COT report for gold: how to read positioning and test it as a rule

Contents

1. How do you read the COT report for gold?
2. How often does the COT report update, and what is the release lag?
3. Managed money vs commercial in gold: who is on which side?
4. Where do you get gold COT data?
5. What do record net longs in gold imply?
6. Can the COT report be used in a backtest?
7. Frequently asked questions
8. References

How do you read the COT report for gold?

In the COT report, gold positioning is one weekly table: how many COMEX gold futures contracts each trader group held net long or net short as of Tuesday. Read the managed-money net position against the commercial net position, and compare each with its own history. It tells you how crowded a trade is, not where price goes next.

The Commitments of Traders (COT) report is a weekly publication of the U.S. Commodity Futures Trading Commission (CFTC) that counts the open futures positions of large traders in each market, grouped by what kind of business they run. A net position is the number of long contracts a group holds minus the number of short contracts it holds. Think of a theatre: the report does not tell you whether the play is good, it tells you how many seats are already taken and who is sitting in them. If every seat is full, there is nobody left to buy a ticket, but the play can still run for months.

For gold, three numbers carry most of the information. The managed-money net position shows how far funds are leaning. The commercial net position shows how much producers and dealers have hedged against them. The week-over-week change shows whether the lean is still growing or already shrinking. The method is the same for every futures market; the Commitment of Traders report guide for systematic futures traders covers the general case, and this article covers what is specific to gold.

How often does the COT report update, and what is the release lag?

The report is compiled as of Tuesday's close and published on Friday afternoon, so a rule built on it can act for the first time at the next session open, three trading days after the positions were counted. The CFTC releases the weekly report at 3:30 p.m. Eastern time on Friday, which is early Saturday morning in Taiwan. In weeks with a US holiday the release moves back by a day or more; the CFTC publishes the shifted dates in advance.

The lag matters for two reasons. First, the positions you read on Friday are already three days old, and in a fast week they can have changed. Second, the first bar on which a rule may use the Friday value is the Monday daily bar (the Sunday evening open in US time). A value attached to any earlier bar is information the trader did not have yet.

Because the data arrives once a week with a lag, COT positioning is a tool for holding periods of weeks, not hours. A trader who wants something that updates every bar is better served by the price-based tools discussed in what testing shows about the best indicator for intraday trading and best indicators for day trading and where each fails. COT data can still filter which weeks an intraday system trades, but it cannot time an entry inside the day.

Managed money vs commercial in gold: who is on which side?

Managed money tends to follow the trend and is most long near price highs, while commercials hedge against it and are most short at the same time, so the two lines are near mirror images and must be read together. Managed money is the CFTC category for hedge funds, commodity trading advisors and other money managers who trade futures for clients. Commercials are the firms that handle physical gold: miners, refiners, jewellers and bullion dealers, who use futures to lock in a price for metal they will produce or need.

In the Disaggregated report the commercial side is split into two categories: "Producer/Merchant/Processor/User" (the physical trade) and "Swap Dealers" (banks that hedge over-the-counter contracts). Many readers add them together to get a single commercial figure. The older Legacy report shows only "Commercial" and "Non-commercial", which is simpler but mixes funds with other large speculators.

Question	Managed money	Commercial (producer/merchant, swap dealer)
Who they are	Hedge funds, CTAs, money managers	Miners, refiners, jewellers, bullion dealers, banks hedging client trades
Why they hold gold futures	To profit from price moves; often trend-following	To fix a price for physical metal they will sell or buy
Typical position near a price high	Large net long	Large net short
Typical position near a price low	Small net long or net short	Small net short or close to flat
How to read an extreme	The crowd is in; new buyers are scarce	Hedgers have sold a lot forward; they are not predicting, they are insuring

The mirror image is not a coincidence. Every futures contract has one long and one short, so the net positions of all groups add to zero. When funds buy, somebody must sell, and in gold that somebody is usually a producer locking in a price. A record commercial net short therefore does not mean that commercials expect a fall; it means they have sold a lot of future production into a rally. Read the fund side for how crowded the speculative trade is, and the commercial side as confirmation that the other side of those contracts is hedging, not betting.

