Energy Radar Field Guide: Every Number on the Panel, and How to Read It

30-second answer: Energy Radar puts two independent measurements on one screen, on two panels. The positioning axis is the COT Index, a 0-100 rank of where Managed Money's net position sits inside its own trailing 3-year range. The fundamentals axis is the 5-year seasonal band, which asks where a stock or storage level sits against the range it has occupied in that same week of the year across the last five years. Every number states a position. It does not state a direction, except in one place: a measured base rate that cleared a stated sample-size and significance bar, shown with its own sample attached. This guide covers every field on both panels.

This is the reference companion to How to Read Energy Radar, which is the shorter introduction. Here we go field by field.

The two panels

Energy Radar renders on two market pages, and each panel reads its own market's positioning. The WTI panel uses the WTI COT Index; the natural gas panel uses the natural gas COT Index. Neither borrows the other's. Reading one market's positioning against another market's fundamentals would be a fabricated number, so a panel either has its own positioning input or it does not render at all.

Forty-five series in total.

The positioning axis: the COT Index

The first input is Managed Money's net position, converted into the COT Index: a 0-100 rank of the current net position against its own trailing 3-year range. 0 means the most net-short it has been in three years; 100 means the most net-long.

The panel shows it as a gauge, with the two crowded thresholds marked at 20 and 80. Those are the same thresholds used everywhere else on COTInsight, so the gauge cannot disagree with the COT Index shown on the rest of the dashboard.

The panel words it in four bands:

A z-score appears beside it. That is a different measurement of the same position: how many standard deviations the current net position sits from its own mean. The COT Index answers "where in the range", the z-score answers "how unusual". They can disagree, and when they do it usually means the range is wide but the distribution inside it is lopsided.

What crowded does not mean. A crowded reading is a statement about how one group is positioned, not a prediction that the position is about to unwind. Crowded positions have stayed crowded for months. The panel reports the level and, separately, what has historically followed that level when the sample is large enough to say.

The fundamentals axis: the 5-year seasonal band

This is the idea the whole panel is built on, so it is worth stating precisely.

For each series that carries a band, we take the value recorded in the same ISO week of the year across the last five years, and take the minimum, the average and the maximum of those five values. That triple is the band for that week. Every week of the year has its own band, so the reference moves as the year moves.

This strips the seasonal shape out. Gasoline stocks are supposed to build through winter and draw across the summer driving season. A summer draw is not news. A summer draw that takes stocks below anything seen in that week for five years is.

Three details that change how you read it.

First, the edges are inclusive. A value sitting exactly on the seasonal floor is at that floor, not in the interior. Strict inequality here was a real bug, fixed in July 2026, which read a level sitting precisely at its five-year low as ordinary.

Second, the band includes the current year. The five years counted are the last five including the one in progress, so the current reading is one of the five values that define its own band. A reading can therefore sit exactly on an edge but never past it. This is why the panel says a series is "at or above the top of its 5-year seasonal band" rather than "above" it, and why the band strip marks edge-touching explicitly instead of leaving room for excursions that cannot occur. When you see a series flagged at its ceiling, it means it is the highest that week has been in five years, not that it has broken out of a range.

Third, the band is descriptive, not directional. It says where a level sits. It does not say a high level is bearish or a low level is bullish. We tested that assumption and the record did not support it; see the base rates section below.

The three positions:

Alongside the position, each banded series reports its percent distance from the five-year average for that week. That number is signed and is what the band strip's right-hand column shows.

Reading the three visuals

The positioning gauge sits at the top. One dot on a 0-100 track, with the shaded zones below 20 and above 80 marking the crowded thresholds.

The band strip is the fundamentals read in one glance. Every banded series gets its own row and its own track, because the series carry different units and different band widths; a shared axis across thousands of barrels, days of cover and billions of cubic feet would be meaningless. Within each row the shaded span is that series' own band for this week, the thin line is its five-year average, and the dot is where it sits now. Rows sitting at an edge are sorted to the top and marked, because they are the rows carrying information this week.

The seasonal band chart plots the panel's headline series over the last 52 weeks against its own envelope: the five-year minimum, average and maximum for each of those weeks. The headline series is commercial crude stocks on the WTI panel and Lower 48 working gas in storage on the gas panel. Marked points are weeks where the line sat exactly on the floor or the ceiling.

