nearly three times the nominal rate. - The probability that at least one peek shows significance is **22.1%** --- meaning more than one in five experiments will produce a false alarm at some point during the test, even when there is no real effect.