[ List Earliest Comments Only For Pages | Games | Rated Pages | Rated Games | Subjects of Discussion ]
Comments by HGMuller
Ecumenical Chess. Set of Variants incorporating Camels and Camel compound pieces. (8x10, Cells: 80) [All Comments] [Add Comment or Rating]
H. G. Muller wrote on Sun, Jan 4, 2009 10:55 PM UTC:H. G. Muller wrote on Mon, Jan 5, 2009 08:20 AM UTC:Orthogonal contiguity is necessary, but not sufficient: The Gnu (Knight + Camel) has it, but has no mating potential.
I am not sure where Joe's conclusion that three orthogonally contiguous squares would be needed came from. Two is sufficient. Even the WD has mating potential on 8x8.
H. G. Muller wrote on Mon, Jan 5, 2009 08:49 AM UTC:I think you can only know this by dynamically searching for the move sequence. For pieces that cover only 2 orthogonally contiguous squares it is also necessary that the can get in a single move from a square where they cover (say) c1 to one where they cover BOTH a1 and b1, where the King on b3 should not be in the way. A Gnu cannot do this.
H. G. Muller wrote on Mon, Jan 5, 2009 10:16 AM UTC:Not really, as the King can already do that. In some endings you achieve the same thing (effectively making a null move) by stepping the King in a symmetric way over the diagonal (e.g. b3-c2) while all other pieces are on the diagonal.
H. G. Muller wrote on Mon, Jan 5, 2009 05:24 PM UTC:It is nonsense to count the square where the piece is standing, as a piece does not defend itself. The fact that a King could defend it is not relevant: to make a checkmate position in the corner the King is already spoken for to cover the squares on the second rank. So the other piece has to attack 2 adjacent first-rank squares. If it cannot do that, there are no checkmate positions.
H. G. Muller wrote on Tue, Jan 6, 2009 10:31 PM UTC:What is required to succesfully drive a King into the corner is very difficult to answer. For one, as the 12x12 checkmates thread shows, it is dependent on the size of the board. When you solve the end-games by retrograde analysis, there are several possibilities: Usually you start with very many checkmates, along the edge, none of them enforcible. (Unless you have something like a Queen). Pieces like Bison that cover only 2 orthogonally contiguous squares do not have those, however, and solely rely on a handful of corner checkmates. Usually all longer mates are ancestors of those corner mates. Sometimes the longer mates die out immediately, like in the case of Gnu. There exist mate-in-1 positions, but no mate-in-2 positions, so afterthat, you are done. This is just an unlucky coincidence of the piece not being able to make the critical step between a position needed to force the bare King to step into a corner, and the square where it needs to be for a corner checkmate. More often the number of longer mates increases very slowly with their duration, and then hovers for a long time around a very moderate number. These are the positions where the bare King is already trapped on the edge, and has to be driven into the corner with very precise play. On large boards, the number of longer mates even tends to decrease again, because it becomes easier and easier for the bare King to actually flee towards the corner voluntarily, the attacking pieces not being able to all follow it quickly enough to keep it trapped there, so that it can then escape along the other edge. So you have to confine the bare King more precicely as you are further away from the corner, not only cutting off its way back to the middle of the edges, but also preventing it gets too much of a headstart towards the corner. This leads to a decrease of the number of positions. On boards that are too big the number of mates actually decreases to zero before you retrogradely reached the middle of the edge, and the game is generally drawn: there is no way to drive the bare King towards the edge without it reaching the edge in the middle between corners, and there is no way to drive it over such a large distance to the corner without it escaping in one direction or the other. If you survive (in retrograde time) until the middle of the edges, though, the number of longer mates suddenly starts to explode. It is usually very easy to drive a bare King to the edge, if you don't care where it will hit the edge. Unless the board is really big. But on 8x8, if your own King is in the center, the opponend is already driven onto the second rank. So in the early phase, from very unfavorable position, almost any sequence of moves that step your King plust its lieutenant towards the center (using opposition to drive the opponent out of it) is a direct route towards the checkmate, and there are very many possibilities for this. On all boards I have tried, (upto 16x16) once you reach the point where the number of longer checkmates starts increasing again, it usually fills up the entire space of positions. Either that, or it slowly peters out before it ever got big.
Fabulous Flying Kittens. Toroidal board using three warp lines with 8 corners tied together. (8x8, Cells: 64) [All Comments] [Add Comment or Rating]H. G. Muller wrote on Wed, Jan 7, 2009 03:23 PM UTC:I don't think it is as simple as that. In fact it is very unclear what exactly is meant. But row 1 is said to be 'joined' with both row 8 and row 5. So there is a T-junction of some sort involved. It is not clear if this implies that row 5 and 8 are also joined.