Because the band includes the current year, the line can touch the envelope but never leave it. Read the marked points, not the gaps.

The WTI panel, series by series

Stocks, all measured against a seasonal band

• Commercial crude stocks. Crude held by refiners and traders outside the SPR. Rises when supply into the system exceeds what refiners run. This is the panel's headline series.
• Cushing stocks. Storage at the WTI delivery point in Oklahoma. Because it is the physical settlement hub, it moves the front of the futures curve more than its size alone suggests.
• Strategic Petroleum Reserve. Government-held crude. Changes reflect policy decisions to release or refill, not the commercial balance. The panel flags it separately for exactly this reason and never lets it rank beside commercial series in the read.
• Gasoline stocks. Finished gasoline and blending components. Strongly seasonal.
• Distillate stocks. Diesel and heating oil together.
• ULSD stocks. Distillate at 15 ppm sulfur or below, road and off-road diesel. The bulk of the distillate pool, tracking freight and industrial activity.
• Heating oil stocks. Distillate above 500 ppm sulfur. A small slice, drawn on by Northeast winter heating.
• Jet fuel stocks. Kerosene-type jet fuel, competing for the same distillate cut as diesel.
• Propane stocks. Driven by US winter heating and by petrochemical export demand.

Days of supply, also banded

Five series: crude, gasoline, distillate, jet fuel and propane. Each divides the stock by current demand, answering how long the system could run on what is in tank.

This matters because a barrel count alone can mislead. A build that only keeps pace with rising demand is not really a build, and days of supply is the series that says so. Where a days-of-supply reading and its underlying barrel count are both outside their bands, the read treats them as one story rather than counting them twice. Propane runs far higher than the others because stocks are built across summer for a winter draw.

Balance flows, not banded

Nine series, reported as levels and as a four-week trend rather than against a seasonal band: crude production, refinery utilisation, refiner crude input, crude imports, crude exports, crude net imports, gasoline production, gasoline exports and distillate exports.

Read these as the mechanism behind a stock move. Exports compete with domestic refiners for the same barrels, so they draw on inventory the same way a refinery run does, which is how domestic distillate stocks can fall while refinery runs look healthy. Weekly import figures are lumpy because a single cargo arriving a day early shifts the week. Weekly production is an estimate and gets revised against the monthly series.

Demand proxies, not banded

Five product-supplied series: total, gasoline, distillate, jet fuel and propane. This is EIA's proxy for demand, measured as product leaving primary storage rather than as end-user consumption. It is noisy week to week, which is why the panel also carries a four-week trend.

Prices and cracks

Five daily spot series: WTI, Brent, NY Harbor gasoline, NY Harbor ULSD and NY Harbor heating oil. Brent's spread to WTI reflects what it costs to move a US barrel to the international market.

Three product cracks are computed against WTI, each with a percentile against the daily history actually held, and the observation count is shown beside it rather than assumed. At the time of writing that is 260 observations, roughly a year, so the percentile is never presented as covering a longer span than it does. Cracks are labelled context only. The crossed cells of positioning against crack level carry an independent sample of around seven, which is nowhere near enough to publish, so no direction is attached to them.

Note that prices are daily while the fundamentals are weekly. The panel carries their dates separately, and the headline "as of" date always refers to the weekly fundamentals, never to the more recent price print.

The natural gas panel, series by series

Storage, all banded and all season-aware

Working gas in underground storage, from the EIA Weekly Natural Gas Storage Report:

• Lower 48 working gas in storage. The national figure, and the gas panel's headline series.
• East, Midwest, South Central, Mountain and Pacific regions. The same figure broken out. East and Midwest demand is the most directly heating-driven. Pacific competes with hydro, and its draw is driven more by summer power burn than by winter heating. Mountain is the smallest and the most exposed to local production swings.
• South Central salt and South Central nonsalt. The South Central total split by storage type. This split is the one worth knowing about: salt caverns can be injected and withdrawn far faster than depleted fields, so they move first when the market needs gas quickly. Read against each other they separate fast-cycling capacity from slow seasonal capacity.

