🎯 Live Odds Scraper — In-Play & Closing Lines for NFL, NBA, MLB and 25 more leagues

Every game's betting line, including the two that other ESPN-based scrapers structurally cannot give you: the line while the game is running, and the line it closed at.

Opening, closing and current moneyline, spread and total for every game — in American, decimal and fractional odds — plus the in-play line while a game is live and the graded result (moneyline / against-the-spread / over-under) once it ends. 154 fields per row. No API key, no login, no browser, no proxy.

---

🧨 Why this one and not the other ESPN odds scrapers

Every other actor in this lane reads ESPN's public scoreboard feed. That feed has one property nobody mentions on their store page:

The scoreboard silently drops the odds the moment a game starts, and never brings them back.

Measured on the MLB slate of 2026-09-02 — 15 events:

Game state	Events	Carrying odds on the scoreboard feed	Carrying odds on the core API
Scheduled	1	1	1
In progress	5	0	5 (count = 2 each)
Final	9	0	9

This actor reads the scoreboard for context (teams, venue, weather, score, status) and ESPN's core sports API for the line. That is where the extra data lives:

• a close block appears on every game once it starts — the actual closing line, the number a betting model is graded against;
• a second provider, 200 — "DraftKings - Live Odds" — appears while the game is running, carrying the real in-play market.

Real example from that run: DET @ MIN, top of the 8th. Pre-match line MIN -120, total 8.5 (opened at 9). In-play line at the same instant: DET -1940, total 7.5. The scoreboard feed showed neither.

Reconciliation check on a 91-game college-football slate: the core API covered 91 of 91 games the scoreboard had odds for, and there were 0 games where the scoreboard had a line the core API lacked. It is a strict superset.

---

📊 What you get

• 🏈 28 leagues, all live-probed: NFL · NCAA Football · NBA · WNBA · NCAA Men's & Women's Basketball · MLB · NCAA Baseball · NHL · MLS · NWSL · Premier League · Championship · LaLiga · Serie A · Bundesliga · Ligue 1 · Eredivisie · Primeira Liga · Liga MX · Süper Lig · Argentine Liga Profesional · Brasileirão · Champions League · Europa League · Conference League · Libertadores · Leagues Cup
• 💰 Three price states per market — open, close, current — for moneyline, spread and total
• 🔢 Three odds formats on every price — American (-175), decimal (1.57), fractional (4/7)
• ⚡ The in-play line (lineType: "live") while a game is running
• 📈 Derived analytics: implied probability, no-vig probability, book hold % on both the moneyline and the total, open→current movement on all three markets, CLV (open→close in probability points), and line direction
• 🏆 Graded results once a game is final: moneylineWinnerSide, spreadResultSide (ATS), totalResult (O/U)
• 🏟️ Full game context: venue + city/state, weather, broadcaster, attendance, team records, live score, period and clock
• 🔗 Direct bet-slip deep links for home / away / over / under / game

---

⚙️ Input

Field	Type	Default	What it does
leagues	multi-select	NFL, NCAAF, NBA, MLB, NHL, NCAAM, MLS, EPL	Which leagues to pull. The enum is a hard-coded, live-probed list — ESPN's slugs are not guessable.
dateFrom	string	today (US Eastern)	YYYY-MM-DD. Past dates work and are how you build a closing-line dataset.
dateTo	string	—	YYYY-MM-DD. Blank ⇒ uses Days ahead.
daysAhead	integer	6	Days past the start date when dateTo is blank.
lineTypes	select	all	prematch, live, or both.
gameStatus	select	any	scheduled / live / final. Use final for a CLV dataset.
maxEvents	integer	50	Stop after N games, earliest kickoff first. 0 = unlimited.
maxConcurrency	integer	10	Parallel requests. 10 measured clean.
proxyConfiguration	proxy	off	Not needed — see Anti-bot below.

⏱ Dates are US-Eastern, because that is how ESPN files a game. A 10pm-ET Tuesday game is 02:00 UTC Wednesday and ESPN still returns it for the Tuesday date. Taking "today" from UTC would skip the entire evening slate — every in-progress game, which is exactly where the in-play and closing lines live.

Example input

{
  "leagues": ["nfl", "college-football", "mlb", "usa.1", "eng.1"],
  "daysAhead": 3,
  "maxEvents": 25
}

Closing-line-value dataset (settled games only, with graded results):

{
  "leagues": ["nfl", "nba", "mlb", "nhl"],
  "dateFrom": "2026-01-01",
  "dateTo": "2026-01-31",
  "gameStatus": "final",
  "maxEvents": 0
}

---

📐 Output — measured field fill

One row per game per line type. A game usually yields one row; a running game or a top-European-soccer match yields two.

