Market vs Reality [Football Season]

← All reports
2026 Season · Week 10

Week 10: Favorites Held Firm—Mostly

Week 10 brought 14 completed games, with favorites finishing 10-4. The market’s average Brier score was 0.1835, better than the 0.2500 baseline. There were no ties, so all 14 games counted in the binary results.

The headline result: the market got most favorites right, though a few road upsets kept the week from being spotless.

Market Accuracy by Week

FOOTBALLBLOGGER / NFL MARKET ACCURACY
===============================================
Average Brier score (LOWER IS BETTER)
Each # represents approximately 0.01 Brier points

WEEK  BRIER   GAMES  CHART
-----------------------------------------------
   3  0.2517     16  #########################
   4  0.2172     16  ######################
   5  0.2236     14  ######################
   6  0.2612     14  ##########################
   7  0.2272     14  #######################
   8  0.1788     14  ##################
   9  0.1166     15  ############
  10  0.1835     14  ################## <-- CURRENT
LATEST VS PREVIOUS: WORSENED (+0.0669)
-----------------------------------------------
50/50 BASELINE: 0.2500
===============================================

Best Call

The market’s strongest prediction was Jacksonville at Tennessee. JAX was given an 83.5% chance to win and delivered, beating TEN 40-23.

That was a convincing result and the week’s cleanest call.

Biggest Miss

The largest miss came in Los Angeles’ visit to Baltimore. The market gave BAL a 67.5% chance, but LAC won 23-20.

A surprising result does not necessarily mean the original probability was unreasonable. A 67.5% forecast still leaves plenty of room for the other team to win; it simply failed to happen this time.

Quarter Kelly

The original Quarter Kelly strategy stayed on the sidelines again. It began the week with $153.44 and ended with $153.44.

All 14 games were examined. None cleared the strategy’s five-percentage-point entry threshold, and no games were excluded because of missing 24-hour snapshots. The entry threshold may be overly cautious in some weeks, but more bets would not automatically be better.

This was a WALK_FORWARD_HISTORICAL test week, using only earlier completed weeks and 24-hour snapshots. Weeks 3 and 4 were training only, and Week 5 was the first test week. The model uses historical calibration adjustments, quarter Kelly sizing, and a 10% bankroll cap. These results do not establish profitability or predictive skill.

Since the experiment began, the original Quarter Kelly strategy has moved from $160.00 to $153.44:

Quarter Kelly 2.5% (Retrospective Comparison)

The 2.5% version is a RETROSPECTIVE historical comparison introduced later, not a prospectively tracked strategy. It uses the same walk-forward model and quarter Kelly sizing, but a lower entry threshold and a separate bankroll.

It placed six wagers and went 4-2:

The wins included KC, IND, SF, and CIN. The losses were HOU and BAL. Since the experiment began, this separate comparison stands at:

That lower threshold should not be called better based on hindsight.

Bet the House

Bet the House divides the carried-forward bankroll equally among eligible favorites, using the final permanent snapshot midpoint as the assumed price. The experiment began with $160.00 bankroll in Week 3, divided equally among eligible favorites.

This week, it wagered on all 14 favorites and went 10-4:

The losses included HOU, GB, BAL, and NYG. Wins such as DET, CAR, and BUF helped produce the strong weekly result.

Overall, Bet the House has grown from $160.00 to $176.65:

What We Learned This Week

All wagers are fictional, midpoint prices may not be executable, and this experiment is not betting advice.

← Week 9 All Reports