I wish I shared your optimism about Howell. The likelihood he goes from well below average to well above average is probably less than 5%.
My calculations show 17.5%
I don't know how you came up with that. I'm getting 13.329% to be precise. And I'm using tried and true Kelly's QB bluebook.
I know
@cymatica and
@Shane Falco are joking about the probability calculations, but there are a few different definitions of probability, all compatible in certain limits, but representing different approaches, and under some of them, different people could legitimately come up with a wide variety of different probabilities for the same event.
Some use the definition that only defines probability in the unreachable limit of infinite observations, specifically, that the probability of a given occurrence is the limit as the sample size goes to infinity of the "relative frequency" (number of times it occurs divided by the total sample size, so if something happens 357 times in a thousand observations, the relative frequency is 35.7%).
Others define probability differently, and some of them make more explicit the subjectivity that comes into assigning probabilities with finite samples, especially small ones. The great Bruno de Finetti famously wrote "probability doesn't exist!" What he meant by that is that probabilities are subjective. He still defined rules for coherently defining and assigning probabilities that are the basis for modern Bayesian statistics. In the limit of large samples, the different definitions of probability all end up converging to the same thing, and de Finetti's approach to probability theory allows a person to make good estimates even when sample sizes are small.
In something like an NFL draft, once you consider things like team situation, player history and age, what other players are available, and what the situations of other teams are, the relevant sample sizes will generally end up small, so the familiar "relative-frequency" definition of probability is basically useless, even though relative frequency is still really important for estimating probability distributions no matter what definition of probability you're using.
So as a result, I suspect different teams really
do have very different assessments of the probability of a given college player succeeding in the NFL. In fact, in some cases, they probably vary more than your estimates of probabilities (a serious "less than 5%," and joking estimates of 17.5% and 13.329%) have varied.
@Shane Falco even brought in another important point, the number of significant figures. Carrying probability estimates like these (even the ones that "use yesterday's
posteriori as today's
priori" and are informed by data from previous drafts and NFL careers) to five significant figures is silly.