COT Positioning Extremes: What 'Crowded' Really Means

A common way to read the Commitments of Traders report treats a positioning extreme as a countdown: speculators are this long, so the turn must be close. That is not what the number measures. On Meaterm’s public COT pages an extreme is a rank. The arithmetic behind it is published on the methodology page, which says in so many words that a label describes where positioning sits against its own history and is not a signal. This post explains the rank, the rules that decide when it is printed and labelled, and how to read one without turning it into a forecast. If the report itself is new to you, start with how to read the COT report.

What “extreme” means in numbers

Every figure starts from one trader category in one market. The net position is that category’s long futures contracts minus its short ones; the spreading positions the CFTC reports in a separate column are left out. On Meaterm’s public COT pages the category is Leveraged Funds from the CFTC’s Traders in Financial Futures report for the S&P 500, the Nasdaq-100, Euro FX and Bitcoin (CME), and Managed Money from the Disaggregated report for gold and WTI crude oil, all from the futures-only files.

The 156-week percentile ranks this week’s net position against the same market’s last 156 weekly net positions: count how many were below this week’s, divide by the number of weeks, multiply by 100 and round. A reading of 96 means this week’s net is higher than in 96% of those weeks, which for a group that is net short means less short than usual, not long. It is a plain empirical rank. It is not a z-score and not a probability, and it says nothing about any other market.

Four details decide what the number can be:

• This week is inside its own window. The 156 observations include the current one, so a week above every other week in the window reads 99, never 100, and a week below every other week reads 0.
• Below 8 weekly observations, nothing is shown. A rank over a handful of points is arithmetic, not information, so the column prints a dash instead.
• A short window caps the rank. With 8 observations the highest possible reading is 88, because at most 7 of the 8 can sit below this week. A rank over less than three years is a real number, and a different one.
• Only weeks in the same units count. From the report of 2 May 2023, the CFTC counts its S&P 500 and Nasdaq-100 Consolidated rows in E-mini contracts, with the micro contracts at a tenth each; the weeks before counted full-size contracts, each worth five E-minis. For those two markets the window holds only the weeks on a report’s own side of that date, so the first report with 156 weeks in the new units behind it was that of 21 April 2026.

That is why the public pages hold two lines. The COT report page prints a percentile once a market has 8 weekly observations. The weekly report, its dated pages and each market’s contract page print a percentile, and a label, only where the full 156 weeks exist, and a dash below that.

Four labels, and the sign of the net decides between them

A rank says where this week’s net sits against its own past, not whether the group is long or short. A group that has been net short for three years ranks near the top in the week it is least short. So the label depends on both the rank and the sign of the net:

• At 90 or above: Crowded long when the net is long, Net short, near 3-year high when it is short or zero.
• At 10 or below: Crowded short when the net is short, Net long, near 3-year low when it is long or zero.
• Anything in between: no label at all.

The “What changed in this week’s positioning” summary on the COT report page uses the same labels, the same thresholds and the same 156-week rule.

Percentile, the min–max “COT Index” and the z-score

There are three common ways to turn a net position into a number you can compare across weeks, and they answer different questions.

• Min–max scaling, often called a COT Index: this week’s net minus the lowest in the window, divided by the highest minus the lowest. It measures where this week sits between the two most extreme weeks.
• A z-score: this week’s net minus the window’s average, divided by its standard deviation. It measures distance from the average, in standard deviations.
• A percentile rank, the one Meaterm’s public pages print: the share of the weeks in the window that sat below this week.

An invented window shows the difference. Take 156 weeks in which one exceptional week reached a net long of 150,000 contracts, every other earlier week sat between 10,000 and 55,000, and the current week is 60,000. The percentile rank is 99: this week is above every week but one. Min–max scaling gives 36, because (60,000 − 10,000) ÷ (150,000 − 10,000) is just over a third: the single exceptional week sets the top of the range and squeezes everything else toward the bottom. A z-score moves too, since that one week lifts both the average and the standard deviation, and it answers in a unit that invites translating the reading into a probability, which assumes the positions follow a bell curve. Nothing says they do.

The rank has its own limit, and it is worth stating. It counts, so it cannot tell a week just above the previous high from one far above it: both read 99. It is used because it says what it means in one sentence, not because it sees everything.

