Sports Betting Odds & Player Props Scraper

Sportsbook lines for NFL, college football, NBA, WNBA, college basketball, MLB, NHL, 219 soccer competitions, UFC, golf, tennis and F1 — the point spread, the moneyline and the total from every book the source lists, plus the things most odds scrapers leave on the table: how far each line has moved since the book opened it, the full player-prop board (about 659 markets on one NFL game), the season futures board with implied probability per runner, and the source's own win projection so you can put the model next to the price.

No login, no API key, no browser. Public endpoints only.

---

What you get

odds — the line, per book, with its movement

One row per fixture per sportsbook.

Field	Example
eventName	Detroit Lions at Buffalo Bills
providerName	Draft Kings
details / spread / overUnder	BUF -5.5 / -5.5 / 54.5
homeMoneyline / awayMoneyline	-245 / +200
homeImpliedProbability / awayImpliedProbability	0.7101 / 0.3333
openSpread → currentSpread (spreadMovement)	+3 → -5.5 (-8.5)
openTotal → currentTotal (totalMovement)	52.5 → 54.5 (+2.0)
favoriteFlippedSinceOpen	true

spreadMovement, totalMovement, favoriteFlippedSinceOpen and the implied probabilities are derived here — upstream publishes the opening and current numbers and leaves the arithmetic to you. The vig is left in, so the two sides of a moneyline sum to a little over 1; that excess is the book's margin, and removing it would be inventing data.

props — the player board

About 659 prop markets on a single NFL game: passing yards, receptions, anytime touchdown scorer, first touchdown scorer, and 70-plus other market types, each with the line it opened at, the line it sits at now and the difference.

futures — season-long markets

MVP, Offensive/Defensive Player of the Year, conference and division winners, championship. Every runner carries its American price and its implied probability, so +400 and -245 sit in the same sortable column.

predictor — model against book

The source's own win projection and predicted margin per fixture. Put homeWinProbability next to homeImpliedProbability × 100, or homePredictedPointDiff next to spread, and the disagreements fall out.

events, providers, reference

The schedule on its own (no odds, cheap), the sportsbook catalogue for a league (71 entries for the NFL), and a probe that checks the source is live and re-tests the assumptions this Actor is built on.

---

Input

Field	Default	Notes
mode	odds	odds, props, futures, events, predictor, providers, reference
league	nfl	A short key, or the source's own sport/league form for anything else
dates	—	20260920, 2026, or a range 20260901-20261001. Empty = the current week only
dateFrom / dateTo	—	Used when dates is empty
eventId / eventIds	—	Scrape specific fixtures instead of a window
season	current year	futures mode
includeUnpricedEvents	true	Keep fixtures no book has priced yet, as hasOdds: false
maxItems	500	Record cap
maxEvents	60	Fixture cap — a whole NFL season window is 320 fixtures
maxConcurrency	4
maxNameLookups	400	Budget for turning ids into names
proxyConfiguration	off	Not needed for ordinary runs

League keys: nfl, college-football, nba, wnba, ncaab, ncaaw, mlb, nhl, epl, laliga, seriea, bundesliga, ligue1, ucl, mls, ufc, pga, atp, wta, f1 — or anything else as sport/league, e.g. soccer/ned.1. There are 17 sports and 219 soccer competitions.

Examples

This week's NFL lines with movement:

{ "mode": "odds", "league": "nfl" }

Every prop on one game:

{ "mode": "props", "league": "nfl", "eventId": "401872932", "maxItems": 700 }

A month of Premier League fixtures:

{ "mode": "events", "league": "epl", "dates": "20260915-20261015" }

The NFL futures board:

{ "mode": "futures", "league": "nfl", "season": 2026, "maxNameLookups": 400 }

---

Known limits — read these before you build on it

• A fixture with no odds is normal. Books post lines close to start time. Fixtures weeks out come back with hasOdds: false and a note; that is the source answering, not a block. Failures are separate ERROR records.
• Coverage follows the season, not the league. During testing NFL, college football, MLB, the Premier League, MLS and UFC carried lines; NBA, NHL, college basketball and tennis did not, because they were out of season.
• Line movement needs both ends. spreadMovement is null when the book has not published an opening number for that market.
• Names are the expensive part. Ids come free in the response; names cost one request each, cached. Past maxNameLookups a record keeps its ids and loses only the name — it is never dropped.
• Prop counts are de-duplicated. The feed publishes repeated rows (659 rows, 521 distinct markets on one game). This Actor emits the distinct ones and logs how many repeats it dropped.
• No historical line archive. The source exposes open, close and current — not a tick-by-tick history. The movement-history endpoint returned count: 0 on every fixture sampled. Run this Actor on a schedule if you want a series.
• providers includes a placeholder. Entry id 0 is literally named "Not Available"; it is passed through rather than hidden, as with every other upstream field.
• Join on providerId, not providerName. The same book is written "Draft Kings" in the odds response and "DraftKings" in the futures response. Names are passed through exactly as the source writes them.
• Book count per fixture varies with kick-off. A sportsbook's live in-play feed appears as a second row near game time and is gone hours later, so one row per fixture and two rows per fixture are both normal.

Technical notes

Python, HTTP only, curl_cffi with a TLS ladder that remembers the winning profile. No WAF answered during testing from either a residential address or Apify's datacenter range, but the ladder stays because the profile that wins locally is not always the profile that wins from a datacenter.

CRAWLING_METHOD.md has the full probe matrix, the ten gotchas found while building (including the two hosts that disagree about date syntax and the internal hostname leaking into pointers), and the candidates probed in the same pass that were skipped with evidence.