Depletion Model — specification (distilled) ¶

Distilled from the main report (§3, §4, §9A, §9B, §11) on Sep 9, 2026 (Day 193); branch weights reweighted Sep 9 (regime event: biggest tit-for-tat wave of the war, exclusion zone enforced, first third-country base hit). The main report is the source of truth; this page is the compact model card — what goes in, what is assumed, what comes out, and how it updates. Review before treating as canonical.

1. Inputs (what feeds the model) ¶

2. Core assumptions ¶

1. Product, not crude, is the scarce asset. US crude ~normal; the constraint is distillates + SPR. Brent $88–99 is carried by demand destruction (−1.6 mb/d IEA), not by physical crude shortage.
2. Three branches, weights reweighted Sep 9 (regime event: 10 Iranian tankers gone in a week, exclusion zone enforced before declared, first strike on a third country; prior weights Sep 7):
   • Corridor holds (extended, shadow flows — now mostly on the Iranian-controlled route, inside the exclusion zone) — ~10%
   • Standoff drift (hotter standoff: published US exchange rate, Iranian exclusion zone; gradual decay) — ~50% ← base case
   • Corridor lapses (no extension) — ~40%
   • Swing factor: the Oman safe corridor (IMO filing claimed Sep 8–9, unverified) — if real, "corridor holds" reconstitutes under Iranian management (less free-flow, more revocable). Verify against IMO circulars before moving weights back.
   • New gauge: the EIA–market spread (EIA 2H26 ~$90 vs spot >$100). Widening past ~$10–15 into October ⇒ market prices lapse >40% ⇒ move the weights with it.
3. SPR draw rate is the leading indicator, not the level. 286.6M (Aug 28), drawing ~0.45M b/d; the ~3M/wk continuation after the IEA program nominally ended is unexplained and is the most policy-dependent line in the model — one DOE announcement flattens the curve.
4. Floor hierarchy (SPR): ~300M cavern-collapse floor (BREACHED Aug 14) → ~250M GEF operational floor → ~180M hard-operable floor → ~70M DOE safe minimum → ~44M unrecoverable. The 300/180 figures are engineering estimates, ±50M, not published DOE numbers.
5. Political breaks precede physical ones. Slovenia/Ireland broke on price, not scarcity. US retail diesel all-time record ($5.90, Sep 7) means the US tripwire is crossed too.
6. The PADD1↔ARA pool is shared (34% of ARA August gasoil imports are US-sourced). The US East Coast and Europe break in the same week, not in sequence.
7. The baseline itself is damaged (§9B): reopening does not restore pre-war supply. Structural residuals: crude −0.6M bpd through 2027 (drift compounds it to 1.5–2.5M), Qatar LNG −17% until 2029–31 (12–14% of EU LNG), 30–55% of Russian refining out permanently by 2027.
8. China is opaque by design. All Chinese inventory figures (SPR est. 1.2–1.4B bbl; commercial est. 0.6–1.0B) are order-of-magnitude estimates. The commercial buffer runway (9–18 months) expires Q1–Q2 2027 → forces the re-import-or-release decision.
9. Days-of-cover ignores production and demand destruction (both extend actual cover).
10. Least stable premise: no Russia→Europe kinetic threshold-crossing. The hybrid front is a dial European politicians control; the one undated breaking point.

3. Branch endpoints (US, end of November 2026) ¶

Gasoline is safe in all branches (net exporter). Jet is fine through November; it snaps first in December.

4. SPR runway (from 286.6M, Aug 28) ¶

At any "nothing-changes" pace the SPR crosses the 180M floor within 6–12 months; by month 12 the reserve is at a 40+ year low with no capacity left to absorb a corridor lapse.

