Export DraftKings data to CSV

Which endpoints export CSV

Native CSV exists on the Pick6 endpoints only: projections and history, along with their aliases. Other datasets return JSON, so convert those on your side.

Native Pick6 CSV

Add format=csv. Each page returns at most 100 rows, with the usual Bearer header, and it costs one request. The file is named pick6-observations.csv and sent as text/csv.

curl --fail-with-body \
  "https://draftkingsapi.com/api/v1/pick6/projections?sport=NHL&stat=Shots%20on%20Goal&limit=100&format=csv" \
  -H "Authorization: Bearer $DRAFTKINGS_API_KEY" \
  -o shots-on-goal.csv

The columns are pick_group_id, pickable_id, pickable_market_id, sport, stat, target_value, period, is_live, is_paused, observed_at, source_updated_at, source_url, capture_hash, selections and entities.

JSON cells and formula safety

selections and entities stay as JSON inside their cells, so a row with two sides and two multipliers keeps its structure. Parse those cells if you need one column per side. Do not flatten them to a single multiplier: that would discard a side or merge different sides.

Values that begin with =, +, -, @, a tab or a carriage return get a leading apostrophe. This stops spreadsheet programs from running a cell as a formula. A consequence is that a negative value can arrive as text with an apostrophe in front, so strip it when you load the file into code.

CSV responses contain no pagination block, and they omit age_seconds and stale. Use observed_at to judge age. For repeated exports, request JSON, read has_more and write CSV yourself.

Page through a longer export

To go beyond one page, increase offset by 100 until a page comes back with only its header. That final request costs one request, so an export of N pages costs N+1 unless you stop on a short page.

offset=0
for page in 1 2 3; do
  curl --fail-with-body -s \
    "https://draftkingsapi.com/api/v1/pick6/projections?sport=NFL&limit=100&offset=$offset&format=csv" \
    -H "Authorization: Bearer $DRAFTKINGS_API_KEY" > "page-$page.csv"
  [ "$(wc -l < "page-$page.csv")" -lt 100 ] && break
  offset=$((offset + 100))
done
head -n 1 page-1.csv > all.csv
for f in page-*.csv; do tail -n +2 "$f" >> all.csv; echo >> all.csv; done

A count of lines is only a rough stop rule, because JSON cells are quoted and stay on one line in practice but are not guaranteed to. Prefer the JSON approach below if accuracy matters. Captures change between pages too, so deduplicate on the three identity columns.

Client-side conversion

For sportsbook, DFS and statistics, read JSON pages and write the columns you want. export-csv.py does that with the standard library, a page budget and the same formula guard.

python3 export-csv.py --resource odds --sport NBA --max-pages 3 --out odds.csv
python3 export-csv.py --resource dfs/players --sport NFL --max-pages 3 --out salaries.csv
python3 export-csv.py --resource stats/player-games --sport NBA --season 2027 --max-pages 3 --out logs.csv

Choose columns deliberately when you adapt it. Odds rows flatten cleanly, DFS data needs specific fields such as salary, displayName and draftableId, and statistics hold a list of stats to pivot by key, using numeric where it is not null.

Before you share a file

A CSV leaves your control once sent. Keep observed_at and source_url in every export, so the capture context travels with the values. Technical access is not a redistribution license, so review your rights before you publish or sell a file built from this data.

As of 2026-11-03, native sportsbook scopes cover New York NFL, NHL, college basketball and college football game lines and New York NHL Shots on Goal, points, assists, saves, goalscorers, power-play points and first-period shots. Read snapshots and streamed changes through the sportsbook endpoints. ESPN-attributed DraftKings moneyline, spread and total observations are also available across eight leagues. Current scope inventory is listed at /api/v1/sportsbook/scopes. Recording began around 2026-11-03. Retained history is short, and it is not a complete opening-to-close archive. It does not contain seasons of past lines, and there is no public graded-results endpoint.

Check the recorded data

• Historical sample data

Related resources