pre-announcement: 10x8 tourney

[Evert]

Ok, I think I fixed the problem. I certainly fixed one nasty bug. Because I had the update for the PV and the time-out code in the wrong order, it could happen that the returned PV contained moves from the previous iteration, which isn't a problem unless we've just changed our mind about what the best move is. This doesn't affect the best move that is returned (because that's tracked separately) but it does affect the ponder move, which is taken from the PV when the search is completed.
That meant the ponder move could be illegal, and since it's not checked for that (how could the ponder move ever be illegal, right?) this could wreck the position, or at the very least the ponder search.
This also appears to have fixed an endless-loop in SEE, although I have no idea how it would have caused that (or how that particular loop could have been infinite at all). I say "appears" because it hasn't happened since fixing the first bug while it happened more regularly before. I'll run a self-play tournament overnight to make sure it's ok.

I normally test with ponder off, so I hadn't noticed this problem before. The bug may be present in Jazz as well.

[Ferdy]

Indeed, in all cases it should be read as 'can'. At least that is how all engines implement it: no Spartan castling, no backward Hoplite moves, no capture on Hoplite double-pushes, and c indicates capture-only.

Started coding now and would like to ask about winning for the spartan side. Can a spartan king able to checkmate the persian king? Sample situation, Spartans still have 2 kings, one king is safe and is not checked while the other spartan king is closing-in/attacking the 1 persian king. If persian king is cornered located at A1 and here is the spartan king moving to A2 with support from other persian pieces. Is this also a win for the Spartans? Thanks a lot.

I only got this rule from the web ref.

4.1 Winning
The Spartan and Persian have different victory conditions.

4.11 Spartan Victory - The Spartan wins when the Persian King is checkmated as in orthodox chess.

[hgm]

Indeed, this is a win. One way to look at it is that the Spartans are only considered to be in check when all their remaining Kings are attacked. So this is a regular checkmate:

[d]8/8/8/8/8/2k5/1k6/1K6 w - - 0 1

White is in check, and cannot escape; Kxb2 is not legal for him because Kb2 is protected by Kc3. The Spartans, however, are not in check, because Kc3 is not attacked.

The following (nonsense) game is perfectly legal:

Code: Select all

[Event "Edited game"]
[Site "CHESS_LAPTOP"]
[Date "2011.10.30"]
[Round "-"]
[White "-"]
[Black "-"]
[Result "1-0"]
[Variant "spartan"]

1. e4 Cd6 2. Qf3 Cb6 3. Qxf7 Wh6 4. d3 Kg8 5. Qxe8#
{White mates} 1-0

lgk1Q1kl/hhhhh1hh/1c5w/8/4P3/3P4/PPP2PPP/RNB1KBNR b KQ - 0 5

with the standard substitutions needed for this forum (L->B, C->R, G->Q and W->N):
[d]bqk1Q1kb/ppppp1pp/1r5n/8/4P3/3P4/PPP2PPP/RNB1KBNR b KQ - 0 5

Black is in check, because both his Kings are attacked by Qe8. There is no move that can alter this. Thus he is checkmated (so called duple-mate). Note how he could simply ignore the attack on Kf8 by 3. Qxf7, because Kc8 was still safe. (Normally you would play 3... Kxf7 anyway, of course, because the Queen is such a tasty bite. This is sort of a Spartan equivalent of fool's mate.) Also note how WinBoard did not put a check sign with Qxf7.

[Ferdy]

Wow, looks like my king attack detection will be getting complicated

[hgm]

Well, it is questionable how important King Safety is when you still have two Kings. Leaving one King exposed to capture by R or Q, protecting it rather than moving it away is a quite common strategy for the Spartans, as the extra King seems to be worth slightly less than a Rook. A good way to calculate King safety could be to calculate the safety of both Kings like it was the only King, and then take the best of the two, possibly giving a bonus if the other one is also quite safe (like you would evaluate castling rights when you can still castle both ways). And perhaps giving a small penalty when the unsafest King is actually attacked.

The fact that you have a second King already provides a lot of King safety in itself. But this is of course incorporated in the piece value of the King, which is nearly 2 Pawns higher than it would be based on its tactical merits. KQKK with connected Spartan Kings is a draw. (Far-apart Kings are toast, of course, to a Queen.)

[smrf]

For some reasons and e.g. maybe to get some testing results of other 10x8 activists I have decided to publish again an actual donationware version of the SMIRF GUI and its engine BC-176f-2 which also may by used by Winboard-F in combination to HGM's smirfoglot adapter.

Download-link: https://files.me.com/r.scharnagl/ktgt2z

Regards, R. Scharnagl

[hgm]

Thanks Reinhard! I am too busy with my Xiangqi engine right now, but as soon as the ICGA Olympiad is behind us, I will start testing the 10x8 engines again.

Any plans for a Smirf that also supports Spartan Chess?

[Evert]

Hopefully this weekend. I'm debating whether to annoucne it on the "technical discussion" or the "general topics" forum...

At the last moment spending half an hour or so on doing the release (compile everything, create packages, update webpage, upload the lot, check that all the links work) lost out to going to sleep.
I'll try squeezing it in tonight.

I did have something odd happening with the Windows version of my perft program, which crashed immediately both in 32 and 64 bit mode. No idea why and it's probably quite hard for me to debug (I don't have Windows). It could be an issue with Wine, but the engine itself runs properly. The Linux and OS X versions are likewise fine. Oh well.

[smrf]

... Any plans for a Smirf that also supports Spartan Chess? :P

Sorry, probably not. There still is a lot to do for me concerning pieces having more traditional gaits .

[Evert]

Version that plays Capablanca is now available. See announcement here and get it here.