All percentages below are measured on real Apify runs on 2026-09-03, not estimated.

• Run A — the default (empty input {}): SUCCEEDED in 3.1 s, 45 rows from 40 games across MLB / NCAAF / EPL / MLS. 4 in-play rows, 15 rows carrying a closing line. chargedEventCounts.odds_row = 45 = row count.
• Run B — wide (10 leagues, 2026-08-29 → 2026-09-06, maxEvents: 200): SUCCEEDED in 3.8 s, 237 rows from 183 games. Books: DraftKings 183, Bet 365 50, DraftKings-Live-Odds 4. 132 rows (56%) carried a closing line. chargedEventCounts.odds_row = 237 = row count.

A% = fill on run A · B% = fill on run B.

Game identity & context

Field	A%	B%	Notes
eventId	100	100	ESPN event id
sport	100	100	football / basketball / baseball / hockey / soccer
league · leagueSlug · leagueAbbrev	100	100	e.g. National Football League / nfl / NFL
seasonYear · seasonType	100	100	e.g. 2026 / regular-season
week	33	14	NFL & college football only
eventName · shortName	100	100	New England Patriots at Seattle Seahawks / NE @ SEA
kickoffUtc	100	100	ISO 8601
statusName · statusState · statusDetail · statusCompleted	100	100	STATUS_IN_PROGRESS / in / Top 8th
statusPeriod · statusClock	93	86	inning / quarter / period, and the clock
neutralSite · conferenceCompetition	93	50	absent on soccer
attendance	100	100
venueName · venueCity	100	100
venueState · venueIndoor	93	50	US venues
venueCountry	40	64
broadcast	100	100	e.g. NBC
weatherCondition · weatherTemperature	73	18	outdoor US games
gameUrl	78	96	ESPN game page

Teams & score

Field	A%	B%	Notes
homeTeamId · homeTeamAbbrev · homeTeamName · homeTeamDisplayName · homeTeamRecord	100	100	same five for away*
homeScore · awayScore	42	72	null until the game starts — ESPN reports "0" for an unplayed game and this actor refuses to ship that as a 0-0 score

Book & line type

Field	A%	B%	Notes
providerId · providerName · providerPriority	100	100	100 DraftKings · 200 DraftKings - Live Odds · 2000 Bet 365
lineType	100	100	prematch or live

Headline line

Field	A%	B%	Notes
details	98	79	e.g. SEA -3.5
spread · overUnder · overOdds · underOdds	98	79
initialSpread · initialOverUnder	60	36
moneylineWinner · spreadWinner	100	100	⚠️ ESPN's own booleans — false on every game we sampled, including settled finals the favourite won. Shipped verbatim because they are real API fields; use moneylineWinnerSide / spreadResultSide for the actual graded result.
homeFavorite · homeUnderdog · homeFavoriteAtOpen (+ away*)	98	79
homeMoneyline · awayMoneyline	87	76	flat American shortcut
drawMoneyline	7	49	soccer only
homeSpreadOdds · awaySpreadOdds	38	43

Moneyline — open / close / current × American / decimal / fractional

Field group	A%	B%
homeMoneylineOpen{American,Decimal,Fraction} (+ away*)	87	76
homeMoneylineClose{American,Decimal,Fraction} (+ away*)	33	56
homeMoneylineCurrent{American,Decimal,Fraction} (+ away*)	89	96
drawMoneylineCurrent{American,Decimal}	7	49
drawMoneylineCurrentFraction	2	21

Spread — the handicap line and the price on it

Field group	A%	B%
homeSpreadOpenLine · homeSpreadOpenOdds{American,Decimal,Fraction} (+ away*)	98	79
homeSpreadCloseLine · homeSpreadCloseOdds{…} (+ away*)	33	56
homeSpreadCurrentLine · homeSpreadCurrentOdds{…} (+ away*)	98	79

*SpreadOpenLine is the handicap (-3.5); *SpreadOpenOddsAmerican is the juice on it (-110). Most competitors collapse these into one column and lose the price.

