Commitments of Traders: Reading CFTC Positioning Data

Article

What is the Commitments of Traders report?

The Commitments of Traders report is a government-published breakdown of who holds what in the futures market. The U.S. Commodity Futures Trading Commission (CFTC) collects position data from firms that report large trader holdings and publishes a summary each week. The CFTC describes the reports as a breakdown of each Tuesday's open interest for futures and options-on-futures markets in which at least 20 traders hold positions at or above the CFTC's reporting levels.

Two terms deserve a plain definition up front.

• Open interest is the total number of contracts that have been entered into and not yet closed out, delivered, or exercised. It measures how large the market is, not which way it leans. Every long contract has a matching short contract, so open interest counts each pair once.
• Reportable positions are the holdings large enough to cross the CFTC's thresholds. Smaller traders are not itemized. In the CFTC's explanatory notes, reportable positions usually represent 70 to 90 percent of open interest in a given market, so the report covers most of a market but never all of it.

That second point matters for interpretation. COT shows a large, well-defined slice of the market, and a reader should keep in mind that the remaining slice is unobserved.

Why does positioning data belong in sentiment research?

Sentiment is usually described as mood: bullish, bearish, fearful, greedy. Futures positioning is different because it records exposure that participants actually carried into a reporting date. That makes it closer to behavior than a survey answer, which costs the respondent nothing.

Exposure is still not the same as conviction. A dealer can be short because the dealer is accommodating customer orders. A producer can be short because futures are hedging output that will be sold later. A leveraged fund can be long because a trend-following rule says so, with no view at all on long-run value. The same number of contracts can therefore carry very different meanings.

For that reason, positioning is best treated as one layer in a wider reading of the market, alongside surveys, options demand, volatility term structure, credit spreads, margin debt and leverage, and market breadth. The parent guide to market sentiment explains how these layers fit together, and the page on positioning versus sentiment versus breadth separates the three ideas more carefully.

What does the CFTC actually publish?

The CFTC publishes several report families, each built for a different kind of market:

• Legacy reports, organized by exchange, with futures-only and combined futures-and-options versions.
• Supplemental reports, covering a selected group of agricultural commodities.
• Disaggregated reports, organized by commodity type such as agriculture, petroleum, natural gas, electricity, and metals.
• Traders in Financial Futures (TFF) reports, which cover financial contracts such as currencies and Treasury securities.

For equity-index, Treasury, currency, and similar financial contracts, the TFF structure is usually the most informative. According to the CFTC's explanatory notes for that report, it sorts reportable traders into four categories: Dealer/Intermediary, Asset Manager/Institutional, Leveraged Funds, and Other Reportables.

That taxonomy prevents one of the most common errors in amateur COT work: treating "speculators" as one crowd. A pension manager, a macro hedge fund, and a bank dealer can sit on different sides of the same contract for unrelated reasons. Splitting them out lets a reader ask a sharper question.

How does the older commercial versus non-commercial split work?

The older Legacy reports sort traders by how they use futures. According to the CFTC, commercial traders use contracts for hedging as defined in its regulations, while non-commercial traders do not primarily hedge a business activity. That split is a regulatory classification, not a statement about skill or sophistication. Readers often hear "commercials are smart money" and "speculators are the dumb crowd." Neither phrase comes from the data. A commercial hedger is managing a business risk, and a non-commercial trader is doing something else. Neither group is trying to forecast for the benefit of a chart reader.

If you want to see how these older categories are tracked as indicators, the site's pages on the commercial hedger net position and the non-commercial net position explain the construction.

When is the data measured and when is it released?

COT is not real time. The position date is a Tuesday, and the CFTC processes the data over the following days before releasing it weekly, in the afternoon on the release day. So a reader looking at the market late in the week is seeing exposure from several sessions earlier.

Two dates therefore belong on any COT chart or note: the position date (what the numbers describe) and the publication date (when the public could first see them). Mixing them up is a classic source of error, especially in historical testing. If you assume a Tuesday position was known on Tuesday, you have given your test information it could not have had. That error is called look-ahead bias, and it makes past results look better than they could have been in real time. For a broader discussion of timing and revisions across data sources, see the guide to data latency and vintages and the source ladder.

The four questions to ask before interpreting any COT series

1. What contract is being measured?

A Treasury futures contract is not an equity-index futures contract, and a VIX futures position is not the same as a direct stock position. Start by naming the economic exposure inside the contract and noting how far it sits from the thesis you care about. A position in an index future says something about index exposure and little about any single stock.

2. Which participant category matters for the question?

If the question is about trend-following risk capital, leveraged-fund data may fit. If the question is about long-only institutional exposure, asset-manager data may fit better. Dealer data can show the other side of customer activity, but a dealer position is rarely a directional opinion. Our indicator pages on the leveraged funds net position and the asset manager net position describe how each series is built.

3. Is the position large because the market itself is larger?

Raw contract counts can climb simply because open interest grows. Before calling a position extreme, divide it by open interest or compare it with its own history.

4. Is the position extreme, changing quickly, or both?

A stable extreme and a fast-changing moderate position tell different stories. The level describes how crowded a trade is. The weekly change describes how fast the crowd is moving. A careful reader looks at both.

How do you calculate a useful COT measure?

Let long positions be L, short positions be S, and total open interest be OI.

• Net position = L minus S.
• Net position as a share of open interest = (L minus S) divided by OI.
• Position percentile = the rank of the current normalized net position inside a rolling historical window. A reading at the 90th percentile means the position is larger than about 90 percent of the observations in the window.
• Position z-score = (current normalized position minus the rolling average) divided by the rolling standard deviation. It expresses how many standard deviations the position sits from its recent norm.

Percentiles are easier for most readers to interpret. Z-scores are handy for analysts comparing series on different scales, but they assume a roughly bell-shaped history, which positioning data often are not. Both depend on the window chosen: a five-year window and a fifteen-year window can give different percentiles for the same week, so the window should always be stated.

Why are percentiles better than fixed thresholds?

No timeless number of contracts means "too bullish" or "too bearish." Futures markets grow, contract specifications change, participation shifts, and macro regimes alter normal hedging demand. A percentile asks a more durable question: how unusual is today relative to this market's own history?

A worked example with illustrative numbers only: suppose a category holds 400,000 long contracts and 250,000 short contracts in a market with 2,000,000 open interest. The net position is 150,000 contracts long, which is 7.5 percent of open interest. If, over the chosen window, 7.5 percent sits higher than about 95 percent of weekly readings, the position is at roughly the 95th percentile. That says the position is large for this market. It does not say price must fall. It says the position deserves a crowding check.

Why can gross positions reveal more than net positions?

Netting is convenient but can hide intensity. Imagine a category with 600,000 long contracts and 500,000 short contracts. The net is only 100,000 long, yet gross exposure is enormous. Another week might show 150,000 long and 50,000 short: also 100,000 net long, but with far less activity and a very different mix of directional and relative-value positioning. (Both examples are illustrative, not drawn from any real report.)

When both gross sides rise together, the market may be attracting more two-sided activity even though the net barely moves. That is useful context for judging liquidity and crowding. A reader who looks only at the net figure would miss it.

The CFTC's explanatory notes also describe spreading, which captures equal long and short holdings by the same trader. In their example, a trader with 2,000 long and 1,500 short contracts shows 500 long and 1,500 spreading. Spread positions are partly self-offsetting, so a category with large spreading is often running relative-value trades rather than a plain directional bet.

How should changes in positioning be broken down?

A change in net positioning can come from four paths:

• long additions (new buying),
• long reductions (selling to close),
• short additions (new selling),
• short reductions (buying to cover).

These paths are not equivalent. A 50,000-contract rise in net long caused by new long buying is a different event from the same rise caused by short covering. The first adds fresh exposure. The second removes an existing bet. Splitting weekly changes into these four flows gives a clearer behavioral picture without pretending to know any trader's motive.

Crowding is a condition, not a catalyst

A crowded trade can stay crowded. If a trend keeps rewarding a position, new participants can join and existing ones can add. Risk appears when the position becomes vulnerable to a change in information, liquidity, volatility, or risk limits.

This distinction matters because many weak sentiment strategies mechanically fade every extreme. A more defensible reading sequence runs like this:

1. Identify an unusual positioning state.
2. Identify the thesis that probably supports it.
3. Identify the evidence that would challenge that thesis.
4. Watch price, breadth, volatility, and liquidity for signs the crowd is being forced to adjust.
5. Treat any unwind as an event to study, not an outcome to assume.

Crowding raises fragility. It does not set the date of a reversal.

Why do dealer positions need special care?

Dealer or intermediary exposure is often misunderstood. Dealers can carry offsetting positions because they facilitate customer trades and manage inventory, basis, options, or balance-sheet risk. A large dealer short is therefore not automatically a bearish call. The same principle holds for many hedging categories: direction alone does not reveal motive.

A good habit is to attach a standing reminder to dealer data: exposure does not equal forecast.

How do asset managers and leveraged funds differ?

Asset-manager positions can reflect benchmark exposure, duration management, asset allocation, or hedging. Leveraged-fund positions are more likely to include macro, relative-value, trend, arbitrage, and tactical exposures. That makes leveraged-fund changes useful for crowding research, though still not pure sentiment.

Comparing categories is stronger than isolating one:

• Asset managers heavily long while leveraged funds are heavily short can signal a disagreement in horizon or strategy.
• Both categories moving the same way can show broad participation.
• A sharp leveraged-fund reversal while asset managers barely change may point to tactical de-risking rather than a wholesale institutional shift.

How do you pair COT with price?

Positioning becomes more informative when price is added. Four simple states help:

The table is a question matrix, not a signal matrix. Its purpose is to tell the reader what to examine next.

How do you pair COT with volatility, breadth, and credit?

Volatility answers a different question: how much uncertainty the market is pricing. The Cboe describes the VIX Index as a measure of market expectations of near-term volatility conveyed by S&P 500 Index option prices. A crowded futures position alongside a sharp rise in implied volatility can indicate rising demand for protection or unstable risk budgets. A crowded position alongside unusually low volatility can indicate complacency, though only in context. The pages on VIX term structure and the VIX term structure indicator show how to read the shape of the volatility curve.

Equity-index positioning can look constructive while internal breadth deteriorates, meaning fewer stocks are carrying the index higher. Credit spreads can widen even while the headline index sits near a high. The ICE BofA US High Yield Option-Adjusted Spread, published on FRED, measures the extra yield investors demand to hold below-investment-grade corporate bonds over Treasuries, so it is a common read on credit risk appetite. These divergences matter because they reveal whether risk appetite is broad or concentrated.

A useful mental model treats COT as one node in a chain:

Positioning, then price, then breadth, then volatility, then credit, then macro regime, then risk controls.

Each link either supports or undercuts the one before it. For the macro end of that chain, the macro economics and market regimes section describes the background conditions that change how a position should be read.

A practical example

Suppose leveraged funds sit at a very high net-long percentile in an equity-index futures contract. The index is rising, breadth is strong, high-yield spreads are stable, and the VIX curve slopes upward in its normal shape. The sound conclusion is not "sell because positioning is extreme." It is closer to this: the trend is well sponsored and crowded, so the cost of a negative surprise may be larger if several participants try to cut exposure at once.

Now change the evidence. The same extreme long positioning remains, but breadth narrows sharply, high-yield spreads widen, and front-month volatility rises above later maturities. The crowding has not changed much, but the fragility evidence has. Recognizing that kind of multi-source shift is the real skill. (All figures and conditions in this example are hypothetical.)

Advanced interpretation: separate exposure from motive

The hardest COT question is not calculating net contracts. It is inferring what those contracts mean. A position can exist because a participant expects a directional move, is hedging another asset, is running a basis trade, is offsetting option exposure, or must hit a portfolio target for duration or beta. The CFTC category narrows the plausible set of motives, but it does not identify the motive of any single participant inside it.

A clean way to keep this straight is to separate three layers:

1. Reported fact: gross long, gross short, spreading where applicable, net position, open interest, and change from the prior report.
2. Historical context: the percentile or normalized position.
3. Interpretation: plausible explanations, listed next to the evidence that would contradict them.

Keeping the layers apart stops factual data from being blended into editorial inference.

Which open-interest denominator should you use?

Dividing net position by total open interest is a good first normalization, but not the only one. Analysts can also divide by reportable open interest, by the category's own gross exposure, or compare against a historical distribution that accounts for contract changes. Each denominator answers a slightly different question.

