How do I build an NFL player prop model?
Build one row per line with its opening line, the line at your decision time and your own features, and label it with the final stat from a box score you trust. NFL lines post early and move for days, so fix one decision time and use the line as it stood then, or your backtest will use information you would not have had.
| NFL market | Moves | Median move |
|---|---|---|
| Receiving Yards | 16,059 | 1.0 |
| Rush Yards | 6,603 | 2.0 |
| Pass Yards | 6,158 | 3.0 |
| Longest Reception | 2,398 | 1.0 |
| Rush+Rec Yds | 1,845 | 1.0 |
| Pass+Rush Yds | 1,493 | 3.0 |
| Receptions | 1,295 | 1.0 |
| Fantasy Score | 857 | 0.5 |
| Time before kickoff | Share of moves |
|---|---|
| More than 24 hours before | 44% |
| 6 to 24 hours before | 18% |
| 1 to 6 hours before | 12% |
| Under 1 hour before | 5% |
| After the start | 21% |
Check meta.leagues_graded for NFL outcome availability and meta.stats_leagues plus meta.coverage for loaded game logs; coverage can differ between those datasets. The /api/v1/training-set endpoint returns labeled rows where grading is available, while /api/v1/game-logs supplies loaded prior-game statistics. Do not turn an empty or ungraded response into a target.
For an NFL prop backtest, reconstruct the line at your fixed decision time from /api/v1/board?as_of or a projection’s history, then join only earlier game logs and time-bounded team or opponent statistics. Split by week or game date, keep all props from one game together, and evaluate against later games. A final closing line must not become a feature for a decision made days earlier.
A line as it moved
curl "https://prizepicksapi.com/api/v1/projections/$PROJECTION_ID/history" \
-H "Authorization: Bearer $PRIZEPICKS_API_KEY"Go further
Related questions
Independent data service, not affiliated with PrizePicks. Information only, not betting advice.