EdgeLine research · Published October 20, 2026
NFL Week 6 model review: what 8 verified picks showed
Week 6 was a losing week in the receipt-backed record. Eight completed Play or Strong selections had matching original pre-kickoff receipts and finished 3–5. Reporting that result plainly is more useful than highlighting only the strongest segment.
Week 6 results by market
| Market | Completed | Record | Win rate |
|---|---|---|---|
| Point spreads | 3 | 1–2 | 33.3% |
| Game totals | 5 | 2–3 | 40.0% |
| All verified Week 6 picks | 8 | 3–5 | 37.5% |
All eight qualifying Week 6 selections in this review were preserved under model version 4.5.1. Non-official leans, missing receipts and reconstructed after-the-game values are excluded.
What the evidence supports
The Week 6 sample shows that both market types missed more often than they won. The totals segment supplied two of the three wins, but it also supplied three losses. The spread segment finished 1–2. With only eight decisions, neither split is large enough to establish a stable rate.
Across the full receipt-backed record available at this snapshot, 17 completed decisions stand at 8–9. Spreads are 5–3 and totals are 3–6. That broader spread record is stronger than the Week 6 spread result, which is exactly why one favorable aggregate should not be used to hide a weaker individual week.
What this review does not prove
This is a grading review, not a claim that the model beats the market. A win-loss record does not measure forecast calibration, and an assumed −110 return is not the same as a return calculated from executable prices saved for every selection.
No decided pick in the current 17-pick receipt-backed sample has a probability preserved on its original receipt. Therefore this review cannot report a valid Brier score, log loss or probability-band calibration. Those probabilities are not reconstructed after the games.
Closing-line data is also incomplete across the broader record: 10 of 17 verified picks have usable saved closing lines. Missing values are not counted as zero, and this page does not infer a Week 6 closing-line result from the partial season sample.
Keeping comparisons fair
A model-versus-market comparison must use the same completed games and information available at the same forecast cutoff. Week 6 is not compared here with a competitor forecast or a later sportsbook line because a complete paired benchmark with matching pregame timestamps is not available for all eight selections.
The separate October 18 model comparison remains a frozen 14-game study covering 13 Week 5 games and one Week 6 game. Its figures are not reposted here as new Week 6 evidence.
What happens next
Future versions will remain separate from version 4.5.1, and new challengers will be judged forward only. No later model change will rewrite these Week 6 receipts. The next useful evidence is a larger set of completed predictions with preserved probabilities, consistent pregame market prices and enough closing-line coverage to evaluate outcomes and calibration together.
Audit the record
- Receipt-backed weekly results and evidence coverage
- Model benchmarks, versions and readiness checks
- Full completed official history
- Current methodology and model limitations
Snapshot source: EdgeLine forward-validation report updated October 19, 2026. Historical results do not guarantee future outcomes. EdgeLine provides independent analysis for informational and entertainment purposes. No wager is guaranteed. 21+ where legal; gamble responsibly.