-<h3>enpassent</h3> If the opponent has performed a double pawn push
+<h3>en passant</h3> If the opponent has performed a double pawn push
and there is now a pawn next to it belonging to the player to move then
"enpassant" is the entry from RandomEnPassant whose offset is the file
of the pushed pawn (counted from 0(=a) to 7(=h)). If this does not
present on e1/e8.
<h2>weight</h2>
In the Polyglot source code this field is called "count" but it is in fact
-a measure for the quality of the move. <b>It should be at least one.</b>
+a measure for the quality of the move.
<p>
The Polyglot book generator sets it to 2*(wins)+(draws), globally
scaled to fit into 16 bits. A move with a weight of zero is deleted
from the book. This is just a convention and book authors
-are free to set this field according to their taste (as long as it is at least one).
+are free to set this field according to their taste.
<p>
If random play is enabled in Polyglot then the probability that
a move is selected is its weight divided by the sum of the weights
of all the moves in the given position.
+In XBoard, which also supports Polyglot books, the weight can be raised
+to a power N (0 <= N <= infinity) before doing the probability calculation,
+enhancing or completely destroying the effect of the weights.
<h3>A note on zero weights</h3>
-Polyglot assumes that weights are at least one since a zero
-weight means the move should not be played, and hence you might as
-well delete the entry from the book (which the Polyglot book generator does).
+Note that older Polyglot versions are defective as to the use of zero weights,
+(meaning the move should never be played),
+and crash when they encounter them.
+This because of a misplaced internal consistency check,
+which really only made sense during book building
+(wich deletes all entries with zero weight from the book).
-Nonetheless, if you really insist, an entry with zero weight can be
-approximated by giving it very low probability. This is what Scid does.
+Yet hand-tuning of a book often leads to the desire to completely disable a move,
+without deleting it. Some software approximates an entry with zero weight
+by giving it very low probability. This is for instance what Scid does,
+but it means the move will still occasionally be played.
+With XBoard you can edit Polyglot books to set the weight to exactly zero,
+as well as delete or add moves.
<h2>learn</h2> "learn" is set to zero by the Polyglot book generator.
If you set "BookLearn=true" in polyglot.ini then this field
-is used to record learning information. However I do not know of any
-program that uses this information. Also this means the Polyglot should
+is used to record learning information. From Polyglot 2.0 on, Polyglot
+can also use this learn information for modulating the playing probabilities
+of the book moves, based on how often they were played and with
+how much success they met. To use this feature Polyglot should
have write access to the opening book which on Unix systems will probably
not be the case if the book has been installed centrally.
+Note that for the book to be opened for write access, the option
+BookLearn will have to be set before the book file is specified!
<p>
<h2>Sample code</h2>
The following sample code is released in the public domain.
(see above) in which the "from square" is the position of the king and
the "to square" is the position of the rook. This is to maintain
compatibility with various Fruit derivatives.
-</body>
+<h2>Extension to other variants</h2>
+Polyglot does not support any Chess variants other than Mad Queen and Chess960.
+XBoard, however, does, and has extended the definition of the Polyglot book format to cover such variants.
+To generalize to various board formats, it also numbers the board square in raster-scan fashion.
+So in Xiangqi, which has a 9x10 board, the square number equals 9*row+file, and can run upto 90.
+The piece numbering is also extended beyond King, keeping the alternation of
+white and black pieces of the same type, to accomodate unorthodox piece types.
+How unorthodox pieces should be mapped on numbers is not yet subject to an official standard.
+But variants that only the orthodox pieces (e.g. Suicide, Atomic) do not suffer from that.
+<p>
+To map a (piece, square) combination to a hash key, however,
+only the square number modulo 64 and the piece number modulo 12 is used as index in the key table.
+The square number divided by 64 and piece number divided by 12 is then used to decide
+how the retreived key should be (bit-wise) rotated to get the actual key.
