If you follow FaT (and if you're bothering to read this, you probably do) you know that we typically recommend two or three wagers each week. On a busy, non-bye week, you might get four bets in, but for the most part our approach has been pretty reserved.
For anyone who's not a nerd, here is all you need to know:
- Underdogos have been adjusted to be marginally less valuable, by about -2%.
- HFA has been adjusted to be less valuable, by about -10%.
- We are reducing our threshold for taking a wager from a 3-point advantage to a 1.89 point advantage. Our pickiness was leaving too much potential value on the table.
- We are eliminating wager scaling, but may reintroduce it on a discretionary basis for wagers that seriously deviate from our expectations.
And we're all about transparency here at FaT, so if you want to know why we're betting more games- a lot more games- each week, here's the technical explanation.
The wagers we take are driven by an attribute we call "wager strength." Each wager's strength is the absolute value, in points, of the residual between the model's prediction and the actual spread each week.
The above regression illustrates the first-pass output of the model, where each game's predicted spread is plotted against the spread that actually existed in the real world. All spreads are quoted in terms of the home team, so the -1.64 regression intercept is what results as home field advantage. The 0.82 coefficient is best described as an "underdog balance" variable. The larger the number, the more the model leans towards favorites. The lower the number, the more the model leans towards dogs.
I could run these all weekly, but for the sake of consistency chose to calibrate them both before and after the completion of the bye weeks. The numbers we see here are all after the last bye week of this season.
Once the regression parameters are taken into account, the model yields a final output:
And the distance between each point and the regression line is the wager strength for that particular wager.
How high a wager strength should we be willing to accept? Coming into the season I based that answer on last year's model, which predicted the actual spread with somewhat less accuracy than this year's model has. As a result, the wager strength we required was set too high, and we ended up taking far fewer wagers each week than I believe we should have.
To close out the 21-22 season, we'll be taking any wagers that exceed the average wager strength the model tends to predict. We've built in a small margin of error based on the bootstrap analysis below. Any wagers that exceed 1.89 points of strength will be accepted, with the caveat that we may revoke a wager based on countervailing news events, such as an injury or major player absence.
Bootstrap Analysis of Wager Strength
Wessa P., (2019), Bootstrap Plot for Central Tendency (v1.0.16) in Free Statistics Software (v1.2.1), Office for Research Development and Education, URL https://www.wessa.net/rwasp_bootstrapplot1.wasp/