Colin Kaepernick, Tactlenecks, and QB Predictiveness























When not showing off the Christian tattoos, Colin Kaepernick is probably rocking the tactleneck.


In an unusual detour from our normal practice, this is a choose your own adventure post where you can read the more esoteric parts of an overly long RotoViz article here but need to go over there to read the rest. Let me know what you think in the comments, or just throw in some bizarre links to stuff I don’t understand. The spam filter won’t catch all of them.

Drafting well is about risk management. . .

Of course, it is also about reward management. One of the debates this offseason has centered on the question: Can you wait to draft a quarterback? A parallel question might be: Is that really the risk-averse strategy if you have a high rating on a QB?

The value the Seahawks and 49ers accrued by selecting Colin Kaepernick and Russell Wilson far outweighs concerns about risk. It appears they both have franchise superstars to build around for a decade. If they’d had any sense of what types of quarterbacks they were drafting, it would have been insane to incur the risk inherent in waiting to select them. Any number of teams could have picked their guy before they had the chance.

Of course, it’s only a risk if the amazing performance was actually predictable. Psychological overconfidence tends to create the impression that we can isolate these types of players much more often than we really can. One of the key points in the battle between scouts and analysts is that analysts tend to be less confident even though the numbers suggest we’re also more accurate. Even if we weren’t more accurate, having less confidence would probably lead to a better allocation of draft resources.

Anyway, that’s a long way of introducing the concept we’re hearing a lot about this year: waiting until the second or third round to take a flyer on your franchise quarterback.

Is NFL Success Predictable at the QB Position?

While simple WR projection algorithms can be shown to dominate scouting-generated rankings, the QB position is not quite so simple. (That said, Jon Moore does an excellent job breaking down Ryan Nassib and explaining how analysts would have pulled the trigger much earlier on Tom Brady.)

Many analyst-generated models for quarterback projection haven’t held up particularly well. For those of us who like stats there’s the temptation to make things complicated, but that’s often a questionable instinct. Karl Popper’s interpretation of Occam’s Razor suggests the simpler the model, the greater the empirical content and the easier to falsify. Many current statistical projection models probably describe noise instead of underlying principles. (Others assign ridiculous leverage to stats that amount to nothing more than historical artifact – Total QBR, please stand up.)

Continue reading Colin Kaepernick, Tactlenecks, and QB Predictiveness