H. G. Muller wrote on Wed, Jan 7, 2009 04:17 PM UTC:I have no idea what 'any squares a Knight usually passes through' means. Being a leaper, a Knight does not pass through any square, but teleports directly to its final destination. If you have a path in mind through which a Knight reaches its desitination, you should specify the path, as this is not standard. Does a Knight move in an L shape, and if so, does the long leg or short leg of the L come first? Or does it move like the Mao or Moa? Or is it a multipath Moo? Or does it move 3 forward and then back diagonally? Also the phrase 'can only jump if it moves like a Knight' is ambiguous. Does it mean the piece is a hopper?
H. G. Muller wrote on Wed, Jan 7, 2009 04:56 PM UTC:So if I understand it correctly, this Ganesha moves as K + N + lame Dababba. That makes me severly doubt the value you put on it (4). The K+N piece alone (on 8x8) is super-strong, about 8 (when Q=9.5). Perhaps it loses a little strength on 10x10, but this should be more than compensated by adding the lame-D moves.
H. G. Muller wrote on Wed, Jan 7, 2009 05:40 PM UTC:lame D = D that cannot jump. K+G vs K+R is a generally-won end-game on 8x8. I am pretty sure it is the same on 10x10, and that it does not matter much if Pawns are present. (Provided, of course, they are not very close to promotion). The reason is that Ganesha can checkmate (forcing it, so not just help mates like Archbishop) without the aid of its King on any size board (or in fact on a quarter-infinite board). It might take some time before it gets close enough to the opponent King to give the first check, but once it does, you are toast... And a Rook can not give perpetuals like a Queen can: when it tries you approach until you attack it diagonally, and then it has to move without checking, and the Ganesha gets one step closer to the opponent King. (Well, two steps, more likely! ;-) )
H. G. Muller wrote on Wed, Jan 7, 2009 06:22 PM UTC:How about this then: K+G vs K+R+N on 8x8 is 95% won with white (= the side with G) to move. With black to move still 35% of all positions is won. (There are many positions where black can capture your G or K on the first move, after which you cannot win anymore. K+R+N on the average cover 6.6+5.25+14 ~ 25 squares or 35% of the board, and either your K or G might be in their line of fire, explaining the bulk of the other 65% positions.) I cannot exclude there are some fortress draws. Do you think 5 is a good value for a piece that almost always slaughters a Rook + Knight in the end-game?
H. G. Muller wrote on Wed, Jan 7, 2009 06:44 PM UTC:Opening? I am talking about K+G vs K+R+N here. With onlyy 5 pieces on the board, I hardly would call that an opening...
H. G. Muller wrote on Wed, Jan 7, 2009 07:58 PM UTC:I think you still don't get it. I am not talking about a single position, but about 228 million different ones. Namely all possible positions with this material. Point is that if you hav King + Rook + Knight, and I have King + Ganesha, I will almost certainly beat you, unless it happened to be your move and my Ganesha happened to be hanging,or it is my move and you happened to have a fork or skewer on my King + Ganesha. In other words, if the Ganesha is trivially lost tactically, so that this is not really a K+G vs K+R+N ending at all. But in a tactically quiet position, the Ganesha almost always wins. That is not bad for a single piece, to defeat Rook + Knight with overwhelming superiority. A Queen cannot do that, for example.
H. G. Muller wrote on Wed, Jan 7, 2009 10:58 PM UTC:Well, I didn't say that it was wrong to have the K + N + lame D piece in there: even if it is worth slightly over a Queen, there is nothing against having 3 Queen-class pieces on a 10x10 board. (Capablanca has 3 Queen-class pieces on 10x8, and that is an OK variant...) Although having a piece capable of solo checkmating might not be good for a variant. (OTOH, Knightmate has that too, and does not really suffer from it.) K+N should be 7.5-8, almost balancing a Rook + Knight in an end-game with otherwise only King and Pawns. At least it is on 8x8. On 10x10 it might lose some value, but so would the Knight, so it might still be a fair match for a Rook and a Knight.
H. G. Muller wrote on Mon, Jan 12, 2009 11:27 PM UTC:K+D should indeed be pretty close to a Rook, perhaps slightly below it. You could weaken it further by making it a non-jumping piece.
