Changes

Written by Johan de Koning, already known for his program which has competed for several years in [[Dutch Open Computer Chess Championship|Dutch tournaments]]. This latest version runs on the [[ARM2]] RISC CPU and represents 2.5 man-years of development. The [[Evaluation functionFunction|evaluation function]] is complex and hence a little slow. Positional [[Score|scores]] can "easily exceed several Pawns". Separate evaluation terms are used for [[Opening|opening]]/[[Middlegame|middle]]/[[Endgame|endgame]] with [[Tapered Eval|smooth transitions]] between them by using 3 weights which sum to 100%, giving a fuzzy definition of the [[Game Phases|phases]]. The [[Search|search]] techniques include [[Check Extensions|check extensions]] and chess-specific static rules for additional [[Extensions|extensions]]. [[Singular Extensions|Singular extensions]] are not used, but [[Lower Bound|lower bounds]] are computed to limit the search in the [[Selectivity|selective]] phase. [[Best Move|Best moves]] from earlier [[Iterative Deepening|iterations]], [[History Heuristic|history heuristic]], [[Killer Heuristic|killer heuristic]], priority to [[Captures|captures]], and a [[Refutation Table|refutation table]] are all used to [[Move Ordering|order moves]], which are [[Move Generation|generated]] in the all-and-sort fashion. Specialized endgame knowledge is included and this is actively being expanded.