Looking Back to Move Forward: Reviewing the 2015 and Anticipating the 2016 NFL Seasons [CORRECTED]
In 2013, our publicly disseminated NFL betting picks fell just shy of the mythical 60% win threshold (recording a 59% win rate). We followed up that impressive performance with a lackluster 26-28 season in 2014. Last year though, we returned to profitability with a 55.6% win rate.
Specifically, our model, which has evolved to lean primarily on a pair of independent power ranking frameworks, managed a 65-52-4 record versus the number and went 25-20-1 picking OVER/UNDERs, demonstrating consistent proficiency across ATS as well as totals bets.
Against the Spread Picks
A look back at weekly performance dissects our (against-the-spread) betting season into distinct cohorts: we experienced four great weeks (where our work netted better than two units), three good weeks, during which we realized greater than one unit of profit each, three bad weeks--each resulting in losses greater than two units (including week 7, where we did not win a single game out of seven plays!), and seven reasonably benign weeks that saw our cumulative net profit shift by less than one unit.
So while significant profits were realized in six instances of 17, meaningful losses were suffered in only three cases. Further, overall, net gains accrued in ten of the 17 weeks of the NFL regular season, while losses were recorded in only seven instances. If the observation that upside is more likely than downside holds, we expect to continue to generate profits going forward.
A review of our weekly ATS performance also reveals that the concept of reversion to the mean (at least, following extreme performances) is evident. Notice that the abysmal week 7 outing occurred on the heels of our best week of the season. Moreover, our second-worst week of the year, week 14, forebode an especially strong showing during the next out.
Moreover, we note our model's successes early and late in the season. The week 1 profitability hints that exploitable inefficiencies exist at the start of a new season. Researchers such as Osborne (2001) have documented such phenomenon. Similarly, the cumulative profit amassed by our work versus the number during the final three weeks of the season is almost double the performance of the next best three-week period. This finding might support the contention of Borghesi (2007), that bias increases as the regular NFL season winds down.
In the totals arena, we suffered only two bad weeks (defined as periods during which we suffered losses of more than one unit): in week 13 we were 2-5 picking OVER/UNDERs, generating just over three units of loss for our trouble. Similarly, week 17's 1-2 outing netted a hair over two units of losses (there were also five weeks in which we recorded nominal losses).
As an offset to these struggles, we experienced four weeks where we earned in the ballpark of one unit, and three weeks of net profits in excess of one unit.
In aggregate, though we experienced our fair share of volatility, our review of the 2015 NFL betting season left us encouraged and excited for the start of the 2016 adventure.