Changes

In 1986, at the [[ACM 1986|17th ACM North American Computer Chess Championship]] in [https://en.wikipedia.org/wiki/Dallas Dallas], Fidelity showed up with two experimental machines. Fidelity Experimental with [[Danny Kopec]] as [[:Category:Opening Book AuthorsAuthor|Book author]] had one [[68020]] CPU, where three models also played the [[WMCCC 1986]] some streets farther with compatible time schedule. [[Chess Challenger|Chess Challenger X]] was a parallel system with a [[Z80]] controller, and 16 or more [[68000]] CPUs, co-authored by [[Ron Nelson]], with a Kopec book as well.

The learning feature is just one facet of a multi-faceted chess algorithm. The program is basically [[Brute-Force|brute force]] in origin with evolution to incorporate extensive positional analysis and [[Extensions|selective extensions]] during the [[Quiescence Search|quiescence search]]. The positional analysis incorporates extensive heuristics for [[King Safety|king safety]] and [[Pawn Structure|pawn structure]]. Numerous end-game specific routines are incorporated, icluding [[KBNK Endgame|mate with bishop and knight]], complete evaluation of [[KPK|king and pawn vs. king]], probably outcome of a [[Pawn Race|pawn race]], [[Rule of the Square|square of the pawn]], [[Wrong color Color Bishop and rook pawnRook Pawn|bishop and rook pawn of the opposite color]], the [[Philidor Position|Philidor]] and [[Lucena Position|Lucena positions]] and others. Dynamic recognition of minimum mating material, [[Fifty-move Rule|fifty move rule]], and [[Repetitions#RepetitionOfPositions|repetition of position]] assist in forestalling heartbreaking draws in otherwise won positions. The search algorithm uses a [[Depth-First|depth first]], [[Alpha-Beta|alpha-beta]] search with the [[Null Window|zero width window]] technique ([[Principal Variation Search|PVS]]). The search proceeds [[Iterative Deepening|iteratively]] with a quiescence search incorporating [[Captures|captures]] and certain threats appended beyond the [[Depth|nominal depth]]. The program will not perform an [[Evaluation|evaluation]] on a position where either king is [[Check|in check]]. The check must first be resolved by showing the existence of an escape move or [[Checkmate|mate]]. Iterations are finally halted under the direction of a [[Time Management|time control algorithm]] which is dynamically incorporated for up to 40 moves in the root position. Two [[Killer Move|killer moves]] are stored at every [[Ply|ply]]. The program performs a preliminary sort on the ply above the quiescence search. The search is supported by extensive [[Transposition Table|transposition tables]] incorporating random numbers selected using [[BCH Hashing|BCH theory]].