In short: Every week the CFTC publishes who is holding what in COMEX gold futures, sorted into categories. The figures below come straight from the CFTC and update themselves when the new report is released. What follows is how to read them, and — more importantly — what they cannot tell you.
How the CFTC sorts traders
Anyone holding a position above a reporting threshold has to be classified. The Legacy report uses three categories.
Commercial traders are the ones with a business in physical gold: miners, refiners, fabricators, and the dealers and banks that finance and warehouse the metal. They use futures to hedge exposure they already have.
Non-commercial traders are, broadly, funds and speculators. They hold gold futures to profit from the price, not because they handle the metal.
Non-reportable positions are everyone below the reporting threshold, lumped together. Every futures contract has a buyer and a seller, so all the net positions in the market must sum to zero. That is why the commercial and non-commercial nets do not mirror each other exactly: the difference is the non-reportable category.
Spreading is the fourth column in that table, and it appears only for non-commercials. It counts traders holding an equal long and short position at the same time, usually in different contract months. Because the two sides cancel, spreading does not show up in the net column at all.
The net figure in the bottom row adds all three categories together, and it should read zero. If it ever does not, something is wrong with the data on this page — not with the market.
Why the weekly change matters more than the level
A net figure of, say, 225,000 contracts tells you where positioning stands. It does not tell you how it got there, and the two are not the same thing.
A net position can fall for two entirely different reasons. Long positions can be closed, which means money is leaving the trade. Or short positions can be opened, which means new money is betting the other way. Both show up as the same number in the net column, and they are not the same event.
That is why the table above splits the change into longs and shorts for each category. Most sites report only the net, and lose the distinction.
Open interest is the third piece. It counts every contract still open across all categories. If positioning moves and open interest rises, new contracts are being created — money is coming in. If positioning moves while open interest falls, existing positions are being closed out. Same price move, different thing happening underneath.
Because the report gives the long and short columns separately, the week sorts itself into one of four readings without any guessing. The panel above works this out each week from the numbers alone.
| Net positioning | Driven mainly by | What that describes |
|---|---|---|
| Falls | Long positions closing | Money stepping back from the trade. |
| Falls | Short positions opening | Money arriving on the other side of it. |
| Rises | Long positions opening | Fresh money on the long side. |
| Rises | Short positions closing | Existing bets unwound, not new ones placed. Covering. |
Open interest sits alongside this rather than inside it, and the distinction is worth holding on to. The net above belongs to one category; open interest counts every category at once. A week where non-commercial longs close while open interest still rises is perfectly ordinary — it just means someone else was opening positions. Reading a rise in the total as evidence about what the funds did is the kind of shortcut that produces confident nonsense.
One further caution. Open interest also moves with the contract expiry cycle, not only with money entering and leaving, so a single week in isolation is weak evidence. It is repetition across several weeks that carries information.
What the percentile adds, and what it cannot tell you
A net figure on its own has no scale. Two hundred thousand contracts net long sounds like a lot, and whether it actually is depends entirely on what this market normally looks like. The percentile above answers that one question and no other: of the last five years of weekly readings, what share sat at or below this week’s?
The method is deliberately plain, so you can reproduce it. Take the weekly net for the last 260 reports, count how many are at or below the current one, divide by 260. If this week is the highest or the lowest of the window, the panel says so rather than quoting a percentile, because “100th percentile” hides the more useful fact that it is a five-year extreme.
Two things it is not. It is not a signal. A position in the 95th percentile tells you the trade is crowded; it does not tell you it is about to unwind, and crowded positions have stayed crowded for months at a time. Anyone who presents a high percentile as a reason to expect a reversal is adding a claim the data does not contain.
And the window is a choice. Five years was picked because it covers more than one cycle while still reflecting the current structure of the market. A one-year window would make this week look far more extreme; a twenty-year window would flatten it. The figure moves with the window, which is why the window is stated on the panel instead of being left implicit.
The two percentiles are also not independent. Every long position has a short on the other side, so when the funds are at a historical extreme the commercials are usually near the opposite one. That is arithmetic, not confirmation.
How to check it yourself
Every figure on this page comes from one public file. Go to the CFTC’s Commitments of Traders page, find Current Legacy Reports, open Metals and Other in either format, and look for GOLD - COMMODITY EXCHANGE INC.
The contract code is 088691. The report is released every Friday at 3:30 p.m. Eastern time, covering positions as of the previous Tuesday — so the newest figure available is always three days old by the time it is published. When a US federal holiday falls in the week, the release slips by a day; the CFTC publishes the schedule in advance.
One boundary worth keeping in mind: this report covers COMEX futures and nothing else. A large share of the world’s gold changes hands over the counter in London, and there is no equivalent weekly report for it.
The panel above uses the Legacy, futures-only report. There is also a futures-and-options version, which includes options positions converted to a futures equivalent, and a Disaggregated report that splits the non-commercial category further into managed money, swap dealers and other reportables. They are different views of the same market and their numbers will not match each other. If you are comparing this page against another source, check which report that source is using before concluding one of them is wrong.
Related reading
Positioning is one input. The macro calendar is another, and the two interact: a crowded position meeting an unexpected figure is a different event from a neutral one meeting the same figure. The Four Releases That Move Gold covers the four scheduled releases that do most of that work.
Sources
- US Commodity Futures Trading Commission, Commitments of Traders — Legacy report, futures only. Contract: GOLD, Commodity Exchange Inc., code 088691.
- Figures on this page are retrieved live from the CFTC’s public reporting API each time the page is loaded. They are not stored or edited here.
- CFTC data is a work of the United States federal government and is in the public domain.
Educational content only. This is not financial advice or an investment recommendation. The author is not a registered financial adviser. Trading leveraged instruments carries risk of loss.