Week 1 recap

These picks were rebuilt after the games, from the model and the betting lines saved before kickoff. They were not posted in advance, so treat this week as a replay, not a track record.

6–4Must plays, +1.9u
4–6Leans, -1.8u

Must plays

Leans

Flat 1 unit per pick. Graded exactly as posted at the last update before each game's kickoff. Players who did not play are voided.

Method changes after this week

Changing

  • Lines now matched only to the game they were posted for (stale-line bug fixed)
  • Picks split into must plays and leans; both tracked
  • Every posted pick is saved with a timestamp and graded as posted
  • Line logging cut to Wednesdays and Saturdays to fit the data budget
  • Edges above 15% held back as likely model errors
  • Week 1 usage review added as a diagnostic, not a model input

Not changing

  • Projection model: unchanged, one week is too small a sample
  • Bet sizing: flat 1 unit, edge size has not predicted results
  • No manual adjustments for Week 1 usage spikes (being tracked instead)

Updated Sep 16, 2026 03:34 UTC

Research project. The model has no demonstrated betting edge yet; results so far are too few to separate skill from luck.