// Account

Platform access requires an active subscription

NatGas Edge
Henry Hub natural gas · Analytics platform
For traders who do their own work
Henry Hub AnalyticsThe missing middle.
Institutional platforms cost £30k+ per year and bundle nat gas inside multi-commodity packages built for procurement teams. Retail tools are free or cheap and stop at charts and opinion.
NatGas Edge is the gap.
Single-commodity depth. Modern infrastructure. Transparent methodology. Individual pricing.
Sixteen years of every Henry Hub fundamental — weather, storage, positioning, renewables, LNG feed gas, price — queryable with seasonal context, forecast revision tracking, and anomaly flags. Synthesised into weekly directional and storage signals with three years of audited live track record.
16 yr
Fundamentals depth
Every signal · Daily
62%
Directional accuracy
156 weeks live · 50% baseline
18 Bcf
Storage MAE · 7 days early
156 weeks live
No account?
Daily fundamentals · Weekly signal
NG1 Front Month$2.755 +0.106
Prev Close$2.649Prior session close
Daily Range$0.0652.36% Today
Volume6,803Today
Resistance 20D$2.888+4.8% above
Support 20D$2.483-9.9% below
EIA Latest WoW+101 Bcf Inject
Storage Surprise+13 Bcfvs NatGasEdge pred
NG1 Front Month$2.755 +0.106
Prev Close$2.649Prior session close
Daily Range$0.0652.36% Today
Volume6,803Today
Resistance 20D$2.888+4.8% above
Support 20D$2.483-9.9% below
EIA Latest WoW+101 Bcf Inject
Storage Surprise+13 Bcfvs NatGasEdge pred
// What this is
Most retail tools give you
charts and opinion.
Most institutional ones
aren't accessible to individuals.
This is the missing middle.

NatGas Edge is the research environment a serious nat gas trader would build for themselves if they had the time and engineering depth.


Every Henry Hub fundamental ingested daily and contextualised against sixteen years of history. Every forecast tracked through its revision history. Every metric displayed with seasonal percentile bands, z-scores against same-week distributions, and anomaly flags surfaced the day they fire.


A directional model and a storage model on top of the data layer, both validated walk-forward across 156 weeks of out-of-sample predictions. The signals aren't the product — they're the synthesis of the product.


For analysts and serious traders who need to see the work, defend a view, or build their own on top.

// The data layer · What sits underneath every signal
Sixteen years.
Every fundamental.
Queryable.
Weather
Regional temperature, HDD, CDD
Historical daily data across South Central, East, Midwest, Mountain, Pacific. 15-day forward forecasts split into 1-5 / 6-10 / 11-15 day horizons. Anomalies vs 5-year and 10-year norms on every region, every horizon.
Renewables
ISO-level gas burn displaced
Hourly TWh by ISO and fueltype across CAISO, ERCOT, MISO, SPP, PJM, NYISO, ISNE. Daily absorption ratios calibrated against EIA-930 actuals, modelled forward at near (D+1-5), mid (D+6-10), and far (D+11-15) horizons. Gas displacement in Bcf surfaced per ISO.
LNG feed gas
Seven major export terminals
Daily feed gas volumes for Sabine Pass, Corpus Christi, Calcasieu Pass, Plaquemines, Cove Point, Cameron, and Elba. Per-terminal and aggregate national flows displayed against seasonal norms with anomaly flags. The dominant bullish narrative in nat gas — quantified.
Storage
Lower 48 underground storage
Weekly storage levels and balance. Five-year min/max/average bands. Z-scores against historical same-week distributions. Surplus/deficit tracking vs five-year norm.
Positioning
Managed money COT
Weekly Commitments of Traders — managed money long, short, net position. Z-scores against 52-week distributions. Regime flags: overbought, oversold, extreme.
Price
NYMEX prompt + forward curve + spot
NYMEX prompt OHLC, panama-adjusted continuous series, full forward curve. EIA Henry Hub spot. Calendar spreads, percentile rankings, seasonal price context.
Every metric, contextualised
Nothing on the platform is just a number. Every fundamental is displayed with rolling seasonal percentile bands, z-scores against historical same-week distributions, and automatic anomaly flags. You don't see "storage is 1,847 Bcf" — you see that it's at the 28th percentile of where storage has been at this week of the year for the last sixteen years.
// Forecast revisions · A capability layered across the platform
The number changed.
That's the signal.

Most retail platforms display whatever the latest forecast says, as if today's number is the only number that exists. But forecasts revise — daily, sometimes dramatically — and the revision itself often matters more than the absolute level.


For every weather, renewables and storage forecast in the system, the day-over-day change is computed and z-scored against the trailing 60 days of revisions. A z-score beyond ±2σ means the forecast just shifted by more than it typically does — the model is updating on new information faster than usual. A cold-weather outlook that's been stable for a week is already priced in; the same outlook revised sharply colder overnight is information the market hasn't fully absorbed.


This is a unique capability layered across the platform, not a single feature. Wherever there's a forecast, there's a revision history beside it.

