How do I download CFTC Commitments of Traders data?

The query reads like a request for a file. It has never been about the file - it is about whether anyone can answer, from evidence rather than chat-room conviction, who actually holds the position in America's exchange-traded futures markets.

One publisher owns that record outright: the U.S. Commodity Futures Trading Commission, whose Commitments of Traders program breaks down weekly futures and options open interest by trader category, market by market. In Datadory's catalog the record appears as CFTC Commitments of Traders (COT) Reports - quality 9 out of 10 against an all-catalog average of 7.81 across 1,744 datasets, a band shared by just 534 records.

The scale runs weekly and wide. Roughly 100-plus contract markets publish in a typical report week across CBOT, CME, CBOE, MIAX, COMEX, NYMEX, ICE Europe and ICE Energy, Coinbase Derivatives and Nodal, generally in both futures-only and combined futures-and-options versions. Every row reflects Tuesday's open-interest snapshot, published at the end of the same week and shifted by holidays - a rhythm Datadory keeps up so you never have to.

One row from the August 2026 research pass - the market label format, the Tuesday report date, and the long/short symmetry that makes the whole schema readable:

market_and_exchange_names   WHEAT-SRW - CHICAGO BOARD OF TRADE
report_date_as_yyyy_mm_dd   2022-09-13
open_interest_all           287046
noncomm_positions_long_all   88091
noncomm_positions_short_all  96219
comm_positions_long_all     119219
comm_positions_short_all    106242
traders_tot_all                315

Request a sample cut to your markets and report families - real rows come back with the complete field dictionary attached.

The same pool of open interest gets sliced four ways, and each cut encodes a different theory of who matters:

Managed Money is the spec-position line quoted in gold, crude and grain commentary; Leveraged Funds plays the same role on the financial side. Picking the wrong family for the question wastes a month - the table below settles the choice in a minute.

Identification comes first, because joins live or die on it. market_and_exchange_names gives the human-readable key (WHEAT-SRW - CHICAGO BOARD OF TRADE), report_date_as_yyyy_mm_dd stamps the Tuesday snapshot, and the six-character contract code (001602) plus exchange mnemonic (CBT) form the stable identifiers other regulatory tables reuse. A commodity rollup (001 / WHEAT) folds related contracts together.

Derived columns ride alongside: change_in_open_interest_all (-2,283 in the sample row) for week-over-week framing, pct_of_open_interest_all (30.7) expressing each category as a share, and traders_tot_all (315) counting distinct reporting traders - suppressed wherever fewer than four active traders exist in a category. Long-format variants repeat the same buckets for _old and _other maturity groups, and concentration ratios ship alongside on request.

Coverage differs sharply by family, which matters more than any single number. Legacy reaches January 15, 1986 - nearly four decades of continuous weekly observations. Disaggregated and TFF both begin June 13, 2006, and CIT begins January 3, 2006, so a factor built on Managed Money or Leveraged Funds carries twenty years of history at most, not forty. At roughly 100-plus markets a week, the archive accumulates thousands of market-weeks per year.

A property rare among financial datasets makes backtests cleaner than usual: records are not revised once published. There are no silent restatements to reproduce around - an archived week stays put forever.

Two scope rules belong in any model as explicit missing-data logic rather than surprises. Individual trader counts vanish below four active traders per category, and a market exits reporting once fewer than 20 reportable large traders remain - absence can be a rule, not a gap.

• Investors and quants backtest positioning signals on nearly forty years of weekly rows - enough history to trust a factor, on data that never revises underneath them.
• Data scientists and ML engineers engineer Managed Money and Leveraged Funds features into models, on a stable schema with documented suppression behavior.
• Competitive-intel and product teams read crowded positioning as confirmation or contrarian signal before bets tied to commodity inputs.
• Journalists and academics explain who is long and short using figures published by the market's own regulator - citable at attribution strength few vendor feeds match.
• Developers and data-product builders wire positioning into dashboards and alerts where the parsing, suppression rules and joins are already done.

On Datadory's relevance rubric the record scores 3 of 3 for investors and quants and for journalists and academics, 2 of 3 for developers and builders.

Positioning explains who moved; four named products supply everything it cannot say:

• CFTC Bank Participation Reports - the monthly banking cut of the same open interest, splitting positions between U.S. and non-U.S. banks in every market where five or more banks hold reportable positions. Pair them and you see both the category and the institution behind it.
• EOD Historical Data APIs - the price leg most positioning strategies need: 150,000-plus tickers of OHLCV bars with thirty-plus years of depth on many instruments. Positioning gives the signal, end-of-day bars give the entry.
• World Federation of Exchanges Statistics - venue-level volume and open-interest context across 60-plus exchanges, useful for normalizing positioning against market size.
• World Bank Capital Markets Indicators (WDI) - seven harmonized market-development series for 200-plus economies annually from 1975, the macro covariates beside microstructure.

For the closest regulatory head-to-head, the COT vs FINRA TRACE comparison contrasts weekly derivatives positioning with corporate, agency and Treasury bond transparency.

Because the record is famous and still awkward. Category taxonomies differ per family, so mixing Legacy buckets with Disaggregated ones corrupts a panel silently. Suppression masquerades as missingness. Futures-only and combined futures-and-options cuts double row counts mid-pipeline. Maturity buckets multiply every position column by three. Each is survivable once; none is fun to re-solve in every new notebook.

Datadory normalizes before delivery: one schema across families or kept separate as specified, suppression surfaced explicitly, join keys stable across decades, price legs pre-matched when the strategy needs them.

Start with the anchor record, CFTC Commitments of Traders (COT) Reports - sample rows, the full field dictionary and coverage chips live on its dataset page, sampled to your markets before anything is committed.

Then see the record in context. The best diversified capital markets datasets ranking scores all ten leaders side by side, and the diversified capital markets data hub indexes every product with its coverage statement - from BIS debt securities outstanding back to 1946 through end-of-day price tapes.