On the Inaccuracy of SackSEER

Jimmy Kempski:

I think just about any knowledgeable football fan that is familiar with Football Outsiders is generally very complimentary of the work they do (as they should be), but SackSEER has its fans and detractors. The model was formed from data collected prior to the 2010 season and was first introduced in 2010. Therefore, the the model pre-2010 is understandably extremely successful. Its fans point to the pre-2010 success. Its detractors feel that predicting a player’s future in the NFL based on a couple jumps and a few passes through some cones at the Combine is a little ridiculous. It also is a pure data model that ignores things like watching game tape.

Jimmy rightly points out the utter failure this model has been in the last two years, although I’d make two additional points. First, in defense of SackSEER, don’t read the projections as predictions, but rather the most likely general range based on past precedent. For example, to knock the model for projecting “only” 36.4 sacks in his first five years for Von Miller is silly. It argued that he would be good and he is. Jerry Hughes, on the other hand…

That said, any model that leans too heavily on something as luck-dominated as sack totals is going to have trouble. I wonder if a better SackSEER could be constructed based on total pressure (sacks, hits, hurries).