First and Thirty has returned for the start of the 2022 - 2023 NFL season. And we fully expect it will be (yet another) season for the dogs.
Several key model parameters must be updated periodically by regressing prior predicted SRS-based spreads against their actual real world spreads. Doing this gives us a window into the value of underdogs relative to favorites and the value of home field advantage. After adjusting for "hooks" around the 3-point and 7-point margins, the model arrives at a final predicted spread.
To start this season off, the model is evaluating spread wagers at roughly 82% of their SRS differential, providing us a sizable overweight to underdogs. Paired with that is an all-time low home field advantage of about +1.5 points. Never before have we used such a small HFA figure, but the data from the past few seasons is indisputable. HFA has been steadily declining for years. I doubt we will ever see an NFL where home field makes no difference, but explanations for the decline are varied.
The underdog effect is more curious. One would expect an efficient betting market to price underdogs appropriately, but favorites have been under a multi-year drought ATS. Until the model sees data indicating otherwise, we will continue to bet with the dogs.
At the end of last year, we reduced our wager threshold significantly based on quantiative data that we weren't taking enough wagers (and Brady continuing to give me the side eye when we'd only recommend 2 bets). Previously, we were recommending wagers only if they had ~3 points of "strength," ie the model predicted the wager to be about a FG stronger than the spread.
This was actually a fairly good starting point, but last year's data proved it was too conservative an estimate. We will start this year taking wagers at a lower cutoff of 2.5626 points. This figure is now model driven (I did not pull 2.5626 out of a hat) under the theoretical principle that we will want to take all better-than-average wagers, with some allotment for variance (2.5626 represents the 95% confidence interval).
I previously speculated that this figure might run as low as 2pts, but a couple of Week 18 outliers biased my estimate. Still, the median wager strength is 1.9, and as our sample grows our confidence interval will narrow, so it is entirely possible we take slightly more bets than we are even taking now. We kickoff Week 1 with 7 wagers, just under half the week's slate. This also just *feels* right- betting around 40% of the season's games.