Use cases & endpoints
CFTC COT report API — weekly positioning data
The COT report answers a question no other public source does: how is open interest distributed across kinds of participants. Commercial hedgers, managed money and other reportables behave differently, and the composition of a market's open interest is often more informative than its level.
Using it well requires respecting two details that are easy to skip. The report describes positions as of Tuesday but is released the following Friday afternoon, and the legacy and disaggregated formats define categories differently, with the newer format available only from its introduction onward.
Hindcast exposes COT with both formats normalized into one schema, category definitions labeled rather than merged where they genuinely differ, and the release timestamp on every record so an as-of query cannot see a report before it was published. Contracts resolve through the shared identifier graph, so positioning sits beside prices and fundamentals in the same question.
hindcast — session
previewhindcast> positioning by category for a contract, as of a past date
→ tool: cot_positioning { contract:"...", as_of:"..." }
✓ the most recent report published by that date
source: CFTC Commitments of Traders, release timestamp retained
What's covered
- Legacy and disaggregated formats in one schema, differences labeled
- Long, short and spreading positions by trader category
- Report date and release timestamp on every record
- Open interest alongside positions for ratio measures
- History across format changes without back-filled categories
- Contracts resolved through the shared entity graph
Join the waitlist — this page's dataset is already selected.
Questions
Why is the Tuesday-to-Friday gap such a problem?
Because three days is enough for the market to move substantially. A study that timestamps positions to Tuesday is implicitly assuming its strategy could act on data that had not yet been published, which inflates results in a way that disappears in live trading.
Which categories are comparable across formats?
Some map cleanly and some do not, and the schema says which is which. Where the definitions diverge the series stay separate rather than being stitched into a continuous line that never existed.
Related
Join the waitlist. Vote the roadmap. First in, first served.