Ask HN: How true is it that ninety-something percent of startups fail?

In particular, what counts as a "startup"? i.e. what's the denominator in (# startups that succeed) / (# of startups)? Given other signals, e.g. if the founders have quit their jobs to work on the idea full-time, how does that ~90% number change? Especially if founders have legitimate credentials such as Stanford CS degree. From experience, the percentage of serious startups (ie founders working full time) started by Stanford CS/MBAs that succeed is orders of magnitude higher than the number that is typically thrown out. Admittedly, that may just be survivorship bias