The COT module inside the terminal is a different view. It shows more trader categories and ranks net positioning as a share of open interest, so its percentiles and labels are not the figures described here, as the methodology page also says.

Why an extreme is not a signal

An extreme describes a stock of positions, not a flow of orders. It says a group already holds more, or less, than usual; it does not say when, whether or in which direction that changes. Crowded positioning has stayed crowded for months, and has persisted while price moved the other way. A fund that follows a trend can stay long a rising market for as long as the trend lasts, and its net position ranks near the top for all of that time.

The calendar adds to it. The report describes positions held on a Tuesday, or on the Monday when a US federal holiday falls on the Tuesday. The CFTC normally publishes it on the Friday of the same week at 15:30 ET, and a US federal holiday moves the release to the following Monday, so the positions are at least three days old when they come out and usually ten days old when the next report replaces them. Meaterm’s copy of the CFTC files refreshes every six hours and the public COT pages rebuild at most once an hour, so a new report can take up to about seven hours to appear on them. Whatever happened after the as-of day is not in the figure.

And the rank is relative to three years only. A market whose last 156 weeks hold no comparable episode can read 99 at a level that would look ordinary over a longer history. Three years is a trade-off: a longer window spans regime changes that make old readings poor comparisons, and a shorter one makes every week look extreme.

What the pages add is context rather than a verdict. Each market’s contract page, such as the Euro FX page, prints how many weeks it has been since the net was last at least this high, or at most this low. A long gap says the reading is unusual. It does not say the reading will end.

Reading an extreme next to price and macro

The 2s10s spread post reads the yield curve next to credit spreads and financial conditions, and names COT positioning as another reading to set beside them. The same habit works from the other end. Put four readings side by side before giving an extreme any weight:

1. The rank. Where this week sits against 156 weeks, and whether the full history is behind it.
2. The week-over-week change. An extreme that is still growing and one that has started to shrink are different descriptions, even at the same rank.
3. Price. A crowded long in a rising market says the move is widely held. A crowded long in a falling market says a group is holding on through weakness. Neither says what comes next.
4. The macro backdrop. Whether rates, the dollar or, for gold, real yields describe the same situation the positioning does. What moves gold works through one market this way.

When the four agree, you have a consistent description. When they disagree, the disagreement is the thing worth understanding. In neither case do you have a forecast.

Which category’s extreme you are looking at

A Leveraged Funds extreme in the S&P 500 and a Managed Money extreme in gold come from different reports and different definitions, so their sizes cannot be compared. Each can only be compared with its own history, which is what the rank does. And every futures contract has a long and a short side, so one category’s net long is matched by net shorts elsewhere in the report, including among positions too small to be reported.

The category does not tell you the intent either. The CFTC classifies traders by their predominant business, not by what a given position is for, and its own notes give the example of a hedge fund, classified under Leveraged Funds, using currency futures to hedge the exposure on an over-the-counter position. How the categories are defined, report by report, is in TFF vs Disaggregated COT reports.

What the data cannot tell you

• Everyone’s position. The report covers markets in which 20 or more traders hold positions at or above the CFTC’s reporting levels, and the CFTC says those reportable positions usually make up 70 to 90 percent of a market’s open interest. Traders below the levels are not sorted by business: the report prints their longs and shorts only as one non-reportable total, so what that total is made of cannot be seen. A contract whose count of reportable traders drops below 20 leaves the report until it is back at 20 or more.
• Options exposure. The public pages read the futures-only files, so a group that is flat in futures and positioned through options shows as flat.
• When, how far or which way positioning unwinds. No column in the report carries it.
• Whether a figure is right. If one of Meaterm’s figures disagrees with the CFTC’s file, the file is right, and the correction goes on the errata page.

An extreme is a useful description of how stretched one group is against its own past. Read it with price and macro, as context. This week’s six markets are on the COT report page, each with its own contract page, and recent reports have dated pages in the weekly archive.

Open the Meaterm terminal free → Charts, a macro dashboard, CFTC COT positioning and fundamentals from SEC filings in one place. Free plan, no card; Pro is €9.99 per month.