Not sure why I'm bothering, buy basic statistics and psychoacouatics here we go...
Sure, any one person in a group of blinded test subjects may in fact be hearing a difference and correctly responding rather than guessing. In fact, many may be. But the results will actually reveal this to be happening... you just can't tell who exactly the magic ears are.
Here's how it works. A test comparing say CD and SACD in this example doesn't ask test subjects which sounds better, it only needs to ask test subjects if they can tell a difference between A and B... between CD and SACD. The answer is either yes or no, either they can or can't. Now, start from the premise that people can't distinguish the two and thus will be guessing. Over a large number of trials with many test subjects, guessing will produce half getting the answer correct by chance and half getting it wrong by chance.
Now I troduce your proposition that some of those people may be getting it correct not because of guessing but because of superior hearing acuity. They really can hear a difference, and thus get the answer correct more often than guessing alone would produce. These people have correct results recorded. But the rest of the subjects are still getting half correct by guessing. When you average together those subjects who always or usually get it right due to good hearing and those guessing, you will find somewhat higher than 50% of the answers are correct. The more people getting it right without guessing, the greater the percentage.
Now, you know if you flip a coin you may get runs of heads or tails. If you flip it ten times, you won't always get 5 of each. If you flip it 100, you will expect to get something around 50 off by only a few. If you flip it 10,000 you will again expect to be off by only a few one way or another. If you test 500 people and are off by only a few from half, this is what you would expect from flipping a coin or guessing. If there are people in that group not relying on guessing, their number must be very small or else it would be highly improbable for the remaining pool of guessing subjects to just happen to get it right too few times to offset these good listeners. The more trials you to, the more confident you become in your results.
In the case of the CD vs SACD test in question, only a couple of people at best could be getting the answer correct without relying on guessing without driving the correct percentage significantly above half. You can calculate the probability of obtaining these results with 'x's number of golden ear test subjects if you like. You find it becomes quickly quite improbable as you add each single good listener.
On the flip side, however, if you could isolate that good listener and then test him individually over a decent number of trials, you shoud quite readily discover that his answers are correct significantly more than half the time. If this is the case for a single listener, then as you increase the number of trials in which he gets it right it quickly becomes highly improbable that his answers are the result of chance.
Thus, while it takes a large number of trials with a large number of listeners producing results near 50%, such as the test referenced earlier, to become confident that at best very few people can hear a difference (and yes, this can never conclusively prove that it is impossible that at least one person may exist somewhere who can really hear the difference), it takes but that one, singular good listener to prove beyond doubt that at least one person can really tell the difference. Just one person to prove without doubt that the difference truly is audible.
This is an important point, because even though it only takes a single listener to effectively end the debate about whether such things as cables, reasonably designed dacs, reasonably designed amplifiers, transports etc may at least sometimes truly sound different, you won't find one documented case of this happening during the decades these debates have been ongoing. Not one single example when the test was properly controlled, conducted and documented. All it takes to prove your cable makes a difference is just one listener. One. Out of the thousands upon thousands who regularly claim to easily hear significant repeatable differences, not one has demonstrated this ability in a controlled manner. Not once. Not for cables, dacs, transports, amplifiers... never. Large sums of money and other incentives have been offered to people who claim the difference is obvious to step forward and prove it. People who say their wives or moms can easily hear or see the difference from the next room. No proof. None. After decades of begging. Why?
Now, put that absolute lack of proof from the thousands who claim the differences are night and day in the context of decades of controlled testing such as the aforementioned CD vs SACD test, all showing at best only a couple of people out of hundreds may not need to rely on guessing. What is the natural conclusion?
Even more convincingly, put all of these listening tests in the context of measurements and what we know about the science of human hearing. Decades of other research and human testing has established what our threshold of hearing is with respect to distortion levels of various types, frequency differences, volume differences, etc. Measurements made with instruments of such accuracy and resolution and repeatability that it really boggles the mind. If two cables have measurable differences in their impact on frequency response or volume that are orders of magnitude smaller than what humans have demonstrated an ability to hear, what do you think of claims that the differences are night and day? Or dacs with distortion and linearity differences mere fractions of what decades of tedious research has established is audible? Either the claims are suspect, or there is something going on in a process as simple as AC current flowing through ordinary wire at quite mundane frequency and power bands that is entirely unknown to science. Science which has relied on intimate and exacting knowledge of how electricity behaves in wire under frequency or power or operating conditions quite extreme compared to audio to make mind bogglingly accurate predictions about how our world works, down to the most amazing minutia. Does the guy assembling cables in his garage really know something profound about this science that the guys at CERN do not?
And finally, if you're asking yourself "but how could all these people be so confident and yet so wrong... myself even"?! Not to fret. There are additional mountains of research documenting the fallacy of human perception and the power of expectation bias, persuasion, and placebo. People can hear, see, feel all sorts of things that are only in their minds. Placebos in medicine can be quite powerful. Persuasion allows magicians to do amazing things that your senses are sure are impossible. And we have documented that expectation bias is a powerful manipulator of what we hear. These things are so powerful, in fact, that they can override changes that really are there. We tend to "prefer" stuff that is bigger, prettier, shinier, more expensive, even when their measured performance is inferior and even when listening tests under blinded conditions demonstrate that we actually prefer them less based on sound alone. When you buy an expensive cable, it is quite likely that it will in fact "sound" better to you, even if underneath the pretty shielding is a rusty coat hanger. That's because hearing is a process in which the brain takes physical sound received by the ears and processes it in an amazingly complex way to mask noise, localize sound, and finally recognize patterns based on experience to put the sound in context. Unfortunately since the brain and its processing are required, we can't separate it from the physical sound, nor isolate it from the influences of suggestion and bias that both plague and rescue all human perceptions. You aren't alone in thinking you hear a difference. I hear one too, even though I know beyond any doubt that none exists physically. I can no more control or eliminate these influences than you can.
If you have made it this far, and are a rational man, do your not at least begin to question what you had already accepted as obvious and "true"? Do you begin to question your own perceptions when it seems so clear? The evidence, on all fronts, is absolutely overwhelming.
Your final question... but why do people bother to make these cables? Why do engineers bother wasting their time designing ever more accurate dacs? Because people will always buy them.