There can´t be a serious estimation on that number; based on my experience i know, that people are very different in their ability and in their self description.
Well, I have tested that ability professionally. When I was at Microsoft I thought we as audiophiles must have better ability to hear compression artifacts than general population. We did a large scale test and found that self-selected audiophiles were no better than the general public. Artifacts that I and our other trained listeners could hear was not audible to them. We were desperate to find more testers through them to help advance our audio technology but the conclusions were quite strong: audiophiles could not hear these non-linear and small artifacts.
Similar data is confirmed by Harman which showed audiophiles, reviewers, sales people, etc. while having similar preference to trained listeners, are quite unreliable in their voting/detecting linear artifacts (frequency response deviations). When I attended a similar test at Harman with a group of audio dealers, they all failed the test at very high levels of impairments. I did much better but Dr. Olive did far better than me. It is clear that audiophiles and their dealers simply don't know how to search, find and identify artifacts properly.
Yes, there is an occasional audiophile that has good ability here but it simply is not the norm.
All of the above listening tests used stimulus that had provable audible impairments. So imagine what I think of the performance of audiophiles when such provable distortions are not there, i.e. cables, power conditions, etc. To the extent they fail the tests I mention yet declare all audio tweaks as night and day, veils removed, etc. tells me that they can't possibly be right. Such differences if audible, will be incredibly subtle. No such description will come from it.
There has also been rare situations where these audiophiles were subjected to modicum of control, e.g. single blind testing, and they all fail and fail catastrophically. The moment they close their eyelids, all of the differences that they say were night and day disappear. I have been in the same boat and know the feeling of being so wrong.
It is on the basis of nearly two decades of experience that I say the outcomes reported in these specific areas is 100% unreliable. Junk if you will. It is not some random hyperbole that I am throwing out there.
Of course I can be swayed to change my opinion. Let's set up a controlled test of the category of devices I mentioned and see the success rate. Folks will bulk and hence the reason I offered my $1,000. In absence of that, I can't give any benefit of doubt here.