QUOTE(rutra80 @ Feb 2 2004, 09:33 PM)
I thought that double blind test is when files under both A & B and X & Y buttons get swapped, and if only X & Y is swapped then it's a signle blind test... Am I wrong?
Yes. Double blind means that the person performing the test (here the listener) as well as the one who provides test objects (here the software

) doesn't know what's tested. The term 'double-blind' makes more sense for e.g. medical tests: both, the patient as well as the doctor who hands him the medicine, don't know if he gets the placebo or the real medicine. Otherwise (if the doctor knew what the patient got,) the way the doctor behaves could influence the result.
QUOTE
And what's with the N/A issue?
If the number of trials is not fixed before the test starts (e.g. "I'll do 20 trials, not more, not less") The p-value (Probability you're guessing) is higher than the one calculated by fb2k (or other ABX tools). The more trials you are willing to take before you give up (if you don't reach a certain p-value), the higher the 'real' p-value gets. I assume that the limit with foobar2000 has been introduced to avoid the danger of false-positive results related to this.