+Square numbers in the range 64-127 rotate the key left by 8 bits,
+while piece values in the ranges 12-23, 24-35 and 36-47 rotate left by an additional 16, 32 or 48 bits, respectively.
+This is sufficient to handle boards of upto 128 squares, and up to 24 piece types.
+<p>
+A number derived from the variant can be XORed to the overall key,
+to prevent positions from variants with equal board and piece types (e.g. suicide, losers or atomic)
+to map to the same key, and invalidates books made for one of those for accidental use by any of the others.
+The table below defines the number that should be used for some common variants.
+Note that some variants that are mutually compatible use the same number,
+so that they can use the same opening book.
+<pre>
+ 0 standard Chess (aka Mad Queen), WildCastle, NoCastle, Fischer Random Chess (aka Chess960),
+ 0 Xiangqi (Chinese Chess),
+ 6 Crazyhouse,
+ 7 Losers Chess,
+ 8 Suicide Chess, Giveaway Chess,
+10 TwoKings,
+12 Atomic Chess,
+13 3Check,
+14 Shatranj,
+23 Shogi (Japanese Chess),
+25 Courier Chess,
+26 Gothic Chess, Capablanca Chess, Capablanca Random Chess,
+28 Knightmate,
+30 Cylinder Chess,
+31 Falcon Chess,
+33 Berolina Chess,
+34 Janus Chess,
+35 Superchess,
+36 Great Shatranj,
+38 Makruk (Thai Chess),
+39 S-Chess (aka Seirawan Chess),
+40 Grand Chess,
+41 Spartan Chess,
+</pre>
+<p>
+For variants with holdings from which pieces can be brought (back) in play (e.g. Crazyhouse, Shogi, S-Chess),
+the holdings are also part of the game state, and must be encoded in the key.
+To do this, the holdings are considered as a single off-board square with the next available number
+(so square number 64 in Crazyhouse).
+The (piece type, square number) pair is than mapped onto a key in the same way as for on-board pieces.
+A complication here is that multiple pieces, possibly of the same type, can be on this 'holdings square',
+so that XORing the corresponding key into the total will not work.
+(Having two pieces of a certain type would then be indistinguishable from having none, as A XOR A equals 0 for any A!)
+Therefore the holdings keys are not XORed but added to the total key,
+after all other key terms have been taken into account through XOR.
+<p>
+The move encoding on non-8x8 boards is generalized to fromSquare*boardSize + toSquare + boardSize*boardSize*promotionType
+(where boardSize = boardLength*boardWidth).
+For large boards this might leave a shortage of promotion types,
+but fortunately promotions are not very common in opening theory.
+The meaning of the promotionType code can be variant-dependent.
+Especially if there is a shortage of codes, the most common promotions should receive the lowest codes.
+(E.g. in Grand Chess the board is 10x10, so the promotion-type multiplier is 10,000,
+while the 16-bit move encoding only runs upto 65,535.
+So there is only room for 6 promotion types, while 7 would be needed for the 6 possible piece types and non-promotion.)
+In asymmetric variants (e.g. Spartan Chess) the same code needs not indicate the same piece type for both sides.
+<p>
+Special moves are indicated by using a (non-existent) promotion choice as an 'escape',
+in which case the from-square number isn't really a from-square, but identifies the special move.
+This system is used (with promotionType = 9, when available) for drops, where the from-square contains the ID of the piece to be dropped
+(A=1, B=2 etc.).
+It is also used to indicate double-moves of a Chu-Shogi Lion (promotionType = 10, when available, but 2 in Chu Shogi),
+where every combination of first and second step gets a unique code 0-63, which is then stored in the from-square.
+</p>
+<p>
+For really large boards (say larger than 144 squares) a more compact encoding system should be used,
+which indicates the from-square relative to the to-square in polar coordinates:
+toSquare + boardSize*directionCode + 4*boardSize*distanceCode.
