Do prior season offensive/defensive rankings have predictive value for Season Win Betting?

We've used rankings successfully as predictive tools in a number of football wagering areas. Does prior season offensive/defensive status change how likely a team is to go over or under the season win line? It would seem in theory to be likely that the lines set for teams in a given season are not only a product of their past history, but also a matter of bias from the linemakers and public towards certain team characteristics.

For instance, one might theorize that high-scoring teams are usually public favorites since the public likes seeing lots of points, and latches on more easily to offensive star players than defensive ones. Consequently, the extension might be that strong offensive teams in the prior season are more likely to go under the season win number the next time out.

We begin with the setup -- we'll use rankings for a team in points scored and points allowed (admittedly an imperfect measure since defensive and special teams scoring is factored in) to judge offensive and defensive capability. Teams that rank from 1-10 get a 'class 1' designation, from 11-20 gets 'class 2' and 21+ ranking is 'class 3' grade.

Add the offensive and defensive ranks together and a team can rate from '2' (top 10 in both offense and defense) to '6' (worst class in both). Season win pushes are ignored in the results.

Comments: the dominant teams of the prior season (good on both points scored and points allowed) have been poor 'over' bets (and therefore profitable 'under' bets) the next season against the season win total. Teams that were weak in the previous season have been generally good 'over' plays, particularly if they rate as average in one of the two domains.

A good start, but let's isolate the offense and defense further:

So our earlier theory is validated by the data -- strong offensive teams in the prior season are solid 'under' bets. Curiously average points scored teams have done fine in besting expectations.

The defensive (or points allowed) breakouts are amazingly similar to the offensive ones -- the good prior year defensive teams are best as 'under' plays, while average defensive teams have gone 'over' with regularity.

Another area for investigation might be the year to year change of a team -- sides that perhaps were good one year at some category, but struggled the next. Do teams bounce back, or is a decline a sign of serious problems?

Teams which were top ten in points scored two season prior, but bottom ten in the last season are just 1 for 9 as over bets! Signifying that when an offense goes south it may be gone for some time! [St. Louis is the team in 2003 with this stigma on them]. Meanwhile, teams that have been top ten in points scored for two straight seasons are 35% over sides, which is basically about the same as looking only at prior season top ten data.

On the other hand, teams which were poor in points scored two years ago, and were either average or poor last season are 63% over types in our sample [Bengals, Bills, Browns, Cowboys, Giants, Jaguars, Lions, Panthers, Redskins). Finally average offensive teams two years back that improved to good in the most recent year have been 27% over plays.

Switching to defense, teams which were top ten two years prior but bottom ten last season have been 77% overs [the Bears qualify on this for 2003, as do the Rams -- nullifying the strength perhaps of the previously mentioned case]. Bad defenses two years previous that were good in the most recent season are just 20% over (look out Colts, Falcons, Panthers in '03). Average defenses two years back are brilliant over plays (81%) if they were average again (Cowboys, Jets), but horrible (25%) if they improved to being good (Browns, Dolphins, Giants, Raiders).

As can be seen from the quick romp through above, it's easy to find conflicting evidence. Nevertheless, we'll be incorporating the prior year O/D ranks trends into our full season win betting selections for the coming campaign.