What is the Commitments of Traders (COT) report?

A weekly report from the U.S. Commodity Futures Trading Commission (CFTC) that breaks down long and short positions by trader type in U.S. futures markets. Positions are measured as of Tuesday and published the following Friday at 15:30 ET.

What is the COT index?

A rescaling of a market's net positioning to 0–100 over a trailing window. 100 means the most one-sided the group has been over that window, 0 the least. cotreport computes it over a trailing 156-week (3-year) window on the group's net contract count, which is already comparable inside that market because the scale is rebuilt from the market's own high and low.

What does the 3-year percentile mean in the COT report?

Where this week's positioning sits within the market's own last three years. cotreport measures it on net % of open interest across a trailing 156-week window: the 97th percentile means 97% of the weekly readings in that window were at or below this week's. At or above 95 counts as crowded long, at or below 5 as crowded short.

What does %OI mean in the COT report?

Any position measured as a share of all open contracts in the market instead of as a raw contract count. cotreport reports every trader group's net on this scale and measures the 3-year percentile on it, because the same contract count means one thing in crude oil and another in lean hogs, and because a market's size drifts over twenty years. Read it in percentage points: +12 %OI means the group is net long by 12% of the market's open contracts.

What is a z-score in the COT report?

How far a reading sits from its own average, counted in standard deviations. Zero is exactly average and +2 is two standard deviations above it. A z-score is sensitive to how spread out the history is, so a market that normally sits still reaches ±2 on a smaller move than a wild one does. That is the difference from a percentile, which counts rank only and ignores the size of the gap.

How often do COT positioning extremes reverse?

The share of positioning extremes that reversed within a horizon. Pooled across all 44 markets, cotreport measures about 39.7% within eight weeks versus a roughly 38% baseline · a positioning extreme barely moves the odds.

What is managed money in the COT report?

Funds, CTAs and momentum traders in the CFTC's disaggregated report. cotreport treats managed money as “the crowd”: the speculative side that tends to chase price and pile into extremes.

Reference

COT report glossary

Every Commitments of Traders term used on cotreport, defined in plain English. Where a term is a measurement, the entry says how this site computes it. New here? Start with the COT report explained. See also the methodology and how traders use this.

Commitments of Traders (COT) report: A weekly report from the U.S. Commodity Futures Trading Commission (CFTC) that breaks down long and short positions by trader type in U.S. futures markets. Positions are measured as of Tuesday and published the following Friday at 15:30 ET.See also: Managed money, Commercials (hedgers), Nonreportable (small traders), Nowcast
Managed money: Funds, CTAs and momentum traders in the CFTC's disaggregated report. cotreport treats managed money as “the crowd”: the speculative side that tends to chase price and pile into extremes.See also: Non-commercials (large speculators), Commercials (hedgers), Disaggregated report
Commercials (hedgers): Producers and merchants who use futures to hedge a physical business. They usually sit on the opposite side of the crowd, so their net position mirrors speculative positioning.See also: Managed money, Nonreportable (small traders), Net position
Non-commercials (large speculators): Large speculative traders in the CFTC legacy report, reported separately from hedgers. The disaggregated report splits this group further into managed money and other reportables.See also: Managed money, Disaggregated report
Nonreportable (small traders): Traders whose positions sit below the CFTC's reporting threshold, counted as one residual bucket instead of being reported individually. cotreport labels this row small traders and prints its net % of open interest and its 3-year percentile beside the big speculators and the commercials. It is context, not the crowd cotreport grades.See also: Commercials (hedgers), Non-commercials (large speculators), Net % of open interest (net %OI)
Disaggregated report: The CFTC breakdown published from 2006 that splits the old speculator bucket into managed money, swap dealers, producers and merchants, and other reportables. Where a market is covered by it, cotreport's crowd series is managed money from this report; where it is not, the legacy non-commercial series is used instead. The finer group table shown on a market page is a display lens only, never an input to the graded record.See also: Managed money, Non-commercials (large speculators), Traders in Financial Futures (TFF)
Traders in Financial Futures (TFF): The CFTC breakdown for financial contracts such as currencies, rates and equity indices. It splits traders into dealers, asset managers, leveraged funds and other reportables, where leveraged funds are the closest equivalent to managed money. cotreport shows it on financial markets as a display lens, the same way it shows the disaggregated table: it does not feed the graded record.See also: Disaggregated report, Managed money
Open interest: The total number of futures contracts outstanding in a market. cotreport uses it to normalise positioning so markets of different sizes can be compared.See also: %OI (percent of open interest), Net % of open interest (net %OI)
Net position: A trader group's long contracts minus its short contracts. Positive means net long, negative means net short.See also: Net % of open interest (net %OI), Open interest
%OI (percent of open interest): Any position measured as a share of all open contracts in the market instead of as a raw contract count. cotreport reports every trader group's net on this scale and measures the 3-year percentile on it, because the same contract count means one thing in crude oil and another in lean hogs, and because a market's size drifts over twenty years. Read it in percentage points: +12 %OI means the group is net long by 12% of the market's open contracts.See also: Open interest, Net % of open interest (net %OI), 3-year percentile
Net % of open interest (net %OI): A group's net position divided by total open interest. cotreport expresses every position this way so a crowded gold market and a crowded hog market are measured on the same scale.See also: %OI (percent of open interest), Net position, 3-year percentile
COT index: A rescaling of a market's net positioning to 0–100 over a trailing window. 100 means the most one-sided the group has been over that window, 0 the least. cotreport computes it over a trailing 156-week (3-year) window on the group's net contract count, which is already comparable inside that market because the scale is rebuilt from the market's own high and low.See also: 3-year percentile, 156-week z-score, Positioning extreme (stretched)
Percentile: Where today's reading sits inside its own history, on a 0–100 scale. cotreport computes it as the share of weeks in the trailing 156-week window whose reading was at or below today's, one market at a time. A percentile here is always a rank against that market's own past and never a ranking of one market against another, and a market needs 78 weeks of history before any percentile prints at all.See also: 3-year percentile, Z-score, %OI (percent of open interest)
3-year percentile: Where this week's positioning sits within the market's own last three years. cotreport measures it on net % of open interest across a trailing 156-week window: the 97th percentile means 97% of the weekly readings in that window were at or below this week's. At or above 95 counts as crowded long, at or below 5 as crowded short.See also: Percentile, %OI (percent of open interest), Positioning extreme (stretched), Weeks in zone
Z-score: How far a reading sits from its own average, counted in standard deviations. Zero is exactly average and +2 is two standard deviations above it. A z-score is sensitive to how spread out the history is, so a market that normally sits still reaches ±2 on a smaller move than a wild one does. That is the difference from a percentile, which counts rank only and ignores the size of the gap.See also: 156-week z-score, Percentile
156-week z-score: How many standard deviations this week's net positioning is from its 3-year (156-week) average. cotreport computes it on the group's net contract count, and ±2 is one of the seven fixed definitions of an extreme. A large positive or negative z-score flags an unusually stretched reading.See also: Z-score, COT index, Positioning extreme (stretched)
Positioning extreme (stretched): A week where a market's speculative positioning reaches the top or bottom of its own 3-year range. cotreport calls these “stretched” and grades what happened next.See also: 3-year percentile, Weeks in zone, De-clustering (episode), Fade
Weeks in zone: How many consecutive weeks a market has been at its extreme, counted back from the newest report. The counter resets the first week the 3-year percentile leaves the zone, so a reading of 9 means nine unbroken weeks of the crowd leaning the same way. It measures persistence, not severity: the percentile says how hard, this says how long.See also: Positioning extreme (stretched), 3-year percentile, Crowd pressure
Crowd pressure: A 0–100 dial that combines four measurements cotreport already publishes: how hard the crowd leans (45% of the score), how many weeks it has leaned that way (20%), how much of this market's turn fingerprint is on the tape today (20%), and whether the live nowcast reads the crowd still adding (15%). When a component cannot be measured its weight folds back into the lean, so a thin market is never flattered by a missing part. It is a reading of the present state, never a forecast.See also: Weeks in zone, 3-year percentile, Turn fingerprint, Nowcast
De-clustering (episode): Consecutive weeks at an extreme are counted as one episode, not many, so a single long stretch doesn't inflate the sample. cotreport's record holds 1,206 de-clustered episodes since 2006.See also: Positioning extreme (stretched), Weeks in zone, Reversal rate (turn rate)
Reversal: A counter-crowd price move of at least 0.75× the market's own volatility within the measured window. It is the outcome cotreport grades each positioning extreme against.See also: Base rate (baseline), Reversal rate (turn rate), Front-month proxy
Base rate (baseline): How often a reversal happens in an ordinary week with no extreme present. cotreport prints every conditional reversal rate next to this baseline, so you can see whether the extreme actually changed the odds.See also: Reversal rate (turn rate), Reversal, Confidence interval (CI)
Reversal rate (turn rate): The share of positioning extremes that reversed within a horizon. Pooled across all 44 markets, cotreport measures about 39.7% within eight weeks versus a roughly 38% baseline · a positioning extreme barely moves the odds.See also: Base rate (baseline), The wait (median weeks to turn), Confidence interval (CI), De-clustering (episode)
The wait (median weeks to turn): When a market did turn, how long it took, as a median in weeks. cotreport measures the wait only across the episodes that actually turned inside the window, so it tells you what the ones that worked cost in patience and nothing about how long today's extreme will take. Most extremes never turn at all, which is why the wait is always printed beside the turn rate and never on its own.See also: Reversal rate (turn rate), Shakeout, Base rate (baseline)
Shakeout: How far a trade ran against you before the turn arrived. cotreport measures it only on the extremes that did turn: the worst move the wrong way between the entry and the moment the reversal threshold was crossed. The median shakeout is what a stop has to survive, which is why the published stop is the larger of the market's own median shakeout and one ATR.See also: Adverse excursion (MAE), ATR (average true range), The wait (median weeks to turn)
Adverse excursion (MAE): The worst the price went against the fade at any point inside the window, whether the case ever turned or not. cotreport publishes the median and the worst on record for each market over the 13-week horizon. It is not the shakeout: the shakeout stops counting at the turn and exists only for cases that turned, while the adverse excursion covers every case for the full window. That is why the worst case is always printed next to the median.See also: Shakeout, ATR (average true range), Reversal
ATR (average true range): The average size of a market's daily range, a plain measure of how far it travels in an ordinary day. cotreport uses a 20-session ATR as the floor under a stop, so a market whose past turns barely shook out never ends up with a stop tighter than one normal day of noise.See also: Shakeout, Adverse excursion (MAE)
Walk-forward test: Rules are fixed on 2006–2018 data and then confirmed on 2019-onward data the rules never saw. Only findings that hold in both eras are trusted.See also: Out-of-sample (confirmation era), Placebo test, FDR (Benjamini–Hochberg)
Out-of-sample (confirmation era): Read on data the rule never saw while it was being built. cotreport fixes every parameter on 2006–2018 and then reads 2019 onward once, as the confirmation era. A result that is strong in the training years and gone out of sample is the textbook signature of overfitting, and cotreport publishes those failures instead of only the fitted half.See also: Walk-forward test, FDR (Benjamini–Hochberg), Placebo test
Placebo test: A check against luck. cotreport draws 2,000 sets of fake events at random dates in the same market, keeping the real event count and the same minimum spacing, then counts how often chance alone matched or beat the real result. The placebo p printed on a backtest is that share. A small p means chance rarely did this well; on its own it is not a claim that the setup works.See also: FDR (Benjamini–Hochberg), Confidence interval (CI), Walk-forward test
Confidence interval (CI): The range a measured rate could plausibly sit in, given how few cases produced it. cotreport prints a 95% interval next to a turn rate, taken from 1,000 bootstrap resamples of that cell's own episodes. A wide interval is the honest signal that the sample is thin: 40% from a dozen episodes and 40% from two hundred are not the same number.See also: Placebo test, Reversal rate (turn rate), Base rate (baseline)
FDR (Benjamini–Hochberg): A multiple-testing correction that controls how many “significant” results are false positives. cotreport applies it at q=0.10; across 930 testable cells, zero survived correction plus out-of-sample confirmation.See also: Placebo test, Out-of-sample (confirmation era), Walk-forward test
Nowcast: A daily estimate of where positioning has drifted since Tuesday's data, because the published COT report is already three days old when it lands on Friday.See also: Commitments of Traders (COT) report, Crowd pressure
Front-month proxy: A continuous price series stitched from the nearest futures contract, used to measure outcomes. Roll gaps are included and no costs or slippage are assumed.See also: Roll gap, Reversal
Roll gap: The price jump that appears when a continuous front-month series switches from the expiring contract to the next one. No holder captured that jump, so it is not profit or loss. cotreport's graded record includes roll gaps rather than quietly smoothing them away, and a live trade whose window crosses a roll prints a roll-adjusted return with the gaps stripped out, labelled as such.See also: Front-month proxy, Adverse excursion (MAE)
Turn fingerprint: The set of conditions present at a market's past reversals, checked against today's conditions. In cotreport's tests the fingerprint gate was one of the few signals that improved outcomes in both eras.See also: Crowd pressure, Positioning extreme (stretched), Walk-forward test
Fade: To bet against the crowd: take the opposite side of a crowded position. cotreport's public trades fade positioning extremes and are graded win or lose.See also: Positioning extreme (stretched), Shakeout, Reversal

Statistics, not advice · Not affiliated with the CFTC · figures per research release r2