On the one hand I agree with the criticisms on survivorship bias and the risks of hypothesis testing. There's also presumably a selection bias. And while Ron Berman's post on methodology [1] talks about how the goals of this stage of the project were to formulate a model that could then be verified, the results are certainly being reported as conclusions.

On the other hand, despite its limitations, it's very interesting data. So the title seems very much an overstatement. "Blindly Believing in Startup Genome Project's Data Considered Harmful" would be a better description -- although far less effective link bait :-)

[1] http://www.systemmalfunction.com/2011/05/deciphering-genome-...