I read somewhere that an IQ level of about 130 is the threshold where one can be expected to be able to successfully self-educate in technically advanced fields.  I believe that Stanford-Binnet IQ ratings are imprecise in gauging features which accurately predict self-education potential, but it's probably a useful value in predicting success anyway.

My experience in the Silicon Valley tech sector substantiate the above.  I would suspect that the lower end of college educated participants was in the 120 range.  The (less numerous) self-educated contingent was probably in the 130-150 range and as a general tendency they excelled (perhaps because those who did not went back to digging ditches.)  The high end of participants tended to hold graduate degrees and I would estimate they to be North of 150 a lot of times.

The more I think back on things, the more I suspect that formal education focuses a lot on instilling certain ideas which are of benefit to a certain group of social engineers.  Their programs were successful in achieving a result.  The self-educated crowd were much more likely to be 'conspiracy theorists' when it came to unrelated political thought patterns for instance.  Also, being deficient in an understanding of certain documented procedures, the self-educated tended to 'go their own way' sometimes, and sometimes it worked out well.  But I worked mostly in 'start-up' companies.  I suspect that in the larger corporate environments it was usually a better strategy to just go with the industry standards.  'Blazing new trails' is a risk/reward ratio which doesn't make as much sense there.