You're just pretending you can calculate inherently incalculable probabilities. How do you determine the probability a company will fail under a variety of potential CEOs so that you could even get a mathematically-justifiable rank order? You can't run trials, you can't really extrapolate from past performance, you're subject to all kinds of confounding variables, you can't actually treat 'failure' as '$0 market cap', etc, etc, etc.

The best you could ever hope to do is "CEO candidates with this kind of background tend to have this kind of record when taking over this kind of company", which I'm sure is quite an illuminating sample.