| --- |
| license: cc-by-4.0 |
| language: |
| - en |
| pretty_name: Line Movement Dataset |
| size_categories: |
| - n<1K |
| task_categories: |
| - tabular-regression |
| - tabular-classification |
| tags: |
| - sports-betting |
| - line-movement |
| - closing-line-value |
| - price-discovery |
| - gambling-economics |
| - tabular |
| configs: |
| - config_name: default |
| data_files: |
| - betbetter_line_movement_h2h.csv |
| - betbetter_line_movement_points.csv |
| - betbetter_distance_to_close.csv |
| --- |
| |
| # Line Movement Dataset — how much betting lines actually move |
|
|
| Aggregate measurements of pre-match betting line movement across **13+ sports and three market |
| types**, built from **11.7 million odds snapshots** recorded between 29 June and 3 August 2026 |
| across 52 online bookmakers. To our knowledge the first open, multi-sport measurement of line |
| movement magnitude — the canonical public evidence on closing-line value rests on a single |
| operator's blog study. |
|
|
| | | | |
| |---|---| |
| | **Source observations** | 11,740,751 pre-match odds snapshots, ~hourly cadence | |
| | **Window** | 29 June – 3 August 2026 (36 days) | |
| | **Books / competitions** | 52 bookmakers, 22 competitions | |
| | **Files** | 3 aggregate CSVs (sport × market × time-bucket grain) | |
| | **Licence** | CC BY 4.0 | |
|
|
| ## Files |
|
|
| ### `betbetter_line_movement_h2h.csv` — open-to-close probability movement |
| One row per sport. Head-to-head (two-outcome) markets: the outcome's **de-vigged** implied |
| probability at the first recorded snapshot vs the last pre-match snapshot, for markets first seen |
| at least 6 hours before start. Columns: `observations`, `mean_lead_hours`, `mean_abs_move_pp`, |
| `pct_moved_gt2pp`, `favourite_signed_drift_pp` (positive = favourites shorten toward the close). |
|
|
| ### `betbetter_line_movement_points.csv` — spread & total line movement |
| One row per sport × market type. The quoted line itself (points/handicap), first vs last |
| pre-match snapshot: `mean_abs_point_move`, `pct_line_changed`, `max_point_move`. |
|
|
| ### `betbetter_distance_to_close.csv` — where price discovery happens |
| One row per sport × time-to-start bucket (`7d_plus, 3-7d, 1-3d, 6-24h, 1-6h, under_1h`): the mean |
| absolute gap, in probability points, between the de-vigged price at that time and the same |
| outcome's closing price. Cells with fewer than 100 observations are suppressed. |
| |
| ## Headline findings |
| |
| - **Movement is sport-specific, by an order of magnitude.** Mean open-to-close move: UFC/MMA |
| 2.76 pts (53% of outcomes move >2 pts), AFL 2.52, WNBA 1.81 — against MLB 0.60 and ATP-tour |
| tennis ~0.5–0.8. |
| - **Price discovery happens 1–7 days out, not at the death.** In-season markets sit ~2.2 pts from |
| their close 3–7 days before start, ~1.4 pts at 1–3 days, 0.6 pts at 6–24h and only 0.33 pts in |
| the final 1–6 hours. |
| - **Favourites shorten.** The opening favourite's probability drifts *up* toward the close in |
| almost every sport (MMA +0.8 pts, AFL +0.5, WNBA +0.44) — consistent with openers overpricing |
| longshots and the close correcting it. |
| - **Spread/total lines move where scoring is high:** AFL handicaps move 2.55 pts on average (76% |
| of lines change; max observed 14), WNBA totals 1.69 pts (77% change) — while MLB run lines and |
| NHL puck lines are essentially static. |
| - **Pre-season lines barely move:** NFL and NHL markets posted months ahead move ~0.3–0.4 pts — |
| lines move when information flows, not with the passage of time. |
| |
| ## Construction and caveats |
| |
| - Probabilities are **de-vigged** (margin removed across the market's outcomes), so movement is |
| price relocation, not margin change. Points (spreads/totals) are the quoted lines themselves. |
| - "Open" = first snapshot in our archive, bounded by the 36-day window — for far-posted markets |
| (NFL) the true open predates the window, so movement figures are floors for those sports. |
| In-season sports (MLB, AFL, WNBA, MMA, tennis, MLS) are fully captured. |
| - Aggregates only: no odds, prices, bookmaker identities, teams, players or event identifiers |
| appear in the data, and no underlying odds feed can be reconstructed from it. |
| - Snapshot cadence is approximately hourly and varies; `under_1h` cells are thin. |
|
|
| ## Licence and citation |
|
|
| CC BY 4.0 — free to use, including commercially, with attribution. |
|
|
| > Bet Better (2026). *Line Movement Dataset: how much betting lines actually move, measured |
| > across 11.7M odds snapshots.* https://betbetter.world/studies/line-movement |
|
|
| Maintained by [Bet Better](https://betbetter.world), which also publishes the |
| [Bookmaker Margin Panel](https://doi.org/10.5281/zenodo.21755380), the |
| [Market Calibration Dataset](https://doi.org/10.5281/zenodo.21764674), a |
| [free sports model API](https://betbetter.world/api/) and an |
| [open AFL dataset](https://doi.org/10.5281/zenodo.21612783). |
|
|
| > 18+. Most people lose money gambling. If gambling is causing you harm: gamblinghelponline.org.au. |
|
|
