The model finished Week 1 at 12-4 straight up—a solid 75% accuracy—but it came with a caveat: the market went 12-4 as well, meaning this was a universally difficult week rather than a solo stumble.
All four misses landed on upsets, and the model never diverged from the market's consensus on any game. When the entire week tilts toward unpredictability, staying in lockstep with published spreads is neither a strength nor a liability; it's simply the prudent default. The real story is how far those consensus calls missed.
The highest-confidence mistake came in ARI @ LAC, where the model backed LAC at 76% conviction on a 9.5-point spread. ARI won outright, 14-26, creating a 21.5-point swing from expectation to reality. Both teams entered with defensive injuries—LAC missing CB Deane Leonard (abdomen), ARI missing CB Garrett Williams (Achilles), G Isaiah Adams (knee), and S Dadrion Taylor-Demerson (ribs)—but the model and market both saw LAC as the stronger side.
The largest margin miss followed in SF @ LAR. The model projected a 3.4-point LAR win on a 3.5-point spread; SF won 7-27, a 23.4-point error. This one stung worse in absolute terms, though the model's confidence (60%) was already tempered.
Two other upsets—NYG over DAL (by 8 on a 3-point spread) and NYJ over TEN (by 13 on a 1-point spread)—rounded out the difficult day. NYG played under new coach John Harbaugh, their first season change, while TEN entered missing LB Cedric Gray to concussion.
Average margin error across all games sat at 11.3 points. The week's core lesson: when the entire market reads the field the same way and reality diverges sharply, it's a market-wide miss, not a model failure. A 75% week against a similarly positioned opponent is exactly what competent prediction looks like.