US Commodity Futures Trading Commission

CFTC Commitments of Traders — Who is long and who is short every futures market, weekly, since 1986

When you want to know positioning rather than price — how long the funds are, whether commercials are hedging the other way, and how today compares with the history of that market. The Tuesday snapshot is released Friday.

positioningcommoditiesfxratesUnited Statesglobal

Asset classespositioning · commodities · fx · rates

Frequencyweekly

History1986 → today

Limit · costNo key · Socrata API, 1,000 rows per page by default

KeyNo key — Optional free Socrata app token lifts the rate limit.

Formats · pull withjson · csv · txt — python, excel, sheets

LicenceUS government work, public domain.

RedistributeYes, with attribution

Best for“Are speculators crowded into a trade?” · “How positioned are funds in the yen, gold, or 10-year futures?”

When to use it

Reach for it when…

Not for: Anything intraday or daily (the report is weekly and three days stale), options positioning by strike, equities that do not trade as futures.

How to read it

Units, revisions, traps

Units. Contracts, by trader category. Legacy report: commercial, non-commercial, non-reportable. Disaggregated: producer, swap dealer, managed money, other. Net = long − short.

Revisions. Rare corrections are republished with a note; the history is stable.

Contract sizes differ across markets — compare net positions to open interest or to their own history, not across markets in contracts.
'Non-commercial' is not 'hedge funds' exactly; the disaggregated report's 'managed money' is closer.
Market names change over time (e.g. 'E-MINI S&P 500' vs 'S&P 500 Consolidated'); filter on the market code.

Classic mistake: Calling an extreme net-long 'bullish' — extremes are contrarian more often than not.

How to use it in your own work

Three recipes

Recipe 1 · Net speculative position in one market
Recipe 2 · This week's full report, one text file
Recipe 3 · Managed money in the disaggregated report

Each recipe: Python · Excel · Sheets, with how to read the result. Python recipes run under pandas; recipe 1 is re-run by the weekly check where the source allows it.

How to use it in your own work

Copy the snippet, change the series id, read the result the way the footer says. Replace YOUR_…_KEY with your own key where one is needed.

Recipe 1Net speculative position in one market

# legacy futures-only report via the Socrata API: 10-year Treasury note, net non-commercial
import requests, pandas as pd
u = ("https://publicreporting.cftc.gov/resource/6dca-aqww.json?$where=cftc_contract_market_code='043602'"
     "&$order=report_date_as_yyyy_mm_dd%20DESC&$limit=104")
df = pd.DataFrame(requests.get(u, timeout=60).json())
df["date"] = pd.to_datetime(df.report_date_as_yyyy_mm_dd)
df["net_spec"] = df.noncomm_positions_long_all.astype(int) - df.noncomm_positions_short_all.astype(int)
print(df.set_index("date")[["market_and_exchange_names", "net_spec", "open_interest_all"]].head(8))

=IMPORTDATA("https://publicreporting.cftc.gov/resource/6dca-aqww.csv?$where=cftc_contract_market_code='043602'&$limit=500")

How to read the resultNet as a share of open interest, and where that sits in its own 3-year range, is the readable number. A record short in Treasuries was the 2023 story.

Recipe 2This week's full report, one text file

# every market, latest week, from the plain-text file the CFTC posts each Friday
import pandas as pd
cols = ["market", "date_yymmdd", "date", "code", "exchange", "x1", "x2", "oi", "nc_long", "nc_short", "nc_spread", "c_long", "c_short"]
df = pd.read_csv("https://www.cftc.gov/dea/newcot/deafut.txt", header=None, usecols=range(13), names=cols)
df["net_spec"] = df.nc_long - df.nc_short
print(df.sort_values("net_spec")[["market", "oi", "net_spec"]].head(10))

=IMPORTDATA("https://www.cftc.gov/dea/newcot/deafut.txt")

How to read the resultThe extremes at the top and bottom of the sorted list are the crowded trades of the week.

Recipe 3Managed money in the disaggregated report

# disaggregated futures-only (dataset 72hh-3qpy): managed-money net in gold (088691)
import requests, pandas as pd
u = ("https://publicreporting.cftc.gov/resource/72hh-3qpy.json?$where=cftc_contract_market_code='088691'"
     "&$order=report_date_as_yyyy_mm_dd%20DESC&$limit=156")
df = pd.DataFrame(requests.get(u, timeout=60).json())
df["date"] = pd.to_datetime(df.report_date_as_yyyy_mm_dd)
df["mm_net"] = df.m_money_positions_long_all.astype(int) - df.m_money_positions_short_all.astype(int)
print(df.set_index("date")["mm_net"].head(8))

How to read the resultManaged money is the trend-follower category; producers lean the other way by construction.

Series → question map

The ids we use from CFTC Commitments of Traders, each with the question it answers. The catalog's compare view reads the concept tags behind these rows.

Series	Answers	Region	Concept
6dca-aqww · 043602	How are funds positioned in 10-year Treasuries?	United States	cot
6dca-aqww · 088691	Is gold crowded?	global	cot
6dca-aqww · 097741	How is the yen positioned?	Japan	cot
6dca-aqww · 13874A	Positioning in S&P 500 futures	United States	cot

Compare with

Same question, different source: Cboe index data, Stooq, FRED. The compare view lines up coverage, frequency, history and access side by side and lists what the combination makes possible.

Open compare: CFTC Commitments of Traders · Cboe index data · Stooq →

Questions readers ask

When is the COT report published?

Friday at 15:30 Eastern, showing positions as of the previous Tuesday.

Legacy or disaggregated?

Legacy has the longest history (1986) and the familiar commercial/non-commercial split; disaggregated (2006 →) separates managed money from swap dealers. Financial futures have their own 'TFF' version.

Where do market codes come from?

The cftc_contract_market_code column in any row; search once by name, then filter by code — names change, codes do not.

Educational only — we explain, we never advise · snippet licence: public domain · corrections to [email protected], fixed within a day and logged in the changelog.