Reading COT Positioning
The three columns
The COT panel on the fundamentals board shows each currency's futures positioning. Three numbers do the work, and reading any one alone is how people get misled.
| Column | Definition | What it answers |
|---|---|---|
| Net position | Speculative long contracts minus short contracts | Direction and size. Positive = the crowd is betting up. This is the only column that carries direction. |
| Net/OI | Net position ÷ total open interest | Concentration. Divides out market size so a 50,000-contract net means different things in a big and a small market. |
| Idx52 | Percentile of today's net between its own 52-week high and low | Abnormality. 0 = most short in a year, 100 = most long. It is a position within its own history and carries no direction on its own. |
The common mistake is reading Idx52 as bullish or bearish. It is neither. An index of 95 on a currency whose net position is deeply negative simply means "less short than usual this year" — still short, just less so.
Fuel gauge, not momentum meter
Picture 174,000 contracts of net short in a single currency. That is not a prediction that the currency falls. It is an inventory of people who must eventually buy it back. If something forces them to start covering, the buying has no natural counterparty — everyone who wanted to be short already is — and the move can be violent. The mechanics are margin mechanics: as price rises against a leveraged short, equity erodes toward the liquidation threshold, and the margin call calculator shows how quickly that point arrives at different leverage levels.
But nothing in the number says when. Positioning can sit at a 52-week extreme for months while price drifts sideways. This is the single most important thing to internalize: crowding is potential energy, and potential energy needs a trigger.
The trigger, empirically, is the weekly change (Δ) flipping sign. A net short that has been growing for eight straight weeks and then posts its first positive week is a different object from one that is simply large.
The asymmetry that matters
A crowded position is never, by itself, a reason to trade against it.
Everyone who has looked at COT has had the thought: positioning is at an extreme, so I will fade it. Tested across our sample, that alone is not profitable. Extremes without a change in direction persist, and fading a trend because it is popular is simply being early — which is indistinguishable from being wrong.
The corollary is more useful and less obvious: selling an asset that speculators are already maximally short is a bad idea even when the technicals look right. You are joining a fully-loaded boat. In our testing this condition alone removed the one-week edge from otherwise valid setups, which is why the site treats it as a veto rather than a minor negative.
What our backtests found
These come from our own study of roughly two years of weekly COT data across the major currencies, plus a 22-year study of the dollar index. They are statistical tendencies with modest sample sizes, not laws — read the t-statistics as a measure of how confident the sample allows us to be, not as a guarantee.
| Pattern | Condition | Result |
|---|---|---|
| Ignition | Idx52 ≤ 10 and Δ flips positive | ≈61% hit rate over the following week — the highest-conviction setup we found |
| Late-stage pile-on | Positioning accelerating in the same direction it already leans | −1.45% over the next 8 weeks (t = −2.8). Piling in late is penalised. |
| Divergence refuels | Price and positioning flow moving apart | +1.20% over 8 weeks (t = 2.2) — the trend tends to continue, not reverse |
| Anti-squeeze veto | Selling an asset with Idx52 ≤ 10 | Removes the one-week edge entirely |
Note what the second and third rows imply together: positioning that is accelerating is a late-cycle warning, while positioning that disagrees with price tends to feed continuation. Neither is the naive "extreme means reversal" story.
The dollar index reads backwards
The panel includes DX, the ICE dollar index future, and it is marked differently from every other row — with ⇈/⇊ for momentum rather than the ⚠ used for reversal fuel. That is deliberate.
Across 22 years and 1,148 weeks, we found the dollar index behaves the opposite way to single currencies. After positioning enters its most-bullish bucket, the dollar index averaged +1.07% over the following 8 weeks (t = 5.18); after its most-bearish bucket, −0.58%. Of 17 events above the 90th percentile, 13 (76%) were followed by a stronger dollar. Extreme dollar positioning is a momentum signal.
The likely reason is structural. A crowded yen short has one counterparty and one squeeze point. A crowded dollar long is dispersed across seven currencies simultaneously, so there is no single forced-buying event to trigger. Dollar extremes also tend to coincide with macro regime changes — 2008, 2014, 2022 — and regimes have persistence.
One consequence worth stating plainly: at the dollar level, the "ignition" pattern above does not exist. We tested it directly and the effect at a one-week horizon was indistinguishable from zero. The dollar's usable effect is a slow drift of roughly +0.13% per week while positioning stays extreme, not a reversal trigger — and a drift that small only matters through accumulation, which the compound interest calculator makes concrete in a minute: a fraction of a percent per week is real money over a quarter and noise over a day. A longer write-up of that study is on jesselau.com.
When it updates
The CFTC publishes the report on Friday afternoon US time, reflecting positions as of the previous Tuesday. That built-in three-day lag is why COT is context rather than a timing tool. The site fetches on a staggered schedule Friday evening and Saturday morning UTC, so the panel is current by Saturday.
Clicking an instrument code opens CME's official daily volume and open-interest page, which lets you check the unwind between weekly reports: price rising while open interest falls means old positions are closing — covering — while price rising with open interest also rising means new money is arriving.
This week's numbers: see the latest COT report table — net position, weekly change, Net/OI and the 52-week index for every currency, metal and Bitcoin, with two-year charts, refreshed every Saturday.