How to read the COT report
A walk through one real row of this week's data — gold — column by column.
To read the Commitments of Traders report, take each trader category's net position (longs minus shorts, with spreads excluded), compare it to the prior week, and place it inside its own history. As of the Jul 21, 2026 report, non-commercial traders in COMEX gold were net long 183,910 contracts against a commercial net of -213,199, with total open interest of 383,368 contracts. The numbers are recorded as of Tuesday's close and published by the CFTC on Friday at 3:30 p.m. ET.
Every figure below refreshes automatically each release. Released Jul 24, 2026; next report expected Jul 31, 2026.
The row we'll read
Here is the current legacy-format snapshot for gold, exactly as it appears on the gold COT page:
| Field | Value (report dated 2026-07-21) |
|---|---|
| Non-commercial net | 183,910 contracts (net long) |
| Commercial net | -213,199 contracts |
| Open interest | 383,368 contracts |
| COT index, 26-week | 58.0 |
| COT index, 52-week | 26.4 |
| COT index, 3-year | 46.1 |
| Z-score, 3-year | -0.49 |
Each of these is derived from the same CFTC source row. Reading them in order is, in effect, the whole method.
Step 1 — net position: long minus short, spreads out
The CFTC reports each category's long contracts, short contracts, and (for
reportable categories) spreading — positions where the same trader is
long one contract month and short another. Net position is simply
long − short; spreads cancel themselves and are excluded. Gold's
non-commercial net of 183,910 means that, summed across every large
speculator the CFTC can identify, longs exceed shorts (or the reverse, if negative) by
that many contracts. Because every long is someone's short, the category nets in any
market sum to approximately zero — a useful sanity check, and one this site verifies
on every ingest (see methodology).
Step 2 — the weekly change
A single week's level tells you little on its own. The report's most-read numbers are the changes: did speculators add or shed exposure since last Tuesday? A large one-week swing in a category's net — especially against a flat price — is the kind of fact the weekly movers page surfaces across all 107 published markets at once. Change in open interest matters alongside it: a net position that grew while open interest shrank means the move came from the other side closing, not from new conviction.
Step 3 — scale by market size: percent of open interest
Raw contract counts are not comparable across markets or across decades. Dividing the net by open interest — for gold this week, 183,910 over 383,368 — turns the position into an intensity: what share of the whole market one category's net stance represents. This is the normalization behind the bias-corrected COT index variant this site publishes, and it is the number to use when comparing, say, gold positioning against crude oil (non-commercials there currently net long 81,689 contracts on open interest of 1,864,487).
Step 4 — place it in history: the COT index and z-score
The COT index maps the current net onto its trailing range: 0 is the most net-short reading of the window, 100 the most net-long. Gold's non-commercial index right now reads 58.0 on 26 weeks, 26.4 on 52, and 46.1 on three years — three windows because a position can be extreme against six months of history and unremarkable against three years. The 3-year z-score (-0.49 this week) asks a related question in different units: how many standard deviations from the 3-year mean. Readings near the edges of these scales describe a crowded market — a condition, not a signal.
Step 5 — read the structure, not just one line
Commercial and non-commercial nets normally sit on opposite sides — in gold this week, -213,199 against 183,910. That mirror image is structural, not sinister: commercials hedge inventory and future production, so as speculators buy, hedgers are the ones selling to them. What analysts actually study is when the gap between the two stretches toward the edge of its historical range, or when it starts closing. The screener and heatmap exist to show that structure across every market in one view.
Common misreads
- Treating Friday's release as Friday's positions. The data is as of Tuesday's close — three days old at publication. Fast weeks can look very different by the time you read the report.
- Confusing net with gross. A category can be modestly net long while holding enormous positions on both sides. The gross long and short columns, and the spreading column, are in the full CSV.
- Ignoring open interest. The same 50,000-contract net means something different in a market of 100,000 contracts than in one of 500,000.
- Reading an extreme as a timing device. Historically, positioning extremes have often persisted for weeks or months. The index says where positioning sits, not what price does next.
- Mixing report variants. Futures-only and futures-and-options combined are different files with different totals; comparisons only work within one variant.
Where the numbers come from
Firms above CFTC reporting levels file positions daily; the Commission compiles the Tuesday snapshot and publishes it Friday at 3:30 p.m. ET in flat files and its public data API (holiday weeks shift — see the release schedule). FuturesBench ingests both channels, cross-checks them, and updates this week's report page and every market page within about a minute of the drop.
- What is the most important number in the COT report?
- There is no single one, but the most-quoted is the non-commercial (large speculator) net position and its week-over-week change. Analysts who work with the report usually read it against open interest and against its own history via a COT index rather than in isolation.
- How current is the COT report when it's released?
- Three days old by design: positions are as of Tuesday's close, published Friday at 3:30 p.m. ET. The lag is for compiling and reviewing the underlying filings.
- Do spread positions count in the net?
- No. Spreading — the same trader long one month and short another — is reported as its own column and cancels out of the net by construction. Net is outright longs minus outright shorts.
- Why don't the category totals match open interest?
- Open interest counts each contract once, while the categories split the long and short sides separately, with spreads counted on both. Summing all category longs (or all shorts, plus spreads) reconciles to open interest; summing nets gives approximately zero.
All figures above are from the CFTC report dated 2026-07-21 and update automatically each week. Positioning data describes conditions; it is not a buy or sell signal. See the disclaimer.