website statistics
Quarterback is a special positions in many ways, and we can use that fact to improve the accuracy of the Bayesian draft prediction model. You may have noticed that throughout the '00s the Colts and Patriots were not burning up first round picks on QBs. With Manning and Brady occupying the starting roles on each team, it would make little sense to do otherwise. And it doesn't have to be a lock-future-Hall-of-Famer under center to be sure a team won't be fishing in the first couple rounds for a QB. While second-tier guys like Roethlisberger, Flacco, Ryan, and other have occupied starting roles, their teams have not drafted QBs in the top rounds.

Additionally, teams that had recently spent a top pick on a QB are not going to go fishing for another very early in the draft. Even when a rookie QB had a shaky first season, for whatever reason--patience maybe, sunk costs perhaps--teams won't take a QB with a high pick in consecutive years.

These are teams I've labeled set at QB. Teams that are "set at QB," with very few exceptions, don't select QBs in the first 3 rounds. Ironically, the exceptions are the teams with the tightest locks on the Hall. Since 2011 only two such teams chose a QB in the first 3 rounds. Denver, which selected Brock Osweiler to backup Manning, and New England, which selected Jimmy Garappolo and Ryan Mallet to back up Brady, took QBs. Mallet was a 3rd rounder, and the other two were taken at the bottom of the 2nd round. Due to his age and neck injury, it's debatable how much of a lock Manning was at QB when Denver took Osweiler. Either way, we can use that information to improve on the team need adjustment recently built into the model.

For the teams determined to be set at QB, the model will now drastically adjust the probability they will take a QB. But to allow for the remote possibility they do, and more likely, allow for the possibility of a trade into the slot in question, it is not set a zero probability. The model typically allows for about a 15% chance of something unexpected happening, which is just above the chance of a trade. To hedge my bets a little, I have only applied the adjustments in the first two rounds. From the 3rd round on, the normal team need algorithm remains in effect.

Admittedly, Set at QB is a judgement call, but it's one of those things where you know it when you see it. Factors such as recent draft selections, contract extensions for current starters, recent trades to get a starter, and age and health of current starters.

To train the model I asked an experienced draft expert to help me assess whether teams were set at QB going into the draft in recent years. I combined those assessments with my own. This year, these are the teams we have determined to be set at QB: ATL, BAL, CIN, DAL, DET, GB, IND, JAX, KC, MIA, MIN, NE, NO, NYG, PIT, SF, and SEA. I recently removed SD from the list given the rumors surrounding Philip Rivers.

It's not an exact science, but I think it gets is closer to the truth in the vast majority of cases. And it results in a more accurate picture of when and where prospects will be selected this weekend.