I was writing a long post about how we think we're really good at detecting "important signals" like if someone's lying, but that we're actually terrible at it. But the new yorker article that was frontpaging yesterday should pretty much explain it, even though Jonah Lehrer is a bit smug for my tastes ;)
TL;DR- the OP sounds like a very shallow-thinking VC who's blissfully unaware of his own biases. It might work out ok for him -- these biases usually work most of the time, and that's why we have them -- but it will probably fail in some very crucial instances where a more thorough, data-driven approach might have worked better. In fact, that may be precisely how the great VC's are differentiated from the merely ok, but that's a different story...
(link: http://www.newyorker.com/online/blogs/frontal-cortex/2012/06...)