COT Data Reversals: Reading Crowded Positioning

COT data reversals are the turns traders hope to catch when the CFTC’s weekly report shows speculators packed into one side of a currency. The report can show you that a trade is crowded. It can’t tell you when the crowd will leave.

That is still worth knowing. A trend runs on new money, and once most of the money that wanted in is in, the trend has few buyers left and a lot of people who may one day want out at the same time.

For currencies, read the Traders in Financial Futures report and watch the leveraged funds: hedge funds, commodity trading advisors and other money managers (CFTC TFF explanatory notes). If that is new to you, start with COT Report Explained. Every currency’s weeks are on COT report for forex.

How a crowded trade comes apart

Say funds have spent months building a large short in a currency. Every one of those shorts is a buy order waiting to happen, because you close a short by buying the contract back.

While the story holds, nobody needs to buy. When it cracks, many of them need to buy in the same few days, and the currency rises faster than the news alone would explain.

The weak point of that logic is timing. Nothing in it says when the story cracks, and positioning can sit at an extreme for months while the trend carries on.

The yen in 2024

Leveraged funds were net short Japanese yen futures, a bet on a weaker yen, in every weekly report from 2 March 2021 to 24 September 2024 (CFTC TFF futures-only data). Over most of that stretch the bet worked: the dollar went from 106.75 yen on 2 March 2021 to more than 160 in the summer of 2024 (Federal Reserve noon rates, via FRED).

By 16 April 2024 the funds were net short 102,380 contracts, and the trade already looked crowded. If you had sold the dollar against the yen on that alone, you would have watched it climb from 154.57 yen that day to 161.73 on 10 July.

On 9 July the net short reached 110,635 contracts, the largest since November 2017 and about 32% of all open yen contracts. Then the turn came. By 5 August the dollar was at 143.95 yen, 11% below its July high. The Bank of Japan had raised its policy rate to around 0.25% on 31 July, in the middle of the fall (Bank of Japan statement).

Here is what the report showed, each figure for the Tuesday named:

• 9 July: net short 110,635
• 16 July: 75,045
• 23 July: 58,596
• 30 July: 70,333
• 6 August: 24,158

The CFTC publishes each Tuesday’s positions on the Friday, at 3:30 p.m. Eastern Time (CFTC COT reports). So the 6 August figure, the first to show most of the unwind, came out on 9 August, after the sharpest part of the fall. The funds did not turn net long until the report for 1 October.

That is the honest picture of a COT reversal. The crowd was a real risk, the report showed it for months before it mattered, and it confirmed the exit after the price had moved.

Measuring an extreme

A contract count on its own is hard to judge. A net short of 110,000 contracts is huge in some markets and ordinary in others. Three measures put it in proportion:

• Net as a share of open interest. The net position divided by all the contracts open in that market, the 32% above. It stays comparable over the years as the market grows or shrinks (What Is Open Interest?).
• A percentile. Where this week ranks in the market’s own weekly history. At the 90th percentile, funds are more net long than in 90% of the weeks in the window.
• A z-score. How many standard deviations this week sits from its average (What Is a Z-Score?).

If positioning followed a normal distribution, a z-score beyond ±2 would turn up about 4.6% of the time, about 2.3% on each side (NIST normal table). It doesn’t. Positioning trends and sits near one end for long stretches, as the yen did for three years. Read ±2 as unusual for that market, never as an exact probability.

The window you measure against changes the answer too. A year of history and ten years can disagree about the same week.

How often an extreme has turned

The yen is one case. For a wider view, take seven currencies against the dollar (the euro, yen, pound, Swiss franc and the Canadian, Australian and New Zealand dollars) and mark every week in which the leveraged funds’ net position first went beyond ±2 on a three-year z-score, using the weekly reports from June 2006 to October 2026 (CFTC TFF futures-only data). Then follow each currency’s dollar rate from the Friday that report came out (Federal Reserve H.10 noon rates).

That gives 91 separate episodes from 2009 to 2026. A new extreme in the same currency only counts once 12 weeks have passed, so no two episodes in one currency overlap. Twelve weeks later the currency had moved against the crowd in 58% of them. Over 4 and 8 weeks the figures were 56% and 54%. In any week of the same years, the currency moved against the funds’ side 50% of the time.

That gap is small, and with 91 episodes it is within what chance alone produces. Two thirds of the episodes also began within four weeks of one in another currency, because one move in the dollar pushes several currencies at once, so there are fewer independent cases than the count suggests. In about four episodes in ten, three months later the currency had moved further the crowd’s way.

So the history gives an extreme reading a slight lean towards a turn at most, and nothing you could lean on. It marks a risk worth watching.

Signs that the crowd is leaving

An extreme tells you when to start paying attention. These four patterns suggest an unwind may be under way. Each has given false alarms, so look for more than one.

• Price and positioning diverge. Price keeps making new highs or lows while the funds’ net position shrinks. The move has less fund money behind it.
• A fast unwind. The yen short went from about 111,000 contracts to about 24,000 in four reports, a cut of almost four fifths. By then the trend had lost most of the money that was driving it.
• The net position crosses zero. Funds go from net short to net long, or back. It is the clearest sign and the latest: the yen funds crossed in the report for 1 October, eight weeks after the sharpest fall.
• An extreme meets a change in the fundamentals. A crowd short a currency whose central bank starts raising rates has lost part of its reason to be there. That is when crowding can turn from a background risk into a live one.

Each currency’s page, such as the Euro COT report or the Japanese yen COT report, prints the net position, its change on the week and its share of open interest for every week, and where the latest week sits in the range of the weeks shown.

What newcomers get wrong

• Fading the trend too early. An extreme is a warning about risk. The yen short looked extreme in April 2024, and the dollar rose another seven yen.
• Ignoring why the crowd is there. Funds short the yen while Japan kept its rates near zero had a reason: until March 2024 the Bank of Japan’s rate was below zero, and then 0 to 0.1% (Bank of Japan, March 2024). An extreme means more once that reason starts to change.
• Reading the wrong group. Leveraged funds in the TFF report for currencies. Gold and other commodities use managed money in the Disaggregated report (COT for Gold Traders).
• Using COT to time a trade. Tuesday’s positions arrive on Friday, and the report shows who is positioned, never when they will leave. It suits a horizon of weeks, and by the time it confirms a turn, price may have done much of the moving.