Every gas storage series is season-aware, meaning the panel names which season the week falls in: withdrawal from November to March, injection from April to October. Naming the season explains why a level is moving. It says nothing about price. The same storage level genuinely means two different things in April and December, which is why the seasonal band matters more here than anywhere else on the product.

Monthly supply and demand, deliberately not banded

Three monthly series, all in million cubic feet: dry gas production, LNG exports and total consumption. They are drawn as a 24-month time series so two full seasonal cycles are visible.

These carry no seasonal band, and that is a deliberate correctness decision rather than an omission. We backtested a five-year month-of-year band on all three. Dry production sat above such a band in 70.4% of months and below it in 0.0%; LNG exports 87.2% and 0.0%; consumption 53.6% and 2.4%, across 125 months each. These series trend with shale output and LNG buildout, so a seasonal band on them detects the trend while carrying a label that says season. A flag that fires 87% of the time and can never fire the other way is not information.

Instead each carries a year-over-year figure, which holds the month of the year fixed and measures against one year of trend rather than five. It states a growth rate as a fact and asserts no extreme.

Two things to keep in mind reading them. They are monthly, so every comparison on the tile is month-on-month and any high or low flag names a 12-month window, not a 52-week one. And they are published roughly three months in arrears, so the panel reports their date separately from the weekly fundamentals date. They describe the recent past, not this week.

Henry Hub spot

The US benchmark gas price, and the price the positioning is read against. Daily.

Degree days

Alongside the storage read, the panel carries one weather sentence built from heating and cooling degree days in the EIA Short-Term Energy Outlook, comparing the season so far against the 10-year normal.

Weather is the dominant driver of gas demand, so a storage level read without it is missing the reason the level is where it is. Two constraints apply: only complete months count, because a month in progress would always look short against a full-month normal, and the source series carries forecast months beyond the present which are dropped rather than presented as observations. Colder or warmer than normal is a statement about weather, not about price.

The read

Above the series, the panel opens with a short read: where positioning sits, what has historically followed that positioning level, where the fundamentals sit, whether any measured inventory outcome applies, and a closing limit statement.

That last statement is the important one. It says plainly that we do not combine the two axes into a single conclusion. Both axes are real and both were measured, but they were never measured against each other, because every crossed combination of a positioning state and an inventory state has fewer than 30 independent observations across the whole history available. Placing a positioning sentence beside a fundamentals sentence invites multiplying them into a conclusion neither supports, so the panel says so rather than leaving you to infer it.

Measured base rates, and the rule behind them

This is the only part of the panel permitted to express a direction, and it does so by reporting what followed historically, never by asserting what will follow.

The publication rule. A relationship renders as a measured base rate only if it clears both thresholds: at least 30 independent observations and an absolute t-statistic of at least 1.0. Everything else renders as context with its count shown and no directional wording.

The independence correction matters. Forward windows are 8 weeks long but sampled weekly, so consecutive observations overlap roughly 8 to 1. A raw weekly count overstates the distinct evidence by that factor, so raw counts are divided by 8 before the rule is applied. The same correction is applied to the positioning buckets, so the two evidence bases are judged by one standard rather than two.

What we tested. Fourteen candidate relationships between an inventory's band position and the price over the following 8 weeks, across WTI as a control, gasoline, distillate, natural gas and natural gas salt storage, with natural gas additionally split by season.

What cleared. Exactly one: WTI commercial crude stocks at or above the top of their seasonal band, independent n = 32, t = +1.49, mean forward return +3.45% over 8 weeks.

Read that carefully. The measured direction is positive after a high inventory reading, which runs opposite to the intuition that a build is bearish. That is precisely why the panel refuses to attach the old direction labels: we checked, and the record contradicted them.

What did not clear, and is therefore absent from the panel: WTI stocks below the band (independent n = 8), every gasoline and distillate relationship tested, and every natural gas relationship, including the seasonal splits, which fell short on independent observations alone. Gasoline stocks above band reached t = +1.55 but only 28 independent observations, below the bar of 30. Distillate above band reached t = +2.11 on 12 observations. Both are interesting and neither is published, because a large t on a small sample is exactly the thing the rule exists to catch.

