Guard Heuristic,
a move ordering and pruning heuristic introduced by Yi-Fan Ke and Tai-Ming Parng in the 1993 ICCA Journal, Vol. 16, No. 2[1] . The guard heuristic was successfully applied in chinese chess and compared with the other heuristics such as the killer heuristic, and also proposed for chess. A similar move ordering algorithm was mentioned by Carl Ebeling in his 1986 Ph.D. thesis All the Right Moves[2] , utilizing the value of the moving piece, the value of the captured piece if any, and the value of the piece and pawn guards. In contrast, the guard heuristics is based on the guard value of each square, and considers those values not only for target squares but also for origin squares of moves to decide whether the move is searched at all, with a reduceddepth, or when.
The guard value of a square is the sum of the capture strengths (CS) of those pieces that are able to capture a potential opponent piece on that square. The capture strength of a piece is roughly inverse proportional to its point value, in white point of view metric, positive values for white and negative values for black. Further, if the square is occupied by a defended black piece, and target of a capture, the (negative) capture strength of that piece is added to the guard value as well.
Pseudo-Code
int guardValueWPOV(enumSquare sq){
isAttacked =false;
isDefended =false;int gv =0;for( all Pieces p controlling sq ){
gv += captureStrength[p];if( isWhitePiece (p))
isAttacked =true;else
isDefended =true;}if( isAttacked && isDefended ){
p = board[sq];if( isBlackPiece (p))
gv += captureStrength[p];}return gv;}
Capture Strength
Chess
King
Queen
Rook
Bishop
Knight
Pawn
White
+1
+1
+2
+5
+6
+9
Black
-1
-1
-2
-5
-6
-9
Chinese Chess
General
Guard
Elephant
Warrior
Horse
Cannon
Soldier
White
+1
+9
+9
+2
+5
+4
+9
Black
-1
-9
-9
-2
-5
-4
-9
Pruning and Ordering
The guard heuristic aims at both rejecting moves at deep nodes and ordering moves to reduce the search time. Moves with unsafe target squares, that is negative guard values less some margin, are pruned near the tips and the remaining moves are ordered by the Escape-first-if-origin-unsafe, giving priority for most valuable piece, and Capture-first-if-target-safe rules, considering MVV-LVA.
a move ordering and pruning heuristic introduced by Yi-Fan Ke and Tai-Ming Parng in the 1993 ICCA Journal, Vol. 16, No. 2 [1] . The guard heuristic was successfully applied in chinese chess and compared with the other heuristics such as the killer heuristic, and also proposed for chess. A similar move ordering algorithm was mentioned by Carl Ebeling in his 1986 Ph.D. thesis All the Right Moves [2] , utilizing the value of the moving piece, the value of the captured piece if any, and the value of the piece and pawn guards. In contrast, the guard heuristics is based on the guard value of each square, and considers those values not only for target squares but also for origin squares of moves to decide whether the move is searched at all, with a reduced depth, or when.
Table of Contents
Guard Values
The guard value of a square is the sum of the capture strengths (CS) of those pieces that are able to capture a potential opponent piece on that square. The capture strength of a piece is roughly inverse proportional to its point value, in white point of view metric, positive values for white and negative values for black. Further, if the square is occupied by a defended black piece, and target of a capture, the (negative) capture strength of that piece is added to the guard value as well.Pseudo-Code
Capture Strength
Chess
Chinese Chess
Pruning and Ordering
The guard heuristic aims at both rejecting moves at deep nodes and ordering moves to reduce the search time. Moves with unsafe target squares, that is negative guard values less some margin, are pruned near the tips and the remaining moves are ordered by the Escape-first-if-origin-unsafe, giving priority for most valuable piece, and Capture-first-if-target-safe rules, considering MVV-LVA.See also
Publications
Forum Posts
External Links
References
What links here?
Up one Level