APFD demonstrated that the random ordering outperformed the repeated coverage maximization for GCC while not for Jikes and Jvac.
As the granularity level decreased, higher APFD values were observed.
They also stressed that choosing a technique with higher APFD is oversimplifying and may not always imply a better technique.
It also proved to be the main motivation behind many evaluation measures such as APFD etc.
10, APFD came forth as a striking measure for computing the Average Percentage of Faults Detected and a massive of 29 studies [2-9, 23, 24, 26-28, 31, 33-35, 41, 46, 47, 53-55, 57, 62, 69, 73, 77] took advantage of the metric directly.
All the metrics spotted in the studies availing APFD or its mutants summed up to be more than 50%.