Every storage headline this year quotes one number: percent full. But a storage site cannot deliver its nameplate withdrawal rate at low inventory. Capability decays with pressure, fastest below 40-50% fill, and fills bottom in late March. We built the index that measures this, backfilled it across 11.6 years, and the record is unambiguous: every measured winter trough lands between 19 March and 1 April, the one season that touched the realized-peak zone was 2018, and on current reads this winter opens nearer that zone than any since the crisis. The market prices Q1-27 as one flat block of risk. The trade is the shape.
Three facts jump out. First, the cliff is seasonal and late: every measured winter trough across eleven seasons lands between 19 March and 1 April, when attention has moved on. Second, comfortable winters stay far from it: 2022/23 and 2023/24 troughed at 87.7% and 89.4% on the days on record (both winters carry a multi-week backfill hole across Jan-Feb, so their true minima could sit somewhat lower). Third, the one season that touched the realized-peak zone was 2018: a late-February cold wave met draining storage and availability fell to 52.9% of technical by 30 March. That is the season this winter's starting arithmetic most resembles.
The standard objection is aggregate and historical: OIES notes that since 2017, EU-wide daily withdrawal has never exceeded roughly 60% of daily withdrawal capacity (Insight 159), and last winter's realized peak was 10.38 TWh/d against a ~20.2 TWh/d EU technical maximum. On that arithmetic, sendout capacity has never been the binding constraint. The objection is correct at the winter's start, and the record shows exactly where it stops being correct.
Last winter is the template: a 96.0%-availability start drew down to a 64.4% trough by 31 March (that percentage is of technical withdrawal capability; by coincidence it matches the sibling note's 30 Aug fill print, a different quantity entirely). The EU fill went from 82.8% on 1 November to 27.9% on 1 April, a draw of 55 fill-points, both endpoints from our own backfilled AGSI record. This winter's central forecasts put the 1 November fill at 70-78%, five to thirteen points behind, and the deferral mechanics our sibling theses document (no forced autumn buyer, thinner commercial buffer, state-held inventory) do not soften the draw. Run last winter's own 55-point draw from this winter's start range: from 78% it ends near 23% fill, from 70% near 15%, the exact fill zone of the March-2018 record. On the current curves those endpoints map the covered index to roughly 63% and 54% of technical respectively (treating the covered set at one uniform fill, an approximation: the live index weights each country at its own fill, and the countries diverge). That is arithmetic on stated assumptions, not a forecast; prices can ration demand harder than last year and lift both numbers. What the arithmetic does not reach, from this starting point, is the comfortable 87-89% of 2023 and 2024.
Fill crosses 40% in late January, a month earlier than last season; DE and FR cross their knees in February. The index troughs 56-63% in late March. Late-winter months reprice against early-winter months as the physical story becomes legible; prompt volatility rises into February-March.
A blocking pattern lands after mid-February on sub-30% fills. Available capability sets a new record low below 52.9%; peak-day need and available sendout converge for the first time on record; late-winter prompt and within-Q1 spreads gap. This is the scenario the flat Q1 strip gives away for free.
A warm, windy winter from a 78% start keeps fills above 45% through February; the index holds 80%+ and the steepener bleeds carry to nothing. This branch is fatter here than in the sibling theses on purpose: a mild winter merely compresses their spreads, but it kills this thesis outright. The loss is bounded by construction; size to this branch.
The market prices Q1-27 as one block; the record says the risk lives at the block's end. Long the months where every measured trough landed, short the months the panic premium already inhabits. Level, Hormuz and the winter-total risk largely cancel across the legs; a parallel repricing of the whole Q1 strip nets to roughly zero here, which is what makes this genuinely a shape trade rather than a second copy of the winter level our sibling theses own. Month marks need your own curve (public prints quote quarters); the platform's spread monitors take BYO marks.
Lineage, stated plainly: the calendar core of our June note, Short the Calm, was restated by The Empty Auction in August on the auction record; its convex overlay (Q1-27 call spreads) is the leg that survived, and this thesis is that leg's instrument. The cold-snap state where those calls pay now has a daily-measured engine; the Hormuz branch the June note also named remains un-instrumented and is not this note's claim. The June structure's funding (the soft summer) is gone, so entering now means paying post-repricing premium at higher strikes: re-underwriting the overlay, not re-running it. A sizing warning: for a desk already long Q1-27 via the Empty Auction or still carrying the June overlay, this leg is a replacement and a monitor, not an addition. Stacking all three is a triple-long on the same March tail.
Aquifers decay before caverns: France's blended curve is the earliest of the five to fall, Germany's knees hardest below 40%. If February-March tightness localizes, PEG and THE late-winter basis against TTF widens before the aggregate story does. No free feed carries those hub curves; this leg exists for desks with their own marks.
This closes a trilogy: The Empty Auction prices this winter's level, The Second Bill the 2027 refill the deferral rolls into, and this note the shape and volatility inside the winter both siblings look across. The legs are designed to net rather than to avoid each other, and one netting is deliberate and named: the steepener's short Jan-27 sits inside the Empty Auction's long Q1-27 (and inside Short the Calm's, where carried). Sized 1:1 per leg, the combined book goes net short January and double-long February-March. That is the thesis, stated: Q1's repricing concentrates late. One honest edge: if the Empty Auction's base case plays early, with January inventory scares doing the repricing, the steepener's short leg pays against it for a stretch. A desk that wants the block AND the tilt sizes the steepener below 1:1.
The index holding at or above 80% through February 2027: the draw never reached the knee, and the steepener’s premise is gone.
The ENTSOG Winter Supply Outlook 2026/27 (due ~October) revising the deliverability curves materially flatter: the cliff was re-surveyed away. The index is re-scored on the new edition either way.
The index crossing below 75% before end-January. That has never happened in eleven measured winters (closest: 76.1% in January 2026, 78.0% in January 2022).
Fills crossing 40% before February; FR/DE basis firming while the TTF front is quiet; the draw pace outrunning last winter’s from a lower start.
The desk treats a rule as armed only when something watches it. The instrument runs live on the Voltstack platform (the index and its 11.6-year history), its alert system can watch these thresholds, and the desk record at app.voltstack.energy/thesis states at any moment whether the rule is armed. The headline on this page reads from the live index.
This note exists because we built the metric first. The storage deliverability index runs live in the Voltstack workspace (per-country curves charted against today's position, trend view over the full backfilled record, provenance and grades printed on the widget face), and the number at the top of this page is served by the same API.
| Capability | Role in This Thesis | Generic Alternative |
|---|---|---|
| Storage deliverability index, live + 11.6y history | The instrument itself: fill-adjusted available withdrawal, per country and covered total. | Does not exist elsewhere |
| AGSI+ storage corridor, by country | The fill trajectory that drives the curves' input. | Weekly CSV, aggregate only |
| Curve & spread monitors (BYO marks) | The steepener and basis legs, tracked against your own month marks. | Excel and screenshots |
| Alerts | The 80%/75% rule thresholds, watchable on the live index. | Reading the chart and hoping |
Storage deliverability index · AGSI+/ALSI+ live · ENTSOG flows · Curve & spread monitors · REMIT II native
voltstack.energy