no edit summary
Quote by [[Chrilly Donninger]] from ''CHE: A Graphical Language for Expressing Chess Knowledge'', 1996 <ref>[[Chrilly Donninger]] ('''1996'''). ''CHE: A Graphical Language for Expressing Chess Knowledge''. [[ICGA Journal#19_4|ICCA Journal, Vol. 19, No. 4]]</ref>
The main design criterion for successor [[Nimzo|Nimzo-3]] was combining the positional play of Nimzo-2 with the [[Tactics|tactical]] strength of a program like [[Fritz]]. Nimzo-2 was developed on an [[x86|486/50 MHz]] [[IBM PC|PC]], which calculated about 3,000 [[Nodes per
second|nodes per second]]. Thereof 60 to 70 percent was spent/wasted in the [[Evaluation|leaf evaluation]] routines. Hence, a major improvement in speed and thus in tactical strength could only be obtained by performing most of the evaluation either in the root or the interior of the [[Search Tree|search tree]]. So, Nimzo-3 became a [[Chess Genius|Genius]]/[[Fritz]]-like program with a complex root evaluation, called Oracle, similar to [[Hans Berliner|Berliner's]] Oracle, and with a very simple, mainly first order evaluation at the [[Leaf Node|leaves]] <ref>This approach seems to have been invented by [[Kaare Danielsen]] for his program [[CXG Star Chess#Advanced|CXG Advanced Star Chess]], Quote from [[Chrilly Donninger|Donninger's]] CHE paper</ref>. Nimzo-3 spends about 10 to 20% on leaf evaluation, its node rate has increased by 400% up to 12,000 nodes/second on the same hardware.