Fireside chat: halftime adjustments

Tl;dr: We will be applying three quantitative adjustments starting Week 6:

  1. The model’s regression coefficient will be reduced from 1.0 to 0.8.
  2. The model’s home field advantage will be reduced from 2.0 points to 1.75 points.
  3. The model will feature a “hook adjustment” around spreads of 3 and 7.

As a result of these adjustments the model’s accuracy increased by an average of 0.36 points per wager.  However, on the top quartile of wagers, the model’s accuracy increased by an average of 1.24 points per wager.


The model has thrown a disproportionate number of favorites in its first 5 weeks: of the 18 wagers over 3.0 points strength, only 3 of them were underdogs.  These figures were more balanced across the entire model, but our betting focus is on the model’s top end.

Regressing the model’s predicted spreads against the actual spreads yielded a pleasantly supplying coefficient of determination in excess of 90%.  This was a highly encouraging result, but there were two key takeaways from the analysis.

First, the slope of the regression was less than one, indicating that actual spreads have been “tighter” than the model’s predictions over the past few weeks.  Second, the y-intercept indicated we should re-evaluate our estimate of home field advantage (HFA). 

Applying the regression coefficient and iterating home field advantage results in a “tightened up” model that I expect will provide more balanced overall results.  Its greater accuracy will also reduce the “strength” of our bets across the board.

The formula for our predicted spread, starting in Week 6 going forward, will be .8[(SRS_away_predicted)-(SRS_home_predicted)] – 1.75.  This will be reevaluated every few weeks as more data comes in, and of course at the outset of 2021, when the reduced HFA should be expected to elevate back to normalized levels as fans (please dear God) return to stadiums.

The reduction in HFA from 2.0 to 1.75 demonstrates just how radically this season has been altered.  HFA adjustments typically hover between 2.5 and 3.0.  Taking a full point away from home teams is a significant modification, and it seems sensible that 70,000 person crowds would be a worth a full point.  The Vikings front office has described the atmosphere as “flat” and found breaks to be “eerily silent.”

The model will also be incorporating a “hook adjustment” around spreads of 3 and 7.  The 3-point hook will be penalized more heavily than the 7-point hook, mostly due to the vastly greater historical importance of 3 compared to 7, but also due to the conversion rule changes made several years ago.

Which wagers to take has always been, and always will be, somewhat of a judgement call.  Under the adjusted model (starting this week), I recommend taking wagers of 2.0 points strength or greater, and scaling wagers exponentially based on their strength.  For example, a $50 baseline on a 2.0 strength wager would escalate to $100 on a 3.0 strength wager and $200 on a 4.0 strength wager. 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Powered by WordPress.com.

Up ↑

%d bloggers like this: