COT Report for Natural Gas Traders: Spreading, Storage, and What the Record Actually Shows

Key takeaways

• NYMEX Natural Gas is the largest energy contract on the board after WTI: 1,762,883 contracts of open interest in the report dated 25 August 2026, against 2,482,663 in NYMEX WTI, 349,514 in RBOB gasoline and 284,157 in NY Harbor ULSD.
• Managed Money holds 605,822 spreading contracts, a spread ratio of 49.8%. Roughly half of that cohort's exposure is calendar structure rather than a directional view, which is the single biggest reading error in gas.
• Positioning has covered hard. Managed Money net went from -126,545 on 4 August 2026 to -71,496 on 25 August, a 55,049-contract move in three reports, taking the 156-week COT Index from 12.0 to 32.7.
• Working gas in storage was 3,184 Bcf for the week ending 21 August 2026, above the five-year average for that week and inside the band, not at either edge.
• Across 657 weekly observations from 2014 to 2026, storage above the seasonal band preceded a mean 8-week gain of +17.9% (n=93) and storage below it -4.9% (n=75). That is the opposite direction to the folk rule, and neither cell clears our publication bar once overlapping windows are accounted for.

Introduction

Natural gas is the market where fundamental data is most abundant and least differentiating. The EIA publishes storage every Thursday, the five-year band is on their own free dashboard, and a dozen desks put out commentary within the hour. Nothing about the storage number is proprietary.

What almost nobody puts next to it is the positioning of the people who have to react to it, and the measured record of what has followed that combination. That is the whole of this guide: which cohorts to read in NYMEX gas, why the headline net figure understates how much of the book is not directional at all, how to place storage against its own seasonal band, and what our own history says when the two are read together.

The Contract

The COT series to watch is NAT GAS NYME (CFTC code 023651), the Henry Hub natural gas futures contract at 10,000 million British thermal units per contract. It is the price the physical gas trade, the LNG export contracts and the power sector all reference, and it settles against Henry Hub in Louisiana.

Two things make it different from every other energy contract on the board.

It is enormous relative to the refined products. Open interest of 1,762,883 contracts on 25 August 2026 is five times RBOB gasoline and six times ULSD, and second in energy only to NYMEX WTI at 2,482,663. Positioning extremes in gas involve genuinely large books.

It is the most weather-driven contract we track. Heating and cooling degree days move the demand side week to week in a way that no refined product experiences. Positioning is a statement about a weather forecast at least as often as it is a statement about supply.

There are also large ICE Henry Hub look-alike contracts, and cash-settled penultimate and last-day financial contracts, filed separately by the CFTC. They are worth knowing about, but the NYMEX physically-delivered contract is the one whose positioning history goes back furthest and whose extremes are comparable across years.

Which Cohorts Matter, and the Spreading Trap

The Disaggregated report splits gas into four reportable groups. From the report dated 25 August 2026:

Read Managed Money for crowding, and read the spreading column before you do. Managed Money holds 605,822 spreading contracts, which is a spread ratio of 49.8%. A spread position is long one delivery month and short another. It expresses a view about the shape of the curve, about winter against shoulder season, not about whether gas goes up. When someone says the funds are short gas, the honest version is that the directional part of their book is net short 71,496 while a book roughly the same size again is sitting in calendar structure.

Swap Dealers are the mirror, and in gas they are large. Net long 208,734 with a long-to-short ratio of 12.81. That is the intermediation of producer hedging and structured flow, not a view.

The Producer/Merchant line is unusually small. Net short 18,811 against 1.76 million contracts of open interest. In crude or in the products the physical trade dominates the short side. In gas a great deal of producer hedging is done through swap dealers instead of directly, which is why the Producer line looks thin and the Swap Dealer line looks fat. Read them together or you will misjudge how much of the gas market is hedged.

Other Reportable is not noise here. Net short 129,197 with 209,039 spreading. In most markets this bucket can be ignored. In gas it is large enough to matter.

The Current Reading

From the CFTC report dated 25 August 2026, Managed Money net position in NYMEX gas:

The 4 August print was the crowded end of this move: a z-score of -1.71 and a COT Index of 12.0, meaning the net position sat in the bottom eighth of its own three-year range. Three reports later the z-score is back at zero.

Open interest rose throughout. From 1,719,031 on 4 August to 1,762,883 on 25 August, up 2.6% while the shorts were covering. Covering into rising open interest means new positions replaced the ones that left, rather than the market simply shrinking. That is the difference between an unwind and a handover, and it is visible in the report before it is obvious on the chart.

The state COTInsight labels here is Flip Zone: the net position has crossed back through its own mean and neither side is crowded. It is the least interesting positioning state in the taxonomy, and saying so is more useful than manufacturing a read from it.

Storage Against the Band

The fundamental half is EIA working gas in the Lower 48, published Thursdays. For the week ending 21 August 2026, from the series COTInsight stores:

• Working gas, Lower 48: 3,184 Bcf. The five-year band for that calendar week runs 2,640 to 3,334 Bcf with an average of 3,098. Gas is inside the band and above its average, by 4.9% against the average of the same week in the five prior years.
• Salt-dome storage, South Central: 248 Bcf, against a band of 185 to 272 and an average of 245.2. Also mid-band.

Salt is worth tracking separately because it cycles fast. It can be injected and withdrawn many times a season, so it is the part of the storage picture that responds to a heat wave inside a week. Non-salt storage moves on the season.

