I have some software that operates non-deterministically. Running a test on the code can result in a pass or fail depending on unknown external factors. Assuming that the odds of passing or failing are equal because of these factors, how many times must I run the test, and how many times must the software pass the test so that I can say with 95% confidence that the software works as expected 90% of the time?
Edit: Assume that the odds that an external factor causes the test to fail are unknown instead of a pass/fail being equal.