Week 1 underdogs went 218-196 ATS — 52.7%, 1999–2025, n=414, asOf 2026-08-17; 10 pushes are excluded from the rate across 424 games. That is worth opening the file for. It is not worth betting before opening the current market.
Every record, percentage, sample size, and season example below comes from data/training-snapshots/nfl/games-*.csv asOf 2026-08-17 under the filters documented at the end. The snapshot contains a stored closing spread and final score, not a complete side-price history.
The full record, without the slogan
AsOf 2026-08-17, the snapshot contains 424 non-pick’em Week 1 regular-season games. The ATS record is 218-196 — 52.7%, 1999–2025, n=414; 10 pushes are excluded from the rate. The same rows contain 144 outright wins, a separate outcome that does not grade the stated ATS question.
Underdogs in every other regular-season week went 3,232-3,102 ATS — 51.0%, 1999–2025 excluding Week 1, n=6,334, asOf 2026-08-17; 179 pushes are excluded. Week 1’s 52.7% rate was 1.6 percentage points higher in this snapshot. That is a modest historical gap, not a separate law of football.
The recent slice went 44-36 ATS — 55.0%, Week 1 of 2021–2025, n=80, asOf 2026-08-17, with no pushes. It is a persistence check and a much smaller sample. It was examined after the broad angle was visible, so it is descriptive rather than independent confirmation.
Road dogs carried the aggregate
| Location | ATS W-L | ATS rate | n | Pushes | Outright wins |
|---|---|---|---|---|---|
| Road underdog | 143-124 | 53.6% | 267 | 8 | 94 |
| Home underdog | 75-72 | 51.0% | 147 | 2 | 50 |
| All underdogs | 218-196 | 52.7% | 414 | 10 | 144 |
Road underdogs went 143-124 ATS — 53.6%, 1999–2025, n=267, asOf 2026-08-17; eight pushes are excluded from 275 games. Home underdogs went 75-72 ATS — 51.0%, 1999–2025, n=147, asOf 2026-08-17; two pushes are excluded from 149 games. The 2.6-percentage-point gap describes this snapshot. It does not isolate travel, crowd, coaching, or schedule quality.
This split is the first brake on the headline. “Week 1 dogs cover” sounds universal. The home group sat near even. Remove the road side and the broad slogan loses most of its shine.
The large-dog bucket did the work
| Points received | ATS W-L | ATS rate | n | Pushes | Outright wins |
|---|---|---|---|---|---|
| +1 to +3 | 86-82 | 51.2% | 168 | 8 | 77 |
| +3.5 to +6.5 | 85-85 | 50.0% | 170 | 2 | 49 |
| +7 or more | 47-29 | 61.8% | 76 | 0 | 18 |
Teams getting seven or more points went 47-29 ATS — 61.8%, 1999–2025, n=76, asOf 2026-08-17, with no pushes. The middle bucket went 85-85 ATS — 50.0%, 1999–2025, n=170, asOf 2026-08-17; two pushes are excluded. Small dogs went 86-82 ATS — 51.2%, 1999–2025, n=168, asOf 2026-08-17; eight pushes are excluded. Remove the +7 group and the broad angle becomes much less dramatic.
The strongest row is also the smallest: n=76 across 27 seasons, asOf 2026-08-17, or fewer than three qualifying games per season on average. The archive was divided into three declared buckets, not searched at every possible cut. A live rule still needs a locked definition and new games that were not used to discover it.
The annual path was noisy
The aggregate did not arrive in a straight line. The selected annual rows were 11-3 ATS — 78.6%, 1999, n=14; 6-10 ATS — 37.5%, 2008, n=16; 11-5 ATS — 68.8%, 2014, n=16; 12-4 ATS — 75.0%, 2021, n=16; 7-9 ATS — 43.8%, 2024, n=16; and 7-9 ATS — 43.8%, 2025, n=16, all asOf 2026-08-17. Those rows are examples from the complete snapshot, not a curated system.
Most season rows in the snapshot contribute 14 to 16 non-pick’em decisions, asOf 2026-08-17. A handful of reversed outcomes can move an annual percentage sharply. The long window is the evidence; the yearly rows show the path.
What the snapshot cannot answer
The source contains no preseason games, so it cannot test whether preseason performance predicted Week 1. It contains no opening spread, so it cannot tell us whether an underdog was available at a better or worse number earlier in the week. It contains the stored closing spread and final score. That is enough for the declared ATS record and not enough for a market-movement story.
The lazy move is to fill those gaps with confidence. The useful move is to label them. Missing opening data is not “no movement.” Missing preseason rows do not prove preseason was irrelevant. They are unanswered questions in this snapshot.
What the record does and does not say
The receipts are clear asOf 2026-08-17: all Week 1 dogs were 218-196 ATS — 52.7%, 1999–2025, n=414; road dogs were 143-124 ATS — 53.6%, 1999–2025, n=267; home dogs were 75-72 ATS — 51.0%, 1999–2025, n=147; and the +7-or-more bucket was 47-29 ATS — 61.8%, 1999–2025, n=76. Pushes are excluded from each n and listed above. Those rows do not establish that every future dog is mispriced or that any historical bucket survives current prices.
The next decision starts with the posted spread. Use the NFL picks page for a graded pick tied to a named number, Analytics to test a locked Week 1 filter, and the ATS guide for settlement basics.
How to test the angle forward
Register one rule before the next completed season: regular-season Week 1, non-pick’em underdogs, graded at a named closing source. Keep road versus home and the three declared spread buckets as diagnostics. Append the new season without editing the old record.
Retain the actual side price in the future ledger. The published output is limited to ATS wins, losses, pushes, percentage, window, sample size, and asOf. No financial claim is made because the snapshot does not supply a complete historical price series.
Future rows also need an honest no-data state. An unplayed schedule row is not a win, loss, push, or zero. A missing close is not permission to choose another number after the game.
The +7 row needs the cleanest holdout
The 47-29 ATS — 61.8%, 1999–2025, n=76, asOf 2026-08-17 large-dog row is the strongest result here and therefore the easiest one to overfit. It was discovered in the same archive used to describe it. The next test should carry the exact +7-or-more rule into new completed Week 1 games rather than splitting it until a prettier subgroup appears.
The same discipline applies to road dogs. Road status remains a diagnostic unless it is declared before the next test. Promoting every favorable split to a separate pick rule multiplies the chance of finding noise and calling it signal.
Where these numbers come from
How we counted: We concatenated the four NFL CSV shards, kept played rows with game_type = REG, seasons 1999–2025, week = 1, and a nonzero closing spread_line. The underdog is the side opposite the favorite implied by the dictionary’s spread sign. Underdog ATS margin is the negative of favorite ATS margin; positive is a cover, zero a push, negative a loss. Points received are abs(spread_line). The comparison with other weeks uses the same rules with week != 1. Cover rate excludes pushes. Source files live under data/training-snapshots/nfl/; definitions and shard hashes are in DICTIONARY.json and manifest.json. Snapshot asOf 2026-08-17.
What to watch in the next holdout
What to watch: the first newly completed Week 1 games appended with a named closing source and real side prices. Keep the +7-or-more and road-dog rules frozen; the new rows should strengthen, weaken, or reverse the historical take without changing the old sample.
NFL ATS cover-margin distribution
Bars count completed NFL schedule rows by closing-spread cover margin using the repository canonical home-margin grading convention.
