Home 
Search 
Today's Posts 
#1




3move repetition and FEN notation
FEN notation is supposed to represent a game state fully. It does
represent castling permissions, enpassant, 50move draw. But it doesn't represent 3move repetition draw. How come ? Given some position in FEN notation, there's no way to know whether we're in the middle of a 3move repetition, so FEN doesn't fully represent a position. Or am I misunderstanding something ? Thks Eli 
#2




3move repetition and FEN notation
Hi, Actually it is not a 3move repetition but a repetition of 3 times the same board situation. This means that you do need a transcription of the entire game (or at least up till the last capture or pawn move), which FEN indeed does not include. Cheers, Arnold "Spur" wrote in message om... FEN notation is supposed to represent a game state fully. It does represent castling permissions, enpassant, 50move draw. But it doesn't represent 3move repetition draw. How come ? Given some position in FEN notation, there's no way to know whether we're in the middle of a 3move repetition, so FEN doesn't fully represent a position. Or am I misunderstanding something ? Thks Eli 
#3




3move repetition and FEN notation
Spur wrote:
FEN notation is supposed to represent a game state fully. It does represent castling permissions, enpassant, 50move draw. But it doesn't represent 3move repetition draw. How come ? FEN only represents a position up to telling you what moves are legal. It also tells you when the 50move counter expires but, as Arnold Meijster says in his followup to your post, you need to know every move since the last irrevocable move (pawn moves or captures) to know whether there's a draw by threefold repetition in the offing. Dave.  David Richerby Strange Metal Lotion (TM): it's like www.chiark.greenend.org.uk/~davidr/ a soothing hand lotion that's made of steel but it's totally weird! 
#4




3move repetition and FEN notation
David Richerby wrote in message ...
Spur wrote: FEN notation is supposed to represent a game state fully. It does represent castling permissions, enpassant, 50move draw. But it doesn't represent 3move repetition draw. How come ? FEN only represents a position up to telling you what moves are legal. It also tells you when the 50move counter expires but, as Arnold Meijster says in his followup to your post, you need to know every move since the last irrevocable move (pawn moves or captures) to know whether there's a draw by threefold repetition in the offing. FEN is indeed enough for legal move generation, but why does it include the 50ply and fullmove counters ? These are of no importance for move generation. It just seems a bit incomplete. 
#5




3move repetition and FEN notation
On Wed, 10 Feb 2004, Spur wrote:
David Richerby wrote: Spur wrote: FEN notation is supposed to represent a game state fully. It does represent castling permissions, enpassant, 50move draw. But it doesn't represent 3move repetition draw. How come ? FEN only represents a position up to telling you what moves are legal. It also tells you when the 50move counter expires but, as Arnold Meijster says in his followup to your post, you need to know every move since the last irrevocable move (pawn moves or captures) to know whether there's a draw by threefold repetition in the offing. FEN is indeed enough for legal move generation, but why does it include the 50ply and fullmove counters ? These are of no importance for move generation. It just seems a bit incomplete. These are both valuable information. For example a chess engine might want to quicky test for the 50move case like this: if(brdhalf_moves=100) return DRAW; rather than browsing through all the stuff in a PGN file. Full moves is important information when chess clocks are used. Indeed, FEN is somewhat incomplete, and the following additional fields would perhaps be quite appropriate: Earlier repetitions of this positions, value '0', '1', '2' or '', where 0 and 1 are 0/1 earlier repetitions, 2 is threefold repetition and therefore immediate draw, '' is unknown. Clock type, value 'c' (conventional), 'i' (incremental), 's' (stmove) or '' (none) and numeric values mps, base and inc and numeric values, how much white and black have time left in their clocks measured in seconds. Again '' if this information is not known. Mikko Nummelin 
#6




3move repetition and FEN notation
Mikko Nummelin wrote:
Indeed, FEN is somewhat incomplete, and the following additional fields would perhaps be quite appropriate: Earlier repetitions of this positions, value '0', '1', '2' or '', where 0 and 1 are 0/1 earlier repetitions, 2 is threefold repetition and therefore immediate draw, '' is unknown. EPD (Extended Position Description) already does this. Clock type, value 'c' (conventional), 'i' (incremental), 's' (stmove) or '' (none) and numeric values mps, base and inc and numeric values, how much white and black have time left in their clocks measured in seconds. Again '' if this information is not known. You need more information than that to specify a time control. Consider the case where you have 90 minutes for the first forty moves and then thirty minutes plus thirty seconds per move for the remainder of the game. Dave.  David Richerby Hungry Sadistic Cat (TM): it's like www.chiark.greenend.org.uk/~davidr/ a cuddly pet but it wants to hurt you and it'll eat you! 
#7




