[21]

Bayesian statistics, there are different priors.

The probability of a convicted person to do something criminal again is not the same as the risk of an random unconvicted person, just like the risk of a flood reoccurring in a previously flooded area:

> According to an April 2011 report by the Pew Center on the States, the average national recidivism rate for released prisoners is 43%.[2]

> According to the National Institute of Justice, about 68 percent of 405,000 prisoners released in 30 states in 2005 were arrested for a new crime within three years of their release from prison, and 77 percent were arrested within five years.

> According to a national study published in 2003 by The Urban Institute, within three years almost 7 out of 10 released males will be rearrested and half will be back in prison.

https://en.wikipedia.org/wiki/Recidivism

[mrtksn]

It's not that simple.

Some ridiculous percentage of startups fail, yet not all startups have the same risk of failure.

That's also why it's not a good idea to dismiss a candidate on a single criterion unless you're randomly hiring.