Total (over / under)

Field group	A%	B%
totalOpenLine · over/underOpen{American,Decimal,Fraction}	98	79
totalCloseLine · over/underClose{…}	33	56
totalCurrentLine · over/underCurrent{…}	98	79

Bet-slip deep links

Field	A%	B%
bettingLinkHome · bettingLinkAway	89	97
bettingLinkOver · bettingLinkUnder · bettingLinkGame	98	79

Derived analytics

Field	A%	B%	Definition
impliedProbHome / Away / Draw	89 / 89 / 7	96 / 96 / 49	100 / decimal, vig included
noVigProbHome / Away / Draw	89 / 89 / 7	96 / 96 / 49	implied, normalised so the book's outcomes sum to 100%
holdPercentMoneyline	89	96	the book's moneyline vig, in points
holdPercentTotal	98	79	the book's vig on over/under
moneylineMoveOpenToCurrent	87	76	home implied-probability move, percentage points (positive = home shortened). Points, not raw American, because American odds do not subtract across the ±100 boundary
spreadMoveOpenToCurrent	98	79	homeSpreadCurrentLine − homeSpreadOpenLine
totalMoveOpenToCurrent	98	79	totalCurrentLine − totalOpenLine
clvHomeMoneylineOpenToClose	33	56	the CLV number — home implied-prob move open→close, in points
lineDirection	100	100	home / away / none

Graded result (null until the game is completed)

Field	A%	B%	Values
homeWon	20	61	true / false / null (draw)
moneylineWinnerSide	20	68	home / away / draw
spreadResultSide	20	53	home / away / push — graded against the closing line, falling back to current
totalResult	20	53	over / under / push
capturedAt	100	100	ISO timestamp of the snapshot

37 of 154 fields are 100% filled on run B; not one field was 0% filled. Every sub-100% number above has a stated structural reason.

---

💵 Pricing

Pay per event — $0.003 per odds line returned. No monthly fee, no platform-usage surcharge.

Event	odds_row — Per odds line returned
Price	$0.003 per row ($3 / 1,000)
Billed for	Each betting line delivered to your dataset
Never billed for	Starting a run · a game with no line posted (skipped before delivery) · the dataset write itself

Billing is budget-aware: rows are delivered through Actor.pushData(items, 'odds_row'), which charges atomically as it pushes and stops cleanly at your charge cap. Delivered always equals billed — verified on the runs above (45 rows → odds_row: 45; 237 rows → odds_row: 237).

The current market rate on the Apify Store for the same event is $0.012 per row. This actor is 4× cheaper.

Worked example: a daily 06:00 snapshot of the default eight leagues ≈ 45–200 rows/day ≈ $4–18/month.

---

⚠️ Honest limits — what this does NOT do

1. ESPN carries ONE sportsbook. This is not a book-comparison product. Verified across 26 events in 5 leagues: DraftKings (provider 100) only for NFL, NCAAF, NCAAB, MLB, NHL, NBA and MLS. Bet 365 (provider 2000) appears as a second book on top European and South American soccer only — measured on EPL, Championship, LaLiga, Serie A, Bundesliga, Ligue 1, Eredivisie, Primeira Liga, Süper Lig, Argentine Liga Profesional, Brasileirão, Champions League, Europa League, Conference League and Libertadores. Requests for other providers (/odds/58 ESPN BET, /odds/1002) return 404 "no instance found" on an NFL event. If you need a five-book line shop, this is the wrong actor — and no ESPN-based actor can honestly claim one.

2. The closing line is null until the game starts. A scheduled game's odds object carries open + current only; the close block appears at kickoff. Every *Close* field, and clvHomeMoneylineOpenToClose, is null on a pre-match row and is never silently mirrored from current. That is why close fill was 33% on the default (mostly upcoming games) and 56% on the wide backfill run.

3. Coverage is line-posting dependent, not game dependent. DraftKings posts most lines roughly a day out — MLB especially. Measured over an 8-day forward window: 8 of 100 MLB games carried a line; the other 92 did not have one posted yet. A far-future date range legitimately returns fewer rows than games. When that happens the run exits SUCCEEDED with a status message naming the reason ("No betting line posted yet… This is a real answer from ESPN, not a block"), never a failure, and never as "no results found".

4. Bet 365 rows are moneyline-only. Provider 2000 ships home/away/draw prices with no spread, no total and no open/close history. Those columns are null on a Bet 365 row by design — that is what ESPN returns, not a parsing gap.

5. ESPN's own moneylineWinner / spreadWinner booleans are dead. They read false on every completed game sampled, including a settled NFL final the favourite won. They are shipped verbatim as real API fields; the usable graded result is in moneylineWinnerSide / spreadResultSide / totalResult, computed here from the final score against the closing line.

6. Leagues deliberately excluded because ESPN carries no line for them: UFC/MMA (27 fights across three cards sampled — zero with odds, including past cards), PGA, F1, NCAA men's hockey and the NBA G League. They are absent from the enum rather than present and empty.

