As expected in the initial generations crossover is highly disruptive (cf. Figure 9) with only 19.5% of crossovers producing a child with the same score as its first parent (i.e. the parent from which it inherits the root of its tree). However after generation 4 this proportion grows steady, by generation 18 more than half have the same fitness. At the end of the run this has risen to 69.8%.
The range of change of fitness is highly asymmetric; many more children are produced which are worse than their parent than those that are better. By the end of the run, no children are produced with a fitness greater than their parent. Similar behaviour has been reported on other problems [12] [21, page 183,] [15, Chapter 7,].
Figure 9:
Fitness relative to First Parent,
Normal runs,
Means of 50 runs.