How it works
60-day rolling baseline · ±2σ threshold
For each forecast, we compute the day-over-day delta and z-score it against the distribution of deltas over the trailing 60 days. The baseline self-adjusts to the recent revision regime — what counts as a shock in calm weeks differs from what counts in volatile weeks.
Where it appears
On every forecasted fundamental
Weather per region per horizon, renewables per ISO per fueltype per horizon, combined weighted demand, storage trajectory. Every chart that shows a forecast also shows its revision history.
// Platform preview
Storage surprise model
Thursday's number,
seven days early.
EIA week-on-week storage change forecast, published seven days before the official release. 18 Bcf mean absolute error across 156 live weeks — comparable to the Reuters and Bloomberg analyst consensus, delivered seven days ahead of the published analyst consensus.
Storage surprise model
Directional model
62% accuracy.
Three years live.
Weekly directional forecast with probability distributions. 62% strict accuracy across 156 walk-forward out-of-sample weeks, rising to 67% on the high-conviction subset (~36 weeks per year). Trained on the full nine-fundamental feature set, validated walk-forward, never fitted to data it later predicted.
Directional model
Composite signal view
Nine signals.
One synthesis.
Every fundamental scored bullish, bearish, or neutral with a written justification grounded in the actual pipeline output. The composite reads all nine and produces a single weekly verdict with reasoning. The synthesis layer that turns data into a view.
Composite signal view
Weather and renewables
What actually
moves gas.
Regional HDD, CDD, mean temperature, wind generation, solar generation — each split across short-range, medium-range, and long-range forecast horizons. Displayed against rolling seasonal percentile bands across sixteen years. The features the directional model actually leans on most.
Weather and fundamentals
LNG feed gas
The bullish narrative,
quantified.
Daily feed gas volumes for the seven major US export terminals — Sabine Pass, Corpus Christi, Calcasieu Pass, Plaquemines, Cove Point, Cameron, and Elba — broken out per facility and aggregated to national flow. Each terminal displayed against its own seasonal norms with anomaly flags when daily flows diverge from typical operation. LNG demand is the structural bullish story in nat gas; this is what it looks like in numbers, not narrative.
LNG feed gas
Anomalies · Active shocks & compound patterns
What's unusual
right now.

The Anomalies page consolidates every active forecast revision shock across the platform into one diagnostic view. Six tabs — Active, Weather, Renewables, Storage, Positioning, Price — each showing every z-score above the configurable threshold, with directional implications for gas marked bullish or bearish.


Beyond individual component shocks, the page surfaces compound patterns — multi-feature configurations measured against sixteen years of forward price data and only published when hit rate, mean return, and median return all beat baseline meaningfully on N≥30 historical instances. Each card displays the live trigger value alongside its measured historical context — what the configuration has done over the following ten days, every time it has fired since 2010.


An as-of date picker navigates the page back through history. Pick any date and the platform shows exactly what would have been visible that day — the same z-scores, the same shocks, the same compound flags. A time machine over the anomaly substrate.

Anomalies · active shocks and compound patterns
Managed money positioning
Where positioning
actually is.
CFTC Commitments of Traders tracked weekly across sixteen years. Managed money long, short, net position, weekly deltas — each z-scored against its own 52-week distribution. Regime flags surface when positioning is at historical extremes, not just where it sits in absolute terms.
COT positioning
User experience
Speed to insight,
not speed to chart.
Every fundamental, every signal, every track record — surfaced through one terminal. Nine signal cards render the week's verdict at a glance; the depth sits one click below. Date filters, seasonal overlays, percentile bands and z-scores are computed continuously, so the contextual analysis is already done by the time you log in.
Full Dashboard
// Why trust this
Every accuracy number
came from a prediction
made before the outcome.
Walk-forward validation
Every published metric comes from a model trained only on data available before the prediction date. No look-ahead. No fitting to outcomes. The accuracy you see is the accuracy a subscriber would have experienced.
Three years of live track record
156 weeks of out-of-sample predictions, every one logged with its probability and the actual outcome. Not a backtest. Not a simulation. Downloadable as CSV.
Documented methodology
How features are constructed, how the model is trained, how the walk-forward is run, where the data comes from. Dated, versioned, public. The credibility comes from the audit trail, not the marketing.
Honest sample sizes
Accuracy claims are made at thresholds where the sample is large enough to support them. The same discipline applies to compound anomaly patterns — only configurations with N≥30 historical instances where hit rate, mean and median all beat baseline are published.
62% directional accuracy across 156 weeks of out-of-sample validation. 67% accuracy on the high-conviction subset (~36 weeks per year). Storage MAE 18 Bcf across 156 weeks. Track record CSV available on the methodology page.
The numbers, in one paragraph
// See the track record · download the CSV · read the methodology
Methodology
// Pricing · Two tiers

Edge is the signal — a structured weekly read on Henry Hub fundamentals delivered after the Thursday EIA release. Built for traders who want the verdict and the headline numbers.


Intelligence is the platform underneath the signal — the full data layer, the forecast revision tracking, the anomalies page, the date-filtered seasonal context, the week-by-week out-of-sample track record. For analysts and serious traders who need to see the work, defend a view, or build their own on top.


Compared against £30k/yr institutional platforms, both tiers are launch-priced — Edge at 0.16% of an institutional contract, Intelligence at 0.8%. The difference is access, not depth.

Edge
£49 /mo · launch
The signal, delivered weekly.
Weekly directional forecast — bullish, bearish, or neutral with conviction
Composite signal across all nine fundamentals
Storage surprise model — EIA prediction with running MAE
Hero cards visible on every dashboard
Email delivery every Thursday after the EIA release
The full platform
Intelligence
£249 /mo · launch
The signal, plus the platform that produces it.
Everything in Edge, plus —
Storage forecast Thursday afternoon ET — five days before Edge
All sixteen years of every fundamental, queryable
Per-signal written justification grounded in pipeline data
Date filtering and seasonal comparison tools
Z-scores, percentile bands, anomaly flags on every metric
Forecast revision z-scores on every weather, renewables and storage forecast
Dedicated Anomalies page with active shocks and validated compound patterns
LNG feed gas across all seven major export terminals
Full week-by-week out-of-sample track record
Accuracy breakdown by season, regime, conviction
CSV export across all data and predictions
Launch pricing through the founding window · sunset to £99 / £349 thereafter · Founding subscribers locked at launch price
Data and analysis only · Not investment advice · FCA disclaimer applies throughout