In the editions where IQ = int x 10, it's not conjecture. Those editions comprise a large majority of D&D.
Can you actually state which editions you mean?
The ones for which I think there is no rule that IQ = INT*10 are OD&D, AD&D (at least 1st ed, maybe 2nd as well), 4e, 5e, and Moldvay/Marsh/Cook B/X. The remark by Brian Blume in a semi-humorous article, written at a time when the main function of Dragon was to publish house rules and advice, is not a rule for any edition of the game (though perhaps Blume handled things that way at his table).
Please re-read my multiple posts on the inappropriateness of using means, SDs, and normal distributions with ordinal data.
<snip>
It is utterly false that an ordinal set of data can posses a true mean, a true standard deviation, or be a true normal distribution.
<snip>
Person A has an IQ of 50. Person B has an IQ of 100. Person C has an IQ of 150. If these values were nominal, you could say that B was twice as smart as A, and C was three times as smart as A, but only half again as smart as B. But IQ is ordinal, which means we cannot tell how much difference there is between numbers, just that the numbers are ranked in order.
Is the above relevant to the discussion with [MENTION=23751]Maxperson[/MENTION] about IQ?
IQ scores provide a ranking, and also indicate a likelihood - so IQ 180 means that X% of people will be at or above that IQ (for some relatively small X, many decimal places less than 1).
So does the 3d6 roll: only 1 in 216 (= (1/6)^3) people will have INT 18.
But 1 in 216 is approximately 0.5%; hence an 18 INT is far more common, in the imagined population of D&D PCs, than is 180 IQ in the real population of human beings.
As far as I can tell, nothing in the above reasoning requires treating IQ, or INT, as measuring some quantity; nor does the notion of a "normal distribution" play a role in the reasoning; nor does the reasoning require treating the mean or SD of any data set as meaningful in the way that you are objecting to.
I'm not a mathematician or statistician, but to me it seems to be a rather straightforward argument about likelihoods.
(And since writing this reply, I see that [MENTION=6787503]Hriston[/MENTION] has said the same thing as above in post 816.)