You wonder if the random selection is really some sort of program / formula that highlights unusual performance spikes as a criterion. Something like [(% annual change xBH)/ |(% avg change in xBH for age]| ]* WTF > 150 trips a test, where 100 is normal.