3move repetition and FEN notation
On Wed, 11 Feb 2004, David Richerby wrote:
Mikko Nummelin wrote: Indeed, FEN is somewhat incomplete, and the following additional fields would perhaps be quite appropriate: Earlier repetitions of this positions, value '0', '1', '2' or '', where 0 and 1 are 0/1 earlier repetitions, 2 is threefold repetition and therefore immediate draw, '' is unknown. EPD (Extended Position Description) already does this. That sounds fine! Clock type, value 'c' (conventional), 'i' (incremental), 's' (stmove) or '' (none) and numeric values mps, base and inc and numeric values, how much white and black have time left in their clocks measured in seconds. Again '' if this information is not known. You need more information than that to specify a time control. Consider the case where you have 90 minutes for the first forty moves and then thirty minutes plus thirty seconds per move for the remainder of the game. Oh yes, the guillotine. I left it out as it is not part of XBoard Chess Engine Protocol and neither recognised in FICS. It would be easy to add the "simple" guillotine type to the clock type list, but more complex definitions would need a full language specification. Mikko Nummelin 
#8




3move repetition and FEN notation
Mikko Nummelin wrote:
On Wed, 11 Feb 2004, David Richerby wrote: Mikko Nummelin wrote: Earlier repetitions of this positions, value '0', '1', '2' or '', where 0 and 1 are 0/1 earlier repetitions, 2 is threefold repetition and therefore immediate draw, '' is unknown. EPD (Extended Position Description) already does this. That sounds fine! (I should clarify: EPD achieves the same effect but by a slightly different mechanism.) You need more information than that to specify a time control. Consider the case where you have 90 minutes for the first forty moves and then thirty minutes plus thirty seconds per move for the remainder of the game. Oh yes, the guillotine. I left it out as it is not part of XBoard Chess Engine Protocol and neither recognised in FICS. It would be easy to add the "simple" guillotine type to the clock type list, but more complex definitions would need a full language specification. PGN has enough of a language for this in the TimeControl tag. It's important as, while FICS doesn't support it, pretty much all tournament and match play has some sort of multipart time control. Dave.  David Richerby Pickled HiFi (TM): it's like a music www.chiark.greenend.org.uk/~davidr/ system but it's preserved in vinegar! 
#9




3move repetition and FEN notation
In article ,
Mikko Nummelin wrote: [...]. For example a chess engine might want to quicky test for the 50move case like this: if(brdhalf_moves=100) return DRAW; rather than browsing through all the stuff in a PGN file. Of course, the 50move rule is not "DRAW" but "DRAW can be claimed". The difference is unimportant in analysis, but matters in a game. [...]. Indeed, FEN is somewhat incomplete, and the following additional fields would perhaps be quite appropriate: Earlier repetitions of this positions, If you really want FEN to be complete, then we need this info not just for *this* position, but for any position that can be reached without pawn moves and captures. Too much like hard work .... value '0', '1', '2' or '', where 0 and 1 are 0/1 earlier repetitions, 2 is threefold repetition and therefore immediate draw, '' is unknown. Again, 3fold repetition is a *claimable* draw, not an actual draw. Even in computer play, it is conceivable that in a time scramble each side thinks it is winning [or could be winning] and therefore wants to play on until the time control is reached, allowing a longer [deeper] look at the position. But for general analysis purposes, even a twofold repetition is a draw, of course. In the days when we had adjudications in the local league, I once had to adjudicate a position where White had an obvious win "on the board", but that winning move led to a position that had already occurred four times [presumably in a scramble]. So the question was whether White still had a win given that several key positions had to be avoided. For computer analysis, all the stuff about 50move rules and 3fold repetition is seriously compromised by transposition tables. You analyse position A, and it's a forced win. So you move on to position B, and you find that it's a win by playing into A. But noone bothers to tell you that the win from A involves a position C [no longer in the TT] that is part of the path from B to A. Later, you analyse C itself, and you find that C is a win by playing to B, then back to C, thence to A [playing back to B now marked "drawn"!], and now the only "win" is to go back to C. Oops. Or else you find that A is a draw because from it you can reach a position that already occurred in the path to A. Later, you analyse B, and assume you can draw by playing to A; but the route via B doesn't go through the repeating position. Oops. Or that A is a win, but it takes 49 moves to conversion, so that the longer route to A via B leads to a draw instead. Getting this absolutely right is a nightmare.  Andy Walker, School of MathSci., Univ. of Nott'm, UK. 
#10




3move repetition and FEN notation
On Thu, 12 Feb 2004, Dr A. N. Walker wrote:
Of course, the 50move rule is not "DRAW" but "DRAW can be claimed". The difference is unimportant in analysis, but matters in a game. To be precise, FIDE has ruled that in some situations, i.e. kbbkq and krbkr, draw can't even be claimed before 100 moves each side due to possibilities of forceful mate lines having over 50 moves before next capture. Mikko Nummelin 