5. Dated breaking points (if nothing changes, Sep 2 baseline; Sep 8 note compresses most ~2–6 weeks earlier; Sep 9 note: regime event — most pull another ~1–3 weeks; the Sep 14/15 "Hormuz normal" bet (3.8%) settles ~0) ¶

1. Sep 30, 2026 — Russia producer-diesel ban expires; Russia cannot cover its own gap (28% of stations stocked, Kirishi down).
2. Late Sep – mid Nov 2026 — corridor expiry window (30–60-day lifespan).
3. Late Oct – Nov 2026 — ARA gasoil crosses trader floor (~8.5–9M bbl); first EU price caps.
4. Late Oct – Nov 2026 — SPR crosses 250M operational floor (pace-dependent).
5. Nov 30, 2026 — Russia jet-fuel ban.
6. Jan 31, 2027 — Russia non-producer gasoline/diesel ban (all three live simultaneously).
7. ~Feb / ~May 2027 — SPR crosses 180M hard-operable floor (0.7 / 0.45 pace).
8. Q1–Q2 2027 — China commercial buffer runway expires → re-import-or-release decision (the largest market event of the war).
9. ~Late Jul / ~late Aug 2027 — SPR at 70M / 44M on the 0.7 path; draw must stop for geology.
10. ~Mar–Apr 2027 — Japan buffer exhaustion risk (203d burned 10–15% faster; METI mandates trigger first).
11. Any date — a calibrated Russia→Europe hybrid attack fails to fail (passenger aircraft, casualties) → Article 5 live; breaks every energy assumption at once.

Headline: the status quo cannot last 12 months intact. By month 9, either the corridor or China's buffer breaks.

6. Key limitations ¶

• This is a stress test, not a forecast — the value is the ordering of breaks, not the dates.
• Traffic figures are a range (3–29 transits/day verified; claimed 8.6–10M bpd dark flows unverified). Model on the verified floor.
• ARA depletion path is an interpolation (GEF publishes the December endpoint, not weekly stock); late-October floor crossing ±3 weeks.
• Every Chinese number is an estimate; China discloses nothing.
• Infrastructure figures: Abqaiq ~7M bpd offline is an upper bound (social-media-derived); South Pars "40% restored" is an Iranian self-report.
• Demand-destruction figures are contested: IEA −1.6 / EIA −1.2 / OPEC +0.58 mb/d — a 2.2 mb/d spread. The spread itself is the signal (the seller is bullish on demand).
• The macro channel (recession/stagflation) is under-weighted: bank recession odds 25–50% vs Polymarket ~7% (narrower window); the Fed is pro-cyclical into the shock (hike camp, 3.50–3.75%). Tripwire: a second consecutive upward revision of 12-mo recession odds, or a Fed hike into the winter draw.
• I am a coding agent, not an energy analyst; verify load-bearing numbers against primary sources.

7. Update procedure ¶

1. Pull: EIA WPSR (weekly), UKMTO/JMIC + PortWatch (weekly), AAA diesel + Brent/WTI settlement (daily), strike/ban reports (weekly), METI/KEA + China customs (monthly), IEA/OPEC/STEO (monthly), prediction markets (as published).
2. Log every new data point in research/<YYYY-MM-DD>.md with source and date. Unverified items are flagged as such and never enter SUMMARY until verified.
3. Re-baseline §3/§4 of the report on each WPSR (levels, WoW, draw rate). The draw rate change is the trigger for branch reweighting.
4. Reweight branches only on regime events (corridor deal/lapse, tanker-war escalation/de-escalation, China move, ban dates landing) — not on weekly noise.
5. Check breaking points: each dated row is both a market event and an escalation tripwire. When one crosses, the table's dominant variable changes; re-order.
6. Update SUMMARY.md at the end of every research pass (standing rule). Update the report's banner only on regime changes; append, don't rewrite.
7. Monthly: re-check the §9B damage inventory, the 11.2M→1.4M shut-in path, ACLED strike pace (if it doubles, the residual-loss estimate doubles), and the reflation risk on the corridor-holds branch.

8. External methods (literature survey) ¶

research/2026-09-09_literature.md maps each model component to its academic literature and off-the-shelf algorithms (buffer-stock theory — Williams & Wright 1978, Dorfman 1969, Pindyck 1979; chokepoint network flows — Pratson 2023; elasticity-driven demand destruction — EIA elasticity review; Bayesian scenario filtering + Brier-score calibration; API-first nowcasting; fat-tailed price paths — GED/GARCH; GPR-style escalation index — Caldara & Iacoviello 2022; ripple-effect graph analysis of the §11 cascade). Includes a prioritized adoption list. Adopted Sep 9: items #1–#3 are now scripts in model/ (elasticity updater, Bayesian branch filter + Brier-score calibration ledger, price-triggered draw rate); see model/README.md for usage and the Sep 9 worked examples.