Why does the Error versus Reject Rate curve tail up at high reject rates when testing some systems (usually poorly performing systems)?
At the tail end approaching 100% Reject, the Accept rate is very low and any error in that range will cause a higher error percentage (because denominator, accepts, is low). As you come away from the end, and add more accepts to the pile, then the error rate is lowered. This is a mathematical anomaly of having an error at a very high confidence rate.
