Week 8 delivered a solid 10-for-14 straight-up performance (71.4%), a clean 8-percentage-point jump above the model's season-to-date average of 63.2%. That's the kind of week that restores confidence—not perfection, but a clear signal the underlying forecast work is tracking well.
The market matched the model exactly: 10-14 correct, identical accuracy, and critically, zero divergence between them. Every game the model picked, the market picked the same way. That alignment, combined with both getting the same four games wrong, suggests Week 8 was genuinely difficult to predict for everyone, not a week where the market stayed sharp while the model alone slipped.
The single worst miscall came in ARI @ DAL. The model backed DAL at 67% confidence and projected a 6-point margin, but ARI won 30-33. That 9-point gap between expectation and reality represents the week's most overconfident pick—and notably, the market agreed entirely, taking DAL by 6 and losing the same game. Both sides underestimated ARI's ability to win in Dallas.
The biggest projection-versus-reality blowout was JAX vs. IND, a game the model got right. It projected JAX to win by 7.5; JAX won 42-7, a 35-point margin. The 27.5-point error shows the model nailed the winner but severely underestimated the dominance—one of those games where the correct pick hides a fundamental misjudgment of the gap between teams.
Average margin error across all Week 8 games: 11 points, a reasonable spread around a week that contained four outright upsets. The model's 10 correct picks, paired with market parity and no exploitable disagreements, points to solid forecasting in a legitimately volatile week. Nothing to apologize for here.