To check a claim of special ability, start by assuming the person has no ability at all and is only guessing. Then ask how likely their results would be if that were true. Only results that pure guessing would rarely produce count as evidence of real skill.
An experiment can be strict enough that only a perfect score rejects 'just guessing': with eight cups, flawless sorting happens by luck about 1.4% of the time. But clearing that bar shows only that some ability exists. It says nothing about how good the ability is.
Whether a small question, like if milk-first tea tastes different, can be answered depends on how the test is arranged beforehand. That means deciding how many trials to run, how to order them, and what result would count as proof. Treating this arrangement as a science of its own gave birth to modern experimental design.