7. Player props and futures are out of scope. ESPN exposes a propBets collection on the same object (242 items on one NFL game) and an FPI predictor. Both are deliberately not scraped — bundling them would make this a combo actor rather than one that does one thing completely.

8. Two league slugs need per-day queries. mens-college-basketball and womens-college-basketball 404 on a date range and return 200 on a single date. The actor walks those day by day automatically, and falls back to day-by-day for any league that starts refusing a range.

🤖 Anti-bot / proxy — measured, not assumed

There is no anti-bot on this feed. No API key, no auth, no Cloudflare, no browser, no cookies.

Path	Result
Home IP → scoreboard	200, 250,817 bytes
Apify proxy groups-AUTO → scoreboard	200, 250,817 bytes — byte-identical
Apify proxy groups-RESIDENTIAL,country-US → scoreboard	200, 250,817 bytes, 0.96 s
Burst: 30 core-API /odds calls @ concurrency 10 through the proxy	30 × 200, 0 non-200, 1.5 s wall
91-game college-football slate @ concurrency 10	91 × 200, 0 failures, 0.8 s

The proxy option is therefore off by default. Turn it on only if your own network blocks ESPN.

📄 robots.txt

Neither API host serves a robots.txt. Quoted verbatim, fetched 2026-09-03:

$ curl -i https://sports.core.api.espn.com/robots.txt
HTTP/1.1 403 Forbidden
<html>
<head><title>403 Forbidden</title></head>
<body>
<center><h1>403 Forbidden</h1></center>
</body>
</html>

$ curl -i https://site.api.espn.com/robots.txt
HTTP/1.1 403 Forbidden
<HTML><HEAD>
<TITLE>Access Denied</TITLE>
</HEAD><BODY>
<H1>Access Denied</H1>
You don't have permission to access "http://site.api.espn.com/robots.txt" on this server.
</BODY></HTML>

There is no crawl directive to honour on either host, and no rule on the separate www.espn.com robots.txt covers these API paths. Data returned is public, unauthenticated sports information. You are responsible for how you use it, and for gambling regulation in your own jurisdiction — this actor reports published prices and is not betting advice.

---

❓ FAQ

What is "closing line value" and why does this actor exist? CLV is the gap between the price you took and the price the game closed at. It is the single best predictor of whether a betting model is actually profitable, and it is why the close block matters more than the current one. Scoreboard-based scrapers cannot give it to you, because that feed drops odds at kickoff.

Which sportsbook are these odds from? DraftKings for everything, plus Bet 365 as a second book on top European and South American soccer. See honest limit #1 — this is not a multi-book comparison tool.

Can I get the in-play line for a game happening right now? Yes — set lineTypes: "live" and gameStatus: "live". Rows come back with lineType: "live" and providerName: "DraftKings - Live Odds".

Can I backfill historical closing lines? Yes. Set dateFrom / dateTo to a past range and gameStatus: "final". Those rows carry the close block and the graded moneylineWinnerSide / spreadResultSide / totalResult.

How do I track line movement over time? You don't need to. ESPN ships open, close and current inline on a single fetch — no cross-run diffing, no key-value store, no scheduled snapshots required for the basic move. Re-run on a schedule only if you want intraday granularity between open and close.

Why are odds in three formats? American for US books, decimal for modelling (implied probability is 100 / decimal), fractional because ESPN ships it. All three are taken straight from the API where present and derived from each other where not.

Why did my run return fewer rows than games? The line was not posted yet for those games. See honest limit #3. The run's status message tells you exactly which case you hit.

Does it need a proxy? No. See the measured table above — home IP, Apify AUTO and Apify RESIDENTIAL all returned byte-identical payloads.

How fast is it? Measured: 45 rows in 3.1 s, 237 rows across 183 games and 10 leagues in 3.8 s, at concurrency 10.

What happens if ESPN changes the payload? There is an integrity gate: if the actor pulls 10+ odds objects and not one yields a moneyline, spread or total, it fails loudly rather than delivering a dataset of empty rows you would still be charged for.

Are the derived probabilities de-vigged? Both are shipped. impliedProb* includes the book's margin; noVigProb* is normalised so the outcomes sum to 100%. holdPercentMoneyline and holdPercentTotal tell you how big that margin was.

---

🔧 Technical

Data source	site.api.espn.com/apis/site/v2/sports/…/scoreboard (context) + sports.core.api.espn.com/v2/sports/…/odds (the line)
Auth	none
Runtime	Node 20, plain HTTP (got-scraping), 1024 MB, no browser
Concurrency	10 by default
Zero-row behaviour	exits SUCCEEDED with an explanatory status message — a block and an empty slate are reported differently and are never conflated