Bored with orthodox Chess?

[George Tsavdaris]

In principle black can promote to any of the Spartan pieces, even to King if he only has one, but I suspect this is never useful.

This is one position where it is very useful to promote to King. For example let's assume that white just played 1.Qc8+
Now black can play 1...Hb1=K saving himself and making the game equal(?). Every other move loses.
But this is not my point here. My point is that is this move(promotion) legal?



As black's King is in check and it is the only one remained in the Spartan army, he obviously have to play as normal FIDE chess and either put a piece in front of the Queen and King to stop the check, or move the King to a non-checking square, or capture the Queen.
So does he have the right NOT to do anything of the above 3 that FIDE rules specifically say?
Because this is not explained at all in the SpartanChessonline site on the "All Rules" section.
It just says:
"When the Spartan has two Kings in play a Spartan King is immune from check. Thus, the Spartan may move a King onto an enemy attacked square, leave a King under attack or move a piece that would expose a King to attack. "

But this is by no means clear. Has 2 Kings mean to have them already prior the move or to have them after finishing his move?

I'm quite sure that the King promotion has to be allowed since it's more close to the rule's and game's philosophy and makes the game better.


I want to report also a bug in Winboard since in Edit mode you can't(i for sure can't) add a Hoplite for black(obviously).
There is not any Hoplite add option in the menu and selecting the black Pawn doesn't work as it puts a negative image of a Pawn(and it's logical).

[hgm]

The way I understand the rules, (and implemented them in Fairy-Max), promotion to King would be allowed here. On my viewerpage I write:

"The Spartans start wih two Kings, and lose when they leave all their Kings in check"

The inventor of Spartan Chess complemented me (in chessvariants.org) about this "succinct an accurate way" to describe he rules. So one assumes he agrees. We had an earlier discussion about how to denote check in SAN, and the conclusion was that he Spartans are only in check when all their Kings are under attack, and that he case where only one ou of wo is under attack should not be considered check, and will no get a '+ 'suffix. This because even when the King would be unprotected, it is not worse than an attack on an unprotected Rook in FIDE chess, and there is also no special notation or warning for that.

So in the given position, the Spartans resolve the check by promoting to King.

As to the piece menu:

The 22 piece types known to WB ome in two series of 11. The second series are in some variants considered the 'promoted' versions of the first series (e.g. in crazyhouse to depict promoted Pawns, or in Shogi for the promoted Bishop and Rook). Note there are 'promote' and 'demote' entries in the menu. When you click a piece, and then use these entries, you change it. Usually into something that was not in the menu.

Code: Select all

basic:    PNBRQFEACWM
promoted: OHIJGDVLSUK

So you get a Lance (L) by promoting an Archbishop (A), (swords -> spear, there seemed some logic to that), a Man (M) by demoting a King (K), a Falcon (V) by promoting an Elephant (E). If you 'promote' an already promoted piece, you change it into the next lower piece, while demoting a basic piece change it into the next higher. So to make a Ferz (F), you can demote a Queen (Q). (In some varants, like Shatranj and Xingqi, you get a Ferz automatically when you ask for a Queen.) To make a Unicorn (U) you can promote a King (K).

Sorry this is so clumsy. Of course there always is the alternative to paste a FEN. In Edit Position mode you can also type a FEN directly to the board (this pops up the move type-in box). The problem is that I want to abandon text-based piece menus (because the same pieces have many different names, depending on the variant), and move to a pictorial representation. (See http://hgm.nubati.net/news.html#tag-B3 .) But we decided to postpone that until we have the GTK front end.

[hgm]

Some preliminary results from the tests so far:

WinBoard does NOT allow promotion to King when the other King is in check! Clearly a bug. It does not really perform the promotion before it tests if you are in check after themove, under the assumption that if the Pawn won't block it, another piece won't block it either. ChessV promoted to King many times, several times while in check.

Fairy-Max has a bug: it does not adapt its King count when the opponent promotes to King. As a result, it then thinks that both Kings re the only King, and allows neither of them to stepinto check, leading to false illegal-move claims.

Fairy-Max also seems to have trouble understanding promotion to general. It sometimes claims false stalemates or mates for the opponent after it.

ChessV crashes if the game is too long. Mostly on move 132, one game on move 158, however.

ChessV seems to draw won position because it is reluctant to push its Pawns.

Catalyst is sorely in need of a hash table. it cannot seepromotions coming early enough without it.

[George Tsavdaris]

Some preliminary results from the tests so far:

WinBoard does NOT allow promotion to King when the other King is in check! Clearly a bug.

Can you please fix it and release another setup with the fixed Winboard.EXE? Or just the fixed Winboard.EXE would be even better.
Thanks.


Winboard has another problem, that if you edit a position with only one King for the Spartans and this King to be in check and it's white to move, then it proceeds like nothing is wrong. Of course this is not a real problem since if you are stupid enough to setup a position with white to move and Spartan to be in check, then it's your fault actually, but it's better to have the legality check to cover this too.

[hgm]

OK, try http://hgm.nubati.net/WinBoard-4.5.beta.zip (a bare executable).

It does not fix the legality testing on entered positions yet (this is a general problem, not specific to Spartan Chess). I guess the desired response is to stay in Edit Position mode and give an Error popup? Can be tricky to do that when the error is caused by pasting a FEN.

[Greg Strong]

Some preliminary results from the tests so far:

ChessV crashes if the game is too long. Mostly on move 132, one game on move 158, however.

ChessV seems to draw won position because it is reluctant to push its Pawns.

Thanks for the feedback! I'll work on these. I'll also try to get ChessV playing Spartan in pure bitboard mode this weekend. I don't think it will take too long and will speed it up a lot...

As always, thanks for your continued effort promoting chess variants and related computer support!

Also your piece evaluations look excellent, but I'm surprised you value the king so highly. Of course having two kings is valuable because your opponent can't so easily force your hand by putting you in check; who knows how much that privilege is worth and your estimate could well be accurate.

[hgm]

The King value was based on the observation that deleting a Spartan King and a Persion Bishop + Pawn disturbed the balance in favor of the Persans, while deleting a Rook in stead hurt the Persians more.

It does not sound unreasonable. The value is about 1.5-2 Pawns above the tactical value, and this is a quite normal value for King Safety. And having a spare King does provide a lot of King Safety, even if the rest of your fortress is lousy.

Another way to look at it is that KQKR is always won, while KQKK is generally a draw (unless the Queen has a quick tactical win by skewering or forking two widely separated Kings).

[PK]

the first working version of Oberon for Spartan chess is available at:

www.koziol.home.pl/oberon.zip

it's a basic pvs search with null move, futility pruning and small tt + material/pcsq/mobility eval.

next week I'll try to make it faster (there is an ample room for improvement)

one question: should I add random component to eval, in order to avoid repeated games against catalyst / chessV?

[hgm]

Some form of randomization would be highly desirable, I think. In Joker I have always used a random term in the eval for this, which worked great in bullet games, but seems to become less effective at long TC. For Fairy-Max I randomize the first 4 moves by adding a bonus in the root. (Which sometimes makes it play very poor moves, like a2a4. But that is mainly because Fairy-Max evaluation is practically material only, so that a2a4 and e2e4 would have nearly equal scores to begin with...) I guess I would pick that system for my future variant engine, but then during the entire game. I saw a very smple implementation of it in XQWLight: just adapt alpha by a small random quantity before searching each move in the root.

[Richard Allbert]

Arrgh. I'm away for a week skiing.... Catalyst doesn't promote to king! I won't be able to fix this until a week tomorrow. Sorry guys!