EnergyScope — Monthly Oil Risk & Strategy Pack · July 2026 update · demonstration edition, written as for a crude desk

Monthly Oil Risk & Strategy Pack — July 2026 · Update, 24 Jul

Second edition of the July pack — same universe (now pinned as watchlist monthly_risk), fresh data, and the first real month-over-month engine diff. What changed leads. 5-minute version →

Generated: 24 Jul 2026 Cutoffs: futures/freight 23–24 Jul · EIA weekly 17 Jul · spots 20 Jul · CFTC 14 Jul (new COT ~19:30 UTC today) Engine: EnergyScope MCP v2.11.0 · 39 signals · fully audited (Appendix)

The tape since edition 1 (data 17 Jul): Brent M1 ran $88.10 → $100.69 (23 Jul) and printed $97.84 today — a −$2.85 reversal that may mark the squeeze's top. Spot-series staleness from edition 1 is fixed upstream (pipeline repair 23 Jul): spots now 4 days fresh, ICE Gasoil current.

1 · Executive summary

Three things that changed this week

The prompt squeeze went vertical. M1–M2 blew out $1.78 → $6.02; M1–M12 to $19.91; annualised roll −73.8%. The headline contradiction escalated elevated → extreme (0.83, Δ+0.49). measured
The state is now policy-managed scarcity. SPR released −5.1 mb in one week (the 172 mb US / 400 mb IEA program in our data); Cushing drew to 19.37 mb; distillate joined crude and gasoline at the 0th seasonal percentile — all three for the first time. measured
The consensus gap is the trade. Conflict ended 18 Jun (US–Iran MOU); the strait is reopening; EIA cut its Q3 Brent forecast $27 to $74. Yet M1 is $97.84 — while M12 ($77.93) sits almost exactly on the consensus. The curve prices both worlds; the −$19.91 spread is the market's estimate of the transition. measured read

Three biggest risks now

Convergence, violently. When logistics normalise, the front converges toward the deferred — up to ~$20 of give-back priced in the spread itself. Yesterday's −$2.85 may be its first day. judgment
Tonight's COT + FOMC (29 Jul). Positioning data is 10 days stale; if the crowd arrived during the rip, the "uncrowded" pillar breaks. 10-yr at 4.70% (+16 bp on the week) into the Fed. measured
Product-led fade. crack_extreme fired (3-2-1 z +3.25) with refiners at 96.1% — no supply response left; a crack collapse drags the complex. measured

Scenarios — Brent M1, ~1 month out

Scenario	Prob.	Brent
Squeeze decays toward the curve (base) — flows rebuild, front converges	50%	82–95
Squeeze persists / extends — restart disappoints, stocks keep draining	30%	95–110
Fast normalisation — agencies right, quickly ($74 EIA Q3 anchor)	20%	72–82

Bands: Brent M1 GARCH 52.4% ann. ⇒ ~15.1% (≈ $14.8) monthly σ; 1-day quantiles Q05 $92.1 / Q95 $102.4. The institutional anchor for the low band is EIA's $74 Q3 average forecast. Probabilities are analyst judgment — kept numeric deliberately: this 50/30/20 is an unscored prior, and the calibration ledger opens with this edition (words can't be Brier-scored). judgment

2 · Changes since last brief — the engine diff first real MoM diff

Signal transitions vs the 23 Jul edition: 3 new firing · 1 stopped · 3 intensifying · headline contradiction elevated → extreme.

Signal transition	Now	Was
crack_extreme new	z +3.25	+2.99 (the "one session" flip called in ed. 1)
gasoline_crack_extreme new	z +3.17	+2.90
distillate_undersupplied new	0th pctl	20th
production_at_record stopped	97.5th pctl	99.8th (output −63 kb/d)
punishing_carry ↑	score 1.00	0.61 (roll −24% → −74%)
steep_backwardation ↑	0.66	0.07 (−$11.07 → −$19.91)
freight_extreme ↑	z +4.32	+4.10
regime_curve_destruction ★	extreme 0.83	elevated 0.34

Metric	24 Jul	17 Jul	Δ
Brent M1	97.84	88.10	+9.74
M1–M2	+6.02	+1.78	+4.24
M1–M12	+19.91	+11.07	+8.84
Cushing	19.37 mb	20.04 mb	−0.67
SPR	311.4 mb	316.5 mb	−5.06
Crude ex-SPR	411.7 mb	409.7 mb	+2.01
US crude exports	3,353	3,721	−368
BWET freight	247.7	227.6	+8.8%
10-yr / DXY	4.70 / 101.4	4.54 / 100.8	+16bp / +0.6