Active versus historical. When a base rate is shown, the panel states whether its condition matches this week's reading. "Active this week" means the series it was measured on is currently in the band position the base rate describes. "Not active" means the row is history and does not describe the current week, and the panel says which position the series is actually in. Never read a base rate as describing today unless it is marked active.

One nuance worth stating for anyone reconciling the numbers: the validation study computed each week's band from the five prior calendar years, excluding the year in progress, while the live panel computes it from the last five years including the current one. The two definitions agree on every current reading at the time of writing, but they are not the same construction, and the base rate should be read as the measured history of a condition rather than as a rule the live classifier reproduces exactly.

What Energy Radar does not do

• It does not forecast price, and it states no win rate, price target or edge.
• It does not combine positioning and fundamentals into a single score. The crossed cells are not measurable at the sample sizes available.
• It does not label an inventory build bullish or bearish. Where a direction has been measured, it appears once, with its sample.
• It does not cover markets beyond WTI and natural gas today. Gasoline, distillate and the other products appear as fundamentals rather than as positioning inputs. That is a limit of what we have built, not of the data: RBOB gasoline and NY Harbor ULSD each carry their own Managed Money positioning in the CFTC report and are already tracked elsewhere on COTInsight. A panel for either needs its own positioning source wired in, which is a change we can make rather than a constraint.

Frequently Asked Questions

What does "at or above the top of its 5-year seasonal band" actually mean?

The value equals or exceeds the highest reading recorded for that same week of the year across the last five years. Because the band is computed over five years including the current one, a reading sits exactly on the edge rather than beyond it.

Why do some series have no band?

Bands are computed for every stock and storage reading, including the days-of-supply series, where a like-for-like seasonal reference is meaningful. That is 22 series, and nothing outside that group carries a band. Flows and demand proxies carry a four-period trend and a change on the period instead. Prices carry a change only, with no trend and no band. The monthly gas series carry a trend, a change and a year-over-year figure: a seasonal band was tested on them and rejected because it detects trend rather than season.

Why is the monthly gas data three months old?

That is the EIA's publication schedule for those series, not a lag on our side. The panel reports their date separately from the weekly fundamentals date so a reader never mistakes one for the other.

What is the difference between days of supply and a stock level?

The stock level counts barrels. Days of supply divides that by current demand, so it answers how long the system could run on what is held. A build that only keeps pace with rising demand shows up as flat days of supply, which is usually the more honest read.

Does a series at its seasonal floor mean the price is going up?

No. The panel states the position and stops. Of fourteen relationships tested between a band position and the price that followed, one cleared our publication bar, and it concerned stocks at the top of the band rather than the base.

Which figure updates when?

Positioning updates weekly with the CFTC report. Petroleum and gas storage fundamentals update weekly with the EIA reports. Spot prices are daily. The monthly gas series update monthly, in arrears. Each carries its own date on the panel.

Is Energy Radar available below Ultimate?

No. It is an Ultimate-tier feature on the WTI and natural gas detail views.

Summary

• Energy Radar shows two independent measurements: the COT Index (0-100 against a 3-year range, crowded below 20 and above 80) and the 5-year seasonal band (per ISO week, edges inclusive, computed over five years including the current one).
• 45 series across two panels: 33 on WTI with 14 banded, 12 on natural gas with 8 banded.
• A band position states where a level sits and attaches no direction. The monthly gas series carry year-over-year instead of a band, because a band on a trending series measures trend while claiming to measure season.
• Of fourteen relationships tested, one cleared the publication rule of 30 independent observations and absolute t of at least 1.0: WTI stocks at or above the band, n = 32, t = +1.49, +3.45% over 8 weeks. The other thirteen are absent because they did not earn a place.
• The two axes are never multiplied into one conclusion, because every crossed cell is too thin to support one.

Data sourced from the CFTC Commitments of Traders report (cftc.gov), the EIA Weekly Petroleum Status Report, the EIA Weekly Natural Gas Storage Report and the EIA Short-Term Energy Outlook (eia.gov), all free public sources. This article describes how to read a COTInsight product feature; it does not forecast price and is not investment advice. Futures trading involves substantial risk of loss.