+Here boardLength is the largest linear dimension of the board (be it height or width),
+and the distanceCodes < boardLength-1 indicate the distance along the indicated ray to the from-square,
+such that the move goes in the direction forward, right, right-forward or left-forward as seen from the player that makes the move.
+When this lies off-board, the distanceCode should be subtracted from boardLength+1 to obtain the distance in the opposite direction.
+The rayCodes 0-3 represent the directions vertical, horizontal, diagonal and anti-diagonal (N, E, NE, SE), respectively.
+The remaining distanceCodes (>= boardLength-1) represent oblique moves, special moves and promotions.
+They always start with the two Knight jumps landing just right of the ray, in the positive and negative direction,
+followed by two (1,3) and two (2,3) jumps.
+</p><p>
+After the Knight jumps the meaning of the distanceCodes becomes variant dependent.
+By default they would contain promotions to various piece types,
+in the order of their piece code (i.e. N, B, R, Q, K, other) in so far they participate in the variant.
+These codes imply a single step in the forward direction along the ray.
+In variants with Shogi-like promotions, where all pieces can promote, but the only choice is promote or not,
+the entire range of distanceCode meanings will be repeated, this time meaning the move will be accompanied by a promotion.
+</p><p>
+The meaning of these variant-dependent distanceCodes can be overruled through the book header.
+After a line with the word "Special" in this header,
+a number of lines can follow, specifying the move as it would be for the vertical ray in the forward direction, starting from e4.
+After that follow codes to indicate additional oblique moves: a pair of Camel jumps, a pair of Zebra jumps, etc.
+With Shogi-like promotions, the 16 promotion choices instead code for two-step Lion moves that start off along the ray.
+Promotions are indicated by adding a number to the distanceCode one larger than the highest code used to indicate any oblique move.
+E.g. for Tenjiku Shogi with its 16x16 board codes 0-14 indicate true (positiove and negative) distances along the ray,
+15 & 16 are Knight jumps, 17-32 are Lion double moves, 33-36 are C and Z jumps (for the area moves),
+and promotions add 37.
+As the board has 256 squares, there are only 64 distance codes available,
+so the highest code for which a promoted version is available is 26 (26 + 37 = 63).
+So there is no room to specify promotion on C and Z jumps, and with part of the Lion double-moves.
+But the only pieces in Tenjiku Shogi that can make C and Z moves do not promote,
+and the Lion double-moves can be ordered such that only the backward moves cannot promote.
+</p><p>
+For Maka Dai Dai Shogi the 8 codes after the Lion double moves
+(36-43, as the board is 19x19 there, shifting up everything 3 places compared to Tenjiku)
+specify Lion Dog linear triple-moves,
+(2-out-1-in, 2+1-out, 1+2-out and 1+1+1-out in both directions along the ray)
+and 44 specifies the moves was made by a 'Universal Leaper' (the ray number indicating which one).
+The latter necessitates the decoding routine to find the corresponding piece on the board.
+('Universal Leaper' in this context means a piece that is not bound to rays,
+and can move almost anywhere, such as Hook Mover, encoded in ray 0, Capricorn (ray 1) or Emperor (ray 2).)
+With 45 distance codes, 4 rays and 361 board squares all available 16-bit codes ar almost used up,
+so there is no way to indicate promotion.
+This is serious, as this variant has promotion on capture,
+making promotion early in the opening quite common.
+Fortunately the pieces are clearly divided into those that promote to upward-compatible pieces
+(replacing all step moves by unlimited sliding in that direction),
+for which you would always promote,
+and those that 'promote' to a worthless Gold, which you would virtually never promote.
+So promotion / deferral is decided by which class the piece belongs to,
+and only moves that do the opposite from what you would normally do cannot be incorporated in the book.
+This again need consulting of the board position before the move can be decoded.
+</p>
+<iframe src="cgi-bin/count.cgi?H=9" width="130" height="40"></iframe>
+</body>
projects / polyglot.git / blobdiff