DraftKings API questions answered

Stored sportsbook observations, Pick6 market rows and separate DFS datasets, plus independent ESPN statistics. Each has its own limits.

Create an account, open the dashboard and create a named key. The secret appears once, and every /api/v1 data request needs it as a Bearer token.

Native sportsbook streams price changes. Scheduled datasets rotate through captures; row timestamps and scope health show current freshness.

Native New York NHL player props cover shots, points, assists, saves, goalscorers, power-play points and first-period shots.

DFS is an authenticated beta on the canonical host. Outside it, or when permission is withheld, the route returns 403. Do not bypass it.

Free is 500 requests a month at 20 a minute; paid plans raise both. Limits are shared across keys and reset on UTC boundaries.

No. Admin, dashboard and auth pages are private surfaces excluded from crawl discovery. Use the docs, playground and status pages instead.

No. This is an independent developer service, not affiliated with or endorsed by DraftKings, and access is not a redistribution license.

Yes. Free includes 500 requests per UTC month at 20 a minute on the same captured datasets, with no stream or webhooks.

This service is not an official API. We do not know of a public DraftKings developer API, and DraftKings has not endorsed this one.

An API key gives technical access, not a redistribution license. Rows carry rights notes, and you must resolve rights for anything you publish.

Line recording began around 2026-11-03, so history is short. It holds no past seasons of lines and is not an opening-to-close archive.

A source value on a Pick6 selection, such as 1.5x. It is not decimal odds, a probability or an established entry payout.

Not reliably. The inverse-odds formula assumes a price. A standings multiplier is a different thing, and nothing here establishes that mapping.

Its last observation is over 15 minutes old. Retained rows stay readable after collection moves on, so stale marks old evidence, not an error.

Pick6 endpoints accept format=csv, up to 100 rows per page. Sportsbook, DFS and statistics are JSON, so convert them on your side.

Yes, for Pick6 changes on the Production plan. Callbacks are signed and delivered at least once. Sportsbook and DFS have no webhooks.

Developer and Production on the monthly contract. Each connection lasts up to 55 seconds, costs one request and resumes from a saved ID.

No. It is plain HTTP with a Bearer header, so cURL, Python urllib and Node fetch all work. There is no vendor package to install.

Keep the key on a server. A browser bundle would expose it, and data endpoints do not send cross-origin headers. Use a server route.

Most lists allow limit up to 100 with offset up to 10,000. Statistics use a cursor with limit up to 200, and Pick6 history uses after.

No. There is no public grading endpoint. Game phase is not settlement, and you must apply your own rules to compare targets and results.

Not over seasons. Line recording began recently and statistics are limited samples, so any result is a small sample rather than a track record.

Native sportsbook uses scope IDs. ESPN mainlines use NCAAF for college football; Pick6 and DFS use CFB. Current inventory is on the status page.

NBA Pick6 rows yes, native sportsbook NBA props no. Pick6 captures are selected and can be stale, so check rows before relying on them.

NFL Pick6 rows yes, including receiving, rushing and passing yards labels. Native sportsbook NFL props are not supplied.

Native NHL shots, points, assists, saves, goalscorers, power-play points and first-period shots, alongside separate Pick6 rows.

Not on the sportsbook side, as of the date stated. Pick6 alternative targets exist, but they are not sportsbook alternate lines.

Pick6 entity_id has matched DFS playerDkId in a bounded NFL check. ESPN athlete_id is separate, and no mapping is supplied, so verify your own.

ESPN-attributed rows use ESPN event numbers. Native sportsbook and Pick6 retain their DraftKings source IDs.

A 401 means the key is missing, malformed or revoked. A 403 means an origin, plan or redistribution boundary, not a problem with the key.

Your account limit was reached. Wait for the next UTC minute, and if it persists the UTC month. Avoid tight retry loops, since keys share one budget.

A 503 means storage or a dataset is not available. An empty page means the query matched nothing. Neither should be read as zero offers.

Yes. The hosted MCP endpoint offers keyless search and fetch tools plus data tools that need a Bearer key in the client configuration.

Yes. The OpenAPI document at /api/openapi is public and needs no key. A reduced document for ChatGPT Actions is at /openapi-chatgpt.json.