Total open interest is the most intuitive default: what share of the market's outstanding contracts does the net position represent? Whichever denominator is chosen, state it. It also helps to know whether combined futures-and-options data are used. The CFTC converts options to a futures-equivalent basis using delta factors, so a reader should avoid layering homemade option-equivalent transformations on top unless the method is spelled out.

How do you compare across contracts?

Comparing related contracts can say more than staring at one series. Equity-index positioning across large-cap and small-cap futures can show whether risk appetite is broad. Treasury futures across maturities can show where duration views are concentrated. Currency futures can add context on dollar positioning. VIX futures positioning can help explain crowding in the volatility market.

Contract economics must be respected. A net long of 20,000 contracts in one market is not comparable with 20,000 in another, because multipliers, volatility, and open interest differ. Use percent of open interest or standardized percentiles for cross-contract comparisons.

How should a historical study of COT extremes be designed?

Anyone testing what happens after extreme positioning should avoid cherry-picking famous reversals. Define the episodes before looking at forward returns. For example: identify every week in which leveraged-fund normalized positioning exceeded its 95th historical percentile, then describe the distribution of outcomes at several horizons. A fair summary shows the median, the interquartile range, the best and worst cases, and the sample size. Overlapping weeks are not independent observations, so a handful of extreme stretches can look like dozens of data points.

Even then, the result is not a stable trading edge. Contract specifications, participants, monetary regimes, and liquidity structures change. The educational value lies in understanding the range of outcomes around crowding, not in manufacturing certainty.

What does a well-documented COT note contain?

Reproducible research beats memory. A note worth keeping records the contract, the report type, the participant category, the observation date, the publication date, gross long, gross short, net, open interest, the normalized net, the percentile, the one-week and four-week changes, the price trend, the volatility, breadth, and credit conditions, the interpretation, the contradictory evidence, and a date to review it again. Written down this way, a reading can be revisited after the next report and judged against what actually happened.

It also helps to treat missing data honestly. If a weekly file is delayed or a market drops out of the report (the CFTC notes that a market leaves the publication when fewer than 20 traders hold reportable positions), mark the series unavailable instead of carrying the last value forward as if it were new.

A six-part checklist for reading any COT series

• Contract: what exact contract is analyzed, and what does it economically represent?
• Participant: which category is relevant, and what motives are common inside it?
• Normalization: is the series divided by open interest and compared with its own history?
• Change: is the information in the level, the weekly change, or a multi-week trend?
• Confirmation: what are price, breadth, volatility, and credit saying?
• Failure condition: what evidence would show the crowding thesis is wrong?

This sequence gives a reader a process without turning the data into a recommendation. For turning any such reading into position limits, the risk management section and the position sizing tools address exposure control separately from market views.

Common mistakes

Treating commercial and speculative categories as good versus bad

The categories describe economic roles and report classifications, not superior or inferior forecasting skill.

Ignoring open interest

Raw contract counts are not comparable across time when the market grows or shrinks.

Using the wrong timestamp

Use the date the information became public, not only the date it refers to.

Fading every extreme

Extremes can persist in strong trends. Crowding increases fragility. It does not set the reversal date.

Double counting related measures

COT, futures open interest, and some dealer-positioning estimates can partly describe the same underlying activity. A composite that adds them together may count one thing twice. The sentiment composite framework discusses how to avoid that.

Forgetting what is missing

COT covers reportable positions only. The unobserved remainder of the market can behave differently, and the report says nothing about options not listed on futures, cash markets, or swaps.

What COT does not tell you

• It does not forecast price. It describes exposure that existed on a past date.
• It does not reveal intent. A category label narrows motives without identifying them.
• It does not reveal the entry price or profit of any position, so it cannot say how much pain a crowd is in.
• It does not show activity between the Tuesday snapshot and the release.
• It does not cover every market. Contracts without enough reporting traders are not published.

Keeping these limits in view is what separates a useful reading from a story told after the fact. If you are unsure how much weight a dataset like this deserves, the source ladder lays out how to rank evidence by how close it is to the primary record.

Frequently Asked Questions

References

• CFTC: Commitments of Traders
• CFTC: Commitments of Traders Explanatory Notes
• CFTC: Traders in Financial Markets Explanatory Notes
• Cboe: VIX Volatility Products
• FRED: ICE BofA US High Yield Index Option-Adjusted Spread