H. G. Muller wrote on Tue, Jan 13, 2009 07:58 AM UTC:You might even underestimate them. A Crowned Bishop (B+W), which also has an enhancement of 4 target squares to an ordinary Bishop, tested as stronger than a Rook on 8x8 (say 5.25). I expect B+D (Hawk) to produce a similar enhancement. F+D (also color-bound), which is included in the Hawk, almost exactly balances a Knight on 8x8. In none of my tests so far color-boundedness seems to suppress the value of a piece very much. It probably just gives a larger pair bonus, which I could not measure well. Adding a single Wazir-like non-capture to a Bishop to break the color-boundedness upped its value by about 1/6, but adding it to a Knight upped its value by a very similar amount. So it is more likely that the enhancement is produced by the increased manoeuvrability than by lifting the color-boundedness. This seemed to be additive, so adding a full non-capture Wazir produced ~2/3, while adding an unrestricted Wazir (to make the Crowned Bishop) produced about 2. So captures seem to be worth about twice as much as non-captures. All this was on 8x8; the Bishop value seems to go up on wider boards. My guess is that the B-N difference, which is already 1/2 on a 10x8 board without the pair bonus, will be driven up to 1 on this large board. Say N=2.5 and B=3.5 excluding the pair bonus (when R=5). That might make the Hawk actually stronger than a Rook, ~5.5 + pair bonus. The Elephant has 12 capture targets, and short-range leapers with 12 moves usually are worth around 4.5, but might devaluate slightly on this big board. The Wazir non-captures will probably add 1.That would make 5.25. So my educated guess is that the Hawk is stronger than the Elephant, and that both are stronger than a Rook.
H. G. Muller wrote on Sat, Jan 17, 2009 10:10 AM UTC:I have nearly completed the first iteration of converting the 'chessd' Internet Chess Server software to handling variants. I will probably leave it on for testing overnight (i.e. starting from 18:00 EST). The server now supports the following variants: normal Chess Crazyhouse (zh) Bughouse (bh) Chess960 (fr) Xiangqi (xq) Capablanca Chess (ca) in various sub-variants (bird, carrera, embassy) CapaRandom Shatranj (sj) Courier Superchess (su) Great Shatranj Knightmate (km) (In parenthesis is the abbreviation you can use.) You can download the needed WinBoard client from http://home.hccnet.nl/h.g.muller/ICSclient.zip The IP number of the server is 80.100.28.169 (but this is pre-programmed in the shortcuts included in the ICSclient download) I will try to leave some bots on the server, so you have automated opponents (HoiXiangqi, Sunsetter (Crazyhouse), DanaSah (FRC), and FairyMax (most other variants)) When on the server, type 'who' to see who is logged in. To start a game, type: match opponentName 5 1 u variant> where 5 1 meand 5 minutes + 1 sec/move (you can pick any other time control that suits you), and variant one of the names given above. Have fun.
H. G. Muller wrote on Sat, Jan 17, 2009 09:57 PM UTC:OK, I have been testing a bit, and most things seem to work. The following details should be mentioned: * Great Shatranj should be played with legality testing in the WinBoard client switched off (through the 'Options->General...' menu; untick 'test Legality'). * If you play Courier Chess through the command 'match shamax 5 1 u courier' it will play in the initial setup that Fairy-Max uses (with a closed rank of Pawns). To play from the historic initial position with some Pawns on 4th/5th rank, you should play the subvariant '1', through the command 'match shamax 5 1 u courier 1'. * I also added category 'fairy' for some minor 8x8 variants including unorthodox pieces, the sub-variant indicating which initial setup will be used. So far only 'fairy archbishop' and 'fairy chancellor' work (where Q is replaced by A or C, respectively). * The category 'great' without mentioning a sub-variant will start with an initial position with a 'Sliding General' (= Mastodon = Q2 = FADW) on d1/d8. This in deviation to the 'official' rules of Great Shatranj, which has a 'General' (= Commoner = FW) there. For the logic of the variant I actually like that better. (As it replaces every slider move of Capablanca Chess by a jump of range 2.) To play with the 'General' on d1/d8 you should use a subvariant specification 'great shatranj' (i.e. start the game with the command 'match shamax 5 1 u great shatranj'.) * Note that in the great' category promotion can be to 'General' or any captured piece, even when the initial setup contained a Sliding General in stead of a General. To this end the captured pieces are displayed next to the board. When you promote, you should click the piece of your choice from the available pieces in the holdings, to finish the move. (The ShaMax bot always promotes to General, even if more valuable pieces were available; This is a limitation of Fairy-Max.)