Where do you get gold COT data?

Gold COT data comes from the CFTC's own Commitments of Traders pages, as a weekly report you can read on screen and as historical files you can download, and the Disaggregated futures-only report is the one that separates managed money from commercials. Gold appears in the market list under its exchange, the Commodity Exchange (COMEX), which is part of CME Group. The Disaggregated files start in 2006; the Legacy files go back further but only show the commercial and non-commercial split.

1. Open the CFTC Commitments of Traders page. It lists the current week's reports and a link to historical data.
2. Choose the Disaggregated report, futures only. A "futures and options combined" version also exists. Pick one and keep it for the whole study, because the two series differ in level.
3. Find the gold row. It is listed under the COMEX exchange name. Check that you have the main gold contract and not a micro or a spread product.
4. Download the historical file. The yearly files contain every market; filter them to gold and keep the as-of date, open interest, and the long and short columns for each trader category.
5. Add a release-date column. For each Tuesday as-of date, write the Friday on which it was published, and adjust for the holiday weeks the CFTC lists. This column, not the as-of date, is what you line up with price bars.

Open interest is the total number of contracts outstanding; it is worth keeping because a net position of a given size means more in a small market than in a large one, so many traders divide the net position by open interest before comparing years. Price data for the same period, and the contract details you need to turn points into dollars, are covered in gold futures backtesting: specs, margin and trading hours.

What do record net longs in gold imply?

A record managed-money net long means the crowd is already positioned and new buying power is thinner, which raises the risk of a sharp pullback but says nothing about timing, so it works as a condition or filter rather than an entry signal. Picture a crowded elevator: one more person can still step in, and the car can still go up several floors, but when it empties, it empties fast. The record tells you the car is full. It does not tell you which floor it stops at.

"Record" is a weak word for a rule because a record can be broken the next week and because gold open interest has grown over the years. A better form is a position measured against its own recent range. The common way to do this is the COT index:

COT index = (current net − lowest net in the lookback) ÷ (highest net in the lookback − lowest net in the lookback) × 100

Here current net is this week's managed-money net position, and the lowest and highest net are the smallest and largest weekly values over a fixed lookback, often one to three years. The result runs from 0 (the most bearish positioning of the lookback) to 100 (the most bullish). A reading above 90 says positioning is near the top of its recent range, whatever the absolute number of contracts is.

Used this way, positioning sits next to the chart-based tools rather than replacing them: the list in technical analysis tools and indicators you can test covers what a price rule can measure, and the COT index adds something those tools cannot see, which is who holds the contracts.

Can the COT report be used in a backtest?

Yes, if each weekly value is aligned to the bar after its Friday release, but because our run uses a price-based N Day rule on ES daily bars as a stand-in rather than COT data on gold, its result shows what a simple position-direction rule does after costs, not what COT positioning earns. The alignment is the part that goes wrong most often. Look-ahead bias is the error of letting a backtest use information that was not yet available at the time of the trade, and a weekly series with a three-day publishing lag is a classic source of it.

1. Build the weekly table with a release-date column, as in the previous section.
2. Join it to daily gold bars on the release date, so that the value first appears on the Monday bar after the Friday release, and carry it forward unchanged until the next release.
3. Compute the COT index from the carried value, using a fixed lookback, so that the index on any bar uses only published weeks.
4. Run the price rule twice: once alone, once with the COT condition as a filter on new entries.
5. Compare average trade after commission and slippage, the number of trades, and the drawdown of the two runs. A filter that improves the average trade but removes most trades needs a longer history before it means anything.

The one-bar check described in look-ahead bias in backtests: common leaks and a one-bar check catches the most common mistake: shift the COT series one bar later and see whether the result collapses. If it does, the original run was reading the future.

