If the most frequent type in the general population makes up 13% of the total, the baseline for your "ability" to type others would be 13%, because you could just type everyone as being of the most frequent type.
A past analysis of the forum population's self-types showed that 15.0% of those who indicated a type considered themselves to be IEI.
If I remember the analysis I once did of the level of correspondence between an individual's declared self-type and their result on the test that I compiled (of which there have been three editions, with one in existence), it was about 40-50%.
Doing the analysis again but based on what I considered the actual type of each test-taker (if I had an opinion, and done independently of test results), I seem to recall it was more like 60-65% for the second edition, and perhaps lower for the other editions. This was based on a small sample, selectively emphasizes one edition over the others, assumes that test-takers were "accurate" (whatever that means) in their answers, that I had a better understanding of the personality of the test-takers than they did themselves, and so on: this paragraph is not meant to be taken seriously, but merely to illustrate a "best possible" scenario.
Based on these considerations and on my years of observation which has seen people being absolutely certain of their own types and the types of others, and having utter faith in their own typing ability, my position is that the best method of typing consistently is with a good test, and that in general, typing ability is unlikely to exceed 40-50%. (There is obviously a slight absurdity in testing validity by something which is not fixed...). A more reliable and more heavily researched personality system, such as the Big Five (or perhaps, exclusively the Big Five) may be able to achieve better results than that, and with greater meaning.
If this is true, then this may mean that when typing (namely, choosing one of the 16 types), in general, more likely than not, you are incorrect, and with no real way of finding out! It may mean that when typing others, especially individuals you do not know or are not especially familiar with, you have even less confidence. (To be fair, you'd expect it to be more difficult to correctly determining one of 16 types rather than one of eight, four, two etc. - more difficult than telling if someone is either highly Agreeable, lowly Agreeable, or Not Readily Determinable: perhaps being able to determine what general flavour a person is can be of great utility, without needing to be precise).
...So, what do you estimate your Typing Ability as, taking into account past experience and perhaps also the Spheres in which you are typing (e.g. people you know, people online, people you don't know...and famous people you don't know)?