24
   Game-phase weight  24 / 160
   Value             [1000 1000]
 11 Unicorn U (U, u)
   Leaper  10030000 (move)
   Leaper  10030000 (capture)
   Move statistics: 4--16 <10>
   Major-class piece
   King attack weight 8
   Game-phase weight  8 / 160
   Value             [333 333]
 12 Fortress F (F, f)
   Leaper  90010000 (move)
   Stepper 00000200 (move)
   Leaper  90010000 (capture)
   Stepper 00000200 (capture)
   Move statistics: 6--20 <13>
   Super-class piece
   King attack weight 20
   Game-phase weight  20 / 160
   Value             [833 833]
 13 Spider S (S, s)
   Leaper  10040000 (move)
   Stepper 00000100 (move)
   Leaper  10040000 (capture)
   Stepper 00000100 (capture)
   Move statistics: 6--20 <14>
   Super-class piece
   King attack weight 24
   Game-phase weight  24 / 160
   Value             [1000 1000]
 14 King K (K, k)
   Leaper  10050000 (move)
   Leaper  10050000 (capture)
   Move statistics: 3--8 <7>
   Royal
   Cannot deliver mate alone
   Royal piece
   Minor-class piece
   King attack weight 8
   Game-phase weight  0 / 160
   Value             [0 0]
 15 Pawn   (P, p)
   Stepper 00000300 (move)
   Stepper 00000400 (capture)
   Move statistics: 0--1 <1>
   No return
   Cannot deliver mate alone
   Pawn-class piece
   White pawn
   Black pawn
   King attack weight 1
   Game-phase weight  0 / 160
   Value             [128 160]

Some of them look rather inflated to me (particularly the pawn), but the ordering should be ok-ish.

[musketeerchess]

Hi
The overall evaluation is globally correct, but it clearly overestimates the pawn, knight, Archbishop, Leopard, Cannon value and under estimates: Hawk, Unicorn.

I don't know what to think about the Spider and Fortress evaluation, i think that they are much overestimated. They are super pieces when there are many pieces and they loose strength when pieces are exchanged. In the endings they are probably less effective than a Rook. This makes their value range between 450 and 750.

[Evert]

Hi
The overall evaluation is globally correct, but it clearly overestimates the pawn, knight, Archbishop, Leopard, Cannon value and under estimates: Hawk, Unicorn.

I don't know what to think about the Spider and Fortress evaluation, i think that they are much overestimated. They are super pieces when there are many pieces and they loose strength when pieces are exchanged. In the endings they are probably less effective than a Rook. This makes their value range between 450 and 750.

There is no "correct" way to calculate the value of a piece based only on how it moves, so typically you need to enter the piece-value along with the rules. If you don't Sjaak II does a crude guestimate based on the attack pattern on a 3x3 and a 5x5 board. The idea behind this is that this tells you roughly how dangerous a piece is in a king-side attack. The score is then modified based on whether a piece is colour-bound or not and whether it can promote or not. The algorithm was calibrated for normal chess and Shogi. It normally produces N=333 on an 8x8 board, Q~3M and Q~2R, which means all exchanges are roughly ok.

For Musketeer, I think the Spider looks about right. From the algorithm the Fortress gets similar statistics as a Queen, so the score comes out the same. It should be something like an Archbishop instead. Archbishop and Chancellor come out the same, which isn't right but isn't completely aweful either. That they come out the same is certainly the best that can be expected from the algorithm. Similarly, the Leopard to the algorithm "looks like" an Archbishop and so gets the same score. This is too much, but perhaps not as bad as you'd expect because distant moves of a Bishop aren't that valuble. Hawk and Unicorn are undervalued because their long jumps are off the radar for the algorithm; long range leapers aren't very strong on an 8x8 board though.

Pawns are supposed to come out as 100, not sure what happened there. I think I may have neglected to unset the value of the pawn, which means it was tweaking the value I passed in rather than estimate the value from scratch.

All in all, the algorithm essentially does what it's supposed to do and does about as well as can be expected. Piece values are only a part of the evaluation, of course, and as long as they don't cause bad trades they're not too important. It's of course better to simply specify the piece values.

[Nordlandia]

I'm not against this variation at all, it's a nice invention to the supplement classic chess. However i belive unless you live around some big club that play chess on regular basic. This variation will not be played unless you introduce the pieces at the club. My experience is that older players (adults) is not that keen to play custom chess variations over the board. For youngsters that's another story.

[musketeerchess]

Hi Jon
I agree with you. Due to very busy schedule i'm not able to go to chess clubs to promote it. But if someone buys 5 or more kits i'm willing to pay for him 1 more kit. This is intended naturally to promote the variant.

If you do so, you have to mail me your order to the House of Staunton. I then will see directly with them to offer you one more kit.

[musketeerchess]

Marry Christmas