DC/TO overlays with prior season records

This in some ways is not a bad thing in that logically it would seem to make sense that big overlays should offer more value than small ones, but it got us to thinking about an often talked about subject in horse racing -- identifying a true overlay as opposed to a false one.

Why would an actual point spread be so far removed from the projection of the DC/TO? Well of course on one hand we use turnover data in a contrarian fashion which can explain a lot of it, and then you encounter situations like the 'star starting QB is out with an injury" where the line is moved a ton from where the statistics might have it.

In general though, the goal is to realize when a number is posted that's based on some less relevant standard or thinking, and with the NFL having so few games the obvious place to start is with the prior season records. People rate teams with last season's results firmly in their minds, when for the NFL things can change radically from season to season.

Here is some performance numbers then against the spread looking at the prior season win number (regular season only) for the team deemed to be the overlay:

Well, the tale of the tape comes in somewhat as expected, with the high-win teams of the prior season being the worst overlay plays overall, although a still profitable 54% (and 58% in the 8+ zone). Teams with less than ten wins in the prior season hit a lofty 60% as 6+ overlays, suggesting that the perceived stats value was genuine and that the holdover expectations and preconceptions about these less than stellar 'year-ago' squads were artifically moving the spread.

Of course it's reasonable to think that this effect would vary some by the stage of the season, so here are the summary numbers by quadrant of the football calender:

The first few weeks of the season marks the most significant carryover of a team's prior season record in terms of expectations, but unfortunately the DC/TO predictions also produce huge overlays in this timeframe since there are so few games played. While two periods show out of the ordinary success rates (weeks 9-12 for the less than seven win teams, weeks 13-17 for the seven to nine win teams) there's not a huge amount of games in those groups to believe that these trends have any strong likelihood of reoccurring.

The conclusion we would take from all this is that when the DC/TO overlay points to a team that was not among the best in the league the prior season, your chances of having a true overlay are good!

Here we see some evidence that the poor prior season teams (less than seven wins) are very strong dog plays when picked from the drive stats projections in the early stage of the season, but taper off thereafter. Surprisingly the mid-quality teams (7 to 9 wins) are fabulous dog overlays in the second half of the season, but less so in the early going.