Wazir. Moves one square orthogonally.[All Comments] [Add Comment or Rating]H. G. Muller wrote on Mon, Jan 26, 2009 03:59 PM UTC:True, the Ferz is more valuable than the Wazir. Not that either of them is worth a lot anyway: embedded in a FIDE-Chess context the Ferz is a about 1.5 Pawn. (A pair of Ferzes in the opening slightly lose to an opponent that started with a Knight in setad.) Wazir might be 1.25-1.4. The reason is that forward moves in practice are about twice as valuable as backward or sideway moves. And moderate color-boundedness is hardly a handicap. Note that in Shatranj or Courier Ferz and Wazir are worth a lot more in terms of Pawns, as the Shatranj Pawn is worth a lot less than a FIDE Pawn. This because it promotes tot he worthless Ferz, in stead of the decisive Queen. So in Shatranj, the Ferz is worth about 2 Pawns.
H. G. Muller wrote on Mon, Jan 26, 2009 09:10 PM UTC:It is too small for me to measure, so far. Fairy-Max is also not a good system for this, as it does not have pair bonuses in its evaluation, and erases the imbalance by trading one Ferz of the pair long before a weak piece like the Ferz can affect the game very much. I really would need a more sophisticated engine for this, that protects the pair it thinks is worth most, so the imbalance lives long enough to manifest itself clearly in the winning probability.
Xiangqi: Chinese Chess. Links and rules for Xiangqi (Chinese Chess). (9x10, Cells: 90) (Recognized!)[All Comments] [Add Comment or Rating]H. G. Muller wrote on Fri, Feb 6, 2009 08:32 PM UTC:I just produced a special Xiangqi version of my general variant engine Fairy-Max. Xiangqi is sufficiently different, because of its subdivided board, deviating promotion, stalemate an repetition rules, to warrant a separate engine, rather a further generalization of Fairy-Max.
The engine is called MaxQi, and is availabe as source code and Windows excutable from my website (download link http://home.hccnet.nl/h.g.muller/MaxQi.zip ). It uses WinBoard protocol to communicate its moves, and so can be run under WinBoard 4.3 ('WinBoard_F'). Other WB engines are HoiXinagqi and TJxiangqi. MaxQi is definitely a lot stronger than HoiXiangqi; I have not had it play many games against TJxiangqi yet, but I expect MaxQi to be weaker than that. Kangaroo (Newton). compound of Knight and Alfil.[All Comments] [Add Comment or Rating]H. G. Muller wrote on Mon, Mar 23, 2009 02:05 PM UTC:I have not actually playtested it, but the general formula for short-range leapers (value = (30+5/8*N)*N, where N is the number of target squares and the value in centi-Pawn) predicts a value of 450, i.e. slightly below a Rook. I would expect it to be a little stronger than the Carpenter, (which also has 12 target squares) though, because it has one more foreward move, and forward moves seem to be approximately 2 times more valuable than backward or sideway moves.
The FIDE Laws Of Chess. The official rules of Chess from the World Chess Federation.[All Comments] [Add Comment or Rating]H. G. Muller wrote on Tue, Mar 31, 2009 03:56 PM UTC:If the Bishops are on like-colored squares, it is an immediate draw even before the flag falls, as no checkmate positions exist. Ifthey are on unlike-colored squares it is more tricky. The other rules you quote seem to contradit each other. (Isn't the second one only for blitz Chess?) In theory a checkmate position exists, but only as a helpmate: you must put your King voluntarily in a corner, your Bishop next to it to block its only escape square, and then you can be mated. This definitely does not count as a 'forced win', unless you are already in that position and the opponent can mate you in one move and it is his move. I always thought that at standard time controls forfeiting on time in KBKB* counts as a loss, but perhaps they changed the rules.
H. G. Muller wrote on Tue, Mar 31, 2009 04:03 PM UTC:I am not sure what exactly you have in mind. You want to drag and drop gifs in an NxM table of checkered background? Without any interpretation of what happens? Or should it be communicated somewhere? I have a Javascript script to view a game it downloads from the web.
H. G. Muller wrote on Wed, Apr 1, 2009 01:22 PM UTC:So if I understand you corectly, you want to have a smart program (not necessarily in Javascript) which is aware of the rules of the variant, so it can parse PGN (properly disambiguating the SAN moves and taking account of their implied side effects). This then should pepare an intermediate format of the game, which can be viewed through a dumb viewer written in Javascript. Sounds very close to what I already have, and was using in my 'Battle of the Goths' Championships. ( http://home.hccnet.nl/h.g.muller/goths.html ) I use WinBoard as the smart PGN parser, which then prepares a data file that is loaded by the Javascript-powered HTML page, and unambiguously specifies how the latter should juggle around gif images to replay the game. Only thing that is missing is buttons to scroll through the game, just one to replay it completely from the beginning. But I guess that would be a comparatively simple addition.
25 comments displayed
Permalink to the exact comments currently displayed.