We have not run COT data through the engine, so the run below tests a stand-in: a price-based N Day rule, built from standard indicators, on ES daily bars, long and short, with one contract, from 2019-01-02 to 2025-06-30. AlphaPrime (Valynox) tried 1,024 candidates in that search and kept one, chosen for its return against its drawdown. The box shows that one candidate. Picking the kept one after the fact flatters it: among many tries, the best will look good partly by chance, as Bailey et al. (2014) show.

Read the box as a description of what a position-direction rule looks like, not as a result of COT data and not as something a reader can expect. Three things stand out. Time in market is 93.14%, so the rule is almost always holding a position and simply flips between long and short; a COT filter would cut that number, because it removes entries in crowded weeks. The max drawdown of $68,465 is more than two thirds of the net profit of $95,582.5, which is the shape many trend-following rules show on index futures. And 44 trades over about six and a half years is a small sample, so the win rate of 65.91% and the profit factor of 2.035 could move a long way if a few trades were different. A gold run with a COT filter would need the same checks: the walk-forward windows and pass thresholds in strategy validation are where a filter like this shows whether it held up on weeks it was not tuned on, and they stop the later checks as soon as a candidate fails one.

Your next step is a two-hour job: download the Disaggregated futures-only file for gold, add the Friday release column, compute a one-year COT index, and run one simple price rule on daily gold bars twice, once plain and once with "no new longs above 90". Write down the average trade after costs for both before you look at the equity curves.

Frequently asked questions

How often is the COT report updated?

Once a week. The CFTC counts positions as of Tuesday's close and publishes the report on Friday at 3:30 p.m. Eastern time, which is early Saturday morning in Taiwan. In weeks with a US holiday the release is delayed, and the CFTC posts the changed schedule in advance. The first bar a rule can act on is the next session open after the release.

What does a record net long in gold mean?

It means managed money, the fund category in the Disaggregated report, holds more long contracts minus short contracts than at any earlier point in the history you are looking at. This marks a crowded trade with fewer new buyers left, which raises the chance of a sharp pullback when positions are closed. It does not say when that happens; price has often kept rising for weeks with positioning at an extreme. Use it as a filter on new trades, not as a sell signal.

Where can I download the COT report?

From the CFTC's Commitments of Traders pages, which offer the current weekly report and historical files by year. For gold, choose the Disaggregated report (futures only, or futures and options combined) and find the gold row under the COMEX exchange name. The Disaggregated files begin in 2006; the older Legacy report goes back further but only separates commercial from non-commercial traders.

Can the COT report be used for backtesting?

Yes, as a weekly filter on a price-based rule. Attach each weekly value to the first daily bar after its Friday release, not to the Tuesday it describes, and carry it forward until the next release. Then run the price rule with and without the COT condition and compare the average trade after costs and the number of trades. A test that shifts the series one bar later and still holds up is a good sign that no future information leaked in.

What is the difference between the Legacy and Disaggregated COT reports?

The Legacy report splits large traders into Commercial and Non-commercial. The Disaggregated report splits them into Producer/Merchant/Processor/User, Swap Dealers, Managed Money and Other Reportables. For gold, the Disaggregated report is more useful because it isolates the fund category (managed money) from banks and other speculators, so its net position is a cleaner measure of how crowded the speculative trade is.

References

• U.S. Commodity Futures Trading Commission. Commitments of Traders. https://www.cftc.gov/MarketReports/CommitmentsofTraders/index.htm
• U.S. Commodity Futures Trading Commission. Explanatory Notes: Commitments of Traders. https://www.cftc.gov/MarketReports/CommitmentsofTraders/ExplanatoryNotes/index.htm
• Bailey, D. H., Borwein, J., López de Prado, M., & Zhu, Q. J. (2014). Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance. Notices of the American Mathematical Society, 61(5), 458–471.