Week 7: Favorites Hold Firm—Mostly
Week 7 brought 14 completed games, with favorites finishing 10-4. The market’s average Brier score was 0.2272, better than the 0.2500 baseline. There were no ties.
The favorites still took some bruises. Tampa Bay won 43-13 at Carolina despite being the underdog, while the Giants shocked Houston 27-21. Seattle also handled Kansas City 20-3.
Best Call
Minnesota was the market’s cleanest prediction. The Vikings beat Indianapolis 16-10, backed by a 74.5% win probability.
That forecast produced a Brier score of 0.0650, the week’s best. It was a strong combination of confidence and correctness, though one successful call alone does not establish that the market is perfectly calibrated.
Biggest Miss
Houston was the biggest disappointment for the market. The Texans carried an 80.5% probability against the Giants but lost 27-21 at home.
That resulted in a Brier score of 0.6480, by far the week’s worst. It was also the clearest example of how a highly favored team can still lose. The surprising result does not necessarily mean the original probability was unreasonable; an 80.5% chance still leaves room for the other outcome.
Other misses included Carolina losing 43-13 to Tampa Bay, San Francisco falling 29-20 at Atlanta, and the Jets dropping a 28-25 decision to Miami.
Quarter Kelly
The fictional Quarter Kelly strategy stayed on the sidelines this week.
It began with a bankroll of $153.44 and ended at $153.44, producing a weekly profit/loss of $0.00 and a return of 0.00%.
- Wagers: 0
- Record: 0-0
- Amount wagered: $0.00
This was a walk-forward historical week, not a training-only week. Weeks 3 and 4 were training-only, and Week 5 was the first test week. The strategy uses only earlier completed weeks and 24-hour snapshots, along with historical calibration adjustments, quarter Kelly sizing, and a 10% bankroll cap.
There were no missing 24-hour snapshots among the 14 games. Instead, all 14 fell below the required five-percentage-point edge threshold, so none qualified for a wager. The strategy’s silence was a rules-based result—not a prediction that every game was unplayable.
Bet the House
The fictional Bet the House benchmark divides its entire carried-forward bankroll equally among eligible favorites, using each game’s final permanent snapshot midpoint as the assumed price. The experiment began with $10 per game in Week 3.
This week, it started with $89.86, placed 14 wagers, and wagered $89.86 in total.
- Record: 10-4
- Weekly profit/loss: $13.25
- Ending bankroll: $103.11
- Return: 14.75%
The benchmark benefited from correctly backing Minnesota, Dallas, Denver, Detroit, and Seattle. But four favorite losses—including Houston at 80.5% and the Jets at 64.5%—kept the result from being even stronger.
What We Learned This Week
- The market beat the 0.2500 baseline overall, but one large miss in Houston significantly weighed on the results.
- The favorite record and the Brier score tell a generally positive story, while the four upsets show that confidence is never certainty.
- Quarter Kelly made no bets because none met its edge requirement, while the simpler favorite benchmark had a profitable week. That contrast is interesting, but one week cannot establish predictive skill or future profitability.
All results here are fictional. Midpoint prices may not be executable, fees are excluded, and no real money was wagered. This is not betting advice.