What the band is, precisely. It is the same calendar week in the five most recent years, so it moves with the recent history rather than with a fixed target. It states where inventories sit relative to their own recent seasonal norm. It does not state whether they are adequate, and it is not a forecast.

What the Measured Record Says

This is the part most gas commentary skips. We ran the pairing point in time over 657 weekly observations from 7 January 2014 to 18 August 2026, with the seasonal band at each date built only from the five prior calendar years, the storage reading being the newest one that would already have been public on the COT release Friday, and entry at the first Henry Hub spot print on or after that Friday. Horizon is 8 weeks.

Baseline for all 649 scorable weeks: mean +4.21%, median -0.81%, higher 47.9% of the time. The mean and the median disagree because gas returns are violently skewed, which is itself a warning about reading any mean in this market.

Storage alone, 8 weeks forward:

Read that twice. High storage preceded strong forward returns and low storage preceded weak ones, which is the reverse of the bullish-draw, bearish-build language the industry uses out of habit. The mechanism is not mysterious: inventories are high because the price has been low, and the price is what mean reverts. It is still the opposite of what the label implies, which is why Energy Radar describes inventory states neutrally, as above, inside or below the band, and never as bullish or bearish.

Positioning alone, 8 weeks forward: crowded long (COT Index at or above 80) n=92, mean -4.87%, higher 37.0% of the time. Crowded short (COT Index at or below 20) n=117, mean +6.64% but higher only 40.2% of the time, another skew warning. Neutral n=440, mean +5.45%.

The season split, which is where gas differs from every other energy market. Splitting the same 656-observation panel by season, and defining high and low storage as more than one standard deviation above or below the mean of the same calendar week in the five prior years, 8-week median returns:

Salt storage sharpens the withdrawal-season version: low salt inventories preceded a median -18.27% over the following 8 weeks (n=36). Salt is the fast-cycling part of the storage picture, so it registers a stress the aggregate number smooths away.

And now the honest part. Weekly observations of an 8-week forward window overlap eight to one, so a raw n of 93 is not 93 independent trials. Correcting for that, and applying our publication rule of at least 30 independent observations and a t-statistic of at least 1 in absolute value:

No natural gas cell clears the bar. Fourteen candidate cells were tested across gas, gas salt storage, gasoline, distillate and WTI, and two qualified. Neither is in gas.

That is a real finding and we publish it as one. Sixteen years of weekly gas data, split by season and by storage state, does not produce enough independent observations to support a claim about what follows. Anyone quoting a confident hit rate for a storage-plus-positioning setup in gas is quoting a number their sample cannot carry.

How to Use It Anyway

The absence of a validated edge does not make the data useless. It changes what the data is for.

Use positioning as a risk statement. A COT Index of 12 says the marginal seller is scarce and that a squeeze, if one starts, has room to run. It does not say one will. The move from -126,545 to -71,496 in three weeks is what that looks like when it happens.

Use spreading as a confidence discount. When half the Managed Money book is in calendar spreads, treat the directional signal as covering half the book it appears to cover.

Use storage for context, and state it neutrally. Where inventories sit against their own seasonal band tells you what the market has already had to absorb. Combined with the record above, it should stop you from calling a build bearish.

Respect the weather clock. Gas positioning turns on forecasts that change in days. Whatever the report said on Tuesday, the temperature outlook for the next two weeks is doing more to the price by the time you read it on Friday.

Where COTInsight Fits

The natural gas market page carries the current positioning read. Inside the app, gas sits on the same board as around 350 other weekly-reporting markets, ranked by extremity, with the z-score, 156-week COT Index, regime state, open-interest trend and divergence check on one screen.

Ultimate adds Energy Radar, which is where the storage side of this article lives: the gas panel puts working gas, the regional and salt breakdown, the seasonal band chart and the positioning gauge in one place, with the base-rate table above shown with its sample sizes and its failures marked as failures. Ultimate also carries the historical outcome statistics per z-score bucket, the twenty-year archive, the AI commentary, and the TradingView indicator so the same z-score and COT Index sit on your gas chart rather than in another tab.

Start with how to read Energy Radar, or the field guide to every number on the panel. Related energy markets: crude oil, RBOB gasoline, NY Harbor ULSD. Pricing is here.

Frequently Asked Questions

Which CFTC contract should I use for natural gas?

NAT GAS NYME, code 023651, the physically-delivered Henry Hub contract. ICE runs large look-alike contracts and there are cash-settled financial versions, all filed separately. The NYMEX contract has the deepest comparable history.

Why is the Managed Money net position misleading in gas?

Because 605,822 of that cohort's contracts on 25 August 2026 were spreading positions, a spread ratio of 49.8%. Spreads express a view on the shape of the curve, not on direction, so the net figure describes a smaller directional book than the headline suggests.

Does high gas storage mean lower prices?

Not in the measured record. Over 657 weekly observations from 2014 to 2026, storage above the five-year seasonal band preceded a mean 8-week gain of +17.9% (n=93) and storage below it preceded -4.9% (n=75). The relationship runs the opposite way to the common label, and once overlapping windows are corrected for, neither cell is statistically strong enough for us to publish as a base rate.

Why is the Producer/Merchant net position so small in natural gas?

Because a large share of producer hedging reaches the market through swap dealers rather than directly. Swap Dealers were net long 208,734 contracts on 25 August 2026 with a long-to-short ratio of 12.81. Read the two lines together.

When is the natural gas COT data updated?

Every Friday at 3:30pm Eastern, reflecting positions as of the preceding Tuesday. EIA working gas in storage is published separately, on Thursdays at 10:30am Eastern.