NFL Week 1 Opening Lines 2026

A free, downloadable dataset: the opening line and the current line, per sportsbook, for every NFL Week 1 game of the 2026 season, across moneyline, spread and total. One row per game, book, market and side. Built from our own archive, published as CSV and JSON, free to use with attribution.

As of 2026-09-01 14:37 UTC · 1701 rows · 16 games · 32 book feeds

What "opening line" means here

An opening line in this dataset is the earliest price we have stored for that combination of game, sportsbook, market and side. It is not a claim about the first number the book ever hung on the board. Our Week 1 series begins 2026-05-12 16:25 UTC, so the deepest rows here reach back to look-ahead boards posted in May, but most books started filing later than that and a large share of the file was first seen on the as-of day itself. Read opening_seen_utc as "first price we hold", not "market open", and check it before you describe any row as an opener: the column is on every row precisely so you do not have to take our word for the depth.

The current line is the most recent price in the same stored series for that same game, book, market and side, as of 2026-09-01 14:37 UTC, and current_seen_utc on every row says when that price was actually filed. A handful of rows carry a current price older than the as-of stamp, because that book stopped quoting; the column tells you which. Opening and current come from one store with one set of book keys, and period markets and alternate-line ladders are stripped before a series is formed, so a row compares a full-game main line to a full-game main line from the same book. The live API is fresher than this frozen extract; this file is a snapshot with a timestamp on it precisely so that anyone who cites it can still check the number later.

One thing worth saying plainly, because it is the reason this is a file and not a live page. The archive it came out of is pruned: it keeps a long thin history and discards most of the dense recent detail as it ages. Re-running the same extract today would read a small fraction of the rows this one read, and every number here would quietly change. So this file is not a view onto a database, it is the record. The as-of stamp is the contract, the file will not be edited in place, and if it is ever rebuilt it goes out under a new date rather than overwriting this one.

1113 of the 1701 rows hold exactly one stored snapshot. For those, opening and current are the same price and notes says single_snapshot_opening_equals_current. They are included because a book quoting a game is a fact worth having, but they carry no movement, and the moves table below ignores them. The 588 rows with two or more snapshots are the ones with real opener-to-now history. Nothing anywhere in this file is interpolated, smoothed or back-filled: a gap is published as a gap.

Prices are American. opening_implied_prob_raw and current_implied_prob_raw are converted straight from the American price with the vig still in them; they are not devigged and should not be read as fair probabilities. Spread rows carry the point from that side's perspective, so a home row of -3.5 means the home team is laying 3.5.

Biggest opener-to-now moves

Consensus move per game and market: the median across every book with two or more stored snapshots, so one book rebalancing its own exposure cannot shape the ranking. A game and market needs at least three such books to appear at all. Spread is measured on the home side and total on the over, so a negative spread move means the home number came down (the home team got stronger) and a positive total move means the total went up.

Per-book rows in the CSV are printed exactly as the archive holds them. That includes cases where two books disagree about which side is laying the points, which is either a genuine look-ahead disagreement or a side swap somewhere upstream, and this file does not pretend to know which. It is the reason the headline uses a median across books rather than any single book's number, and it is why the snapshots and history_days columns are there: check them before you quote a single row.

Columns

• game_id, kickoff_utc, sport_key, away_team, home_team: the game.
• book: our bookmaker key, the same key GET /v1/bookmakers serves. The unibet_* keys are separate regional feeds of one brand, not separate brands.
• market: moneyline, spread or total. side: home, away, over or under.
• opening_point, opening_price_american, opening_implied_prob_raw, opening_seen_utc.
• current_point, current_price_american, current_implied_prob_raw, current_seen_utc.
• point_move: current point minus opening point, blank for moneyline. implied_prob_move_pct_points: the change in raw implied probability, in probability points.
• snapshots: how many stored prices back this row, each one from a distinct timestamp. history_days: days between the opening and the current snapshot, rounded to a tenth, so a row refreshed twice inside a couple of hours reads 0 while still holding two real timestamps. The two _seen_utc columns are the authority on depth.
• notes: honest flags, space separated. Exactly two values occur. single_snapshot_opening_equals_current means we hold one price for this row, so it carries no movement. price_missing_on_one_end means the archive holds the line but not the price at one end, so the line move is real and the implied-probability move is left blank rather than filled in.

What we left out, and why

Readings we could not place on a side, a period or a line with certainty were dropped rather than guessed at, because guessing any of those three changes the number:

• label_not_resolvable_to_a_side: 72 source rows
• period_market_under_full_game_key: 43 source rows
• spread_composite_label_side_ambiguous: 81 source rows
• alternate_line_ladder_main_line_not_identifiable: 798 readings

Two of those need a word. Some books file first-half and quarter markets under the full-game market key, distinguishable only by a suffix on the label, and a 1H total of 22.5 published as a full-game opener would be nonsense; those are removed. And some books do not quote a single current line, they post the whole alternate-line ladder in one instant, a dozen rungs from one side of the number to the other, each with its own price. A ladder is a menu, not a history, and which rung is the main line is not recoverable from the row, so the whole instant is dropped. That is why a book you would expect to see on a given game and market is sometimes absent from it: we would rather be missing a book than invent its opener.

License and attribution

Free to use, including commercially: republish it, chart it, load it into a model, ship it in a newsletter. The one condition is attribution with a working link back. Copy either line:

Data: ParlayAPI NFL Week 1 opening lines dataset, https://parlay-api.com/nfl/week-1-opening-lines (as of 2026-09-01 14:37 UTC)

Data: <a href="https://parlay-api.com/nfl/week-1-opening-lines">ParlayAPI NFL Week 1 opening lines</a> (as of 2026-09-01 14:37 UTC)

If you build something with it we would genuinely like to see it. No permission request needed, no sign-up needed to download, and no tracking pixel in the files.

curl -s 'https://parlay-api.com/v1/sports/americanfootball_nfl/odds?regions=us&markets=h2h,spreads,totals' \
  -H "X-API-Key: $PARLAY_API_KEY"

Related

NFL Week 1 odds board · Biggest line moves (free feed) · Historical odds coverage · NFL odds API · All free datasets · No-vig calculator