The honest ledger — three calls from edition 1, scored. (1) The first print leaned against us: headline crude built +2.0 mb — said plainly, that was the draw-thesis missing its first check. The decomposition then matters: SPR released 5.1 mb the same week, so the commercial build is a policy transfer — the total system (commercial + SPR) still drew 3.0 mb. Thesis intact, but only after adjustment, and the adjustment is now part of the record. (2) The export paradox resolved — the arb model won: we flagged the weekly export print as the tiebreak between our balance model (exports +0.7 mb/d) and our arb model (economics shut). Exports fell 368 kb/d. Balance-model forecasts carry a demerit this edition. (3) Our own backtest corrects our own prose: edition 1 justified "stand aside on freight" with "extremes mean-revert" — the v0 backtest (§4) says in-sample they preceded rallies. Claim withdrawn; stance survives on different grounds.

3 · The headline contradiction — now extreme

regime_curve_destruction · extreme · score 0.83 (Δ+0.49) · 1 of 15 active · dominant lens

The market pays $6/bbl per month for barrels now — while pricing normal barrels in a year

Storage economics are not merely discouraged, they are inverted at maximum severity: punishing_carry is saturated (score 1.00, roll −73.8% annualised) and the backwardation itself is at −$19.91. Inventory holders liquidate working stock into the squeeze, deepening it — the loop from edition 1, running hotter. What is new is the shape of the resolution: the deferred curve never confirmed the rally (M12 moved $0.90 all week) — the market's own pricing says this ends by convergence, not by contagion. measured

Curve — M1–M12 +$19.91, M1–M2 +$6.02, roll −73.8% ★★★★★

Positioning — 34th pctl but 10 days stale; COT tonight ★★

Inventories — crude, gasoline AND distillate all 0th pctl; Cushing 19.4 ★★★★★

Macro — VIX 18.7, VIX–Brent corr +0.20: no contagion ★

Freight — BWET z +4.32, 99.9th pctl; arb still shut (−1.08) ★★★★

GSADF — even the July rip prints no explosive episode ★★

Policy — SPR releasing 5 mb/wk against the squeeze ★★★

Deferred curve — M12 flat: the far end votes for normalisation ★★★

Implication — high conviction: the squeeze is real, physical, and *prompt-local*. Every confirming dial lives in the front of the curve and the physical system; every non-confirming dial (macro, GSADF, deferred prices, agency forecasts) says the back end is anchored. Positions should be shaped for convergence: the question is not whether M1 meets M12 somewhere lower — it is when, and from what peak. judgment

4 · What the signals' own history says signal_backtest v0 point-in-time · de-clustered · base-rated

First implementation of the backtest spec: three of this pack's standing claims, tested against every historical firing. Two of three came back inconvenient — reported anyway.

Signal (firing rule)	n	Forward (21 td)	Median	Base rate	Edge	Honest read
crack_extreme (3-2-1 z>3, 2021–)	3 (2 cl.)	crack Δ$/bbl	+6.40	+0.64	n/a	Too few independent episodes to score — 3 onsets, all 2022, and after them the crack went higher, not lower. History offers no reversion evidence in either direction. The hedge stance rests on the asymmetry of an all-time-range margin (tail insurance), not on a backtested edge. (Restated r3 per amended spec: rolling de-clustering; the earlier "n=17" counted pseudo-replicated firings.)
freight_extreme (BWET z>3, 2024–)	3 (2 cl.)	Brent %	n/a	+1.0%	n/a	Not evidence-grade in either direction — 3 episodes, 2 independent, one regime. The v0 "n=23, edge +7.0%" was pseudo-replicated (overlapping forward windows; the spec's overlap rule was added by adversarial review after this finding). Ed. 1's "extremes mean-revert" stays withdrawn — and no opposite claim is asserted.
gasoline_undersupplied (below all prior-5-yr same-week, 1996–)	28 (28 cl.)	WTI %	−0.1%	+0.8%	−1.0%	The statistically sound one: 30 years, 28 fully independent episodes. The 0th-percentile gasoline print alone carries no bullish edge for crude — demoted from evidence to context; its weight comes only from the combination with curve and cracks.

Method: client-side v0 of signal_backtest.md — expanding-window (point-in-time) signal evaluation, ≥250-obs warmup, firings de-clustered into episodes (5-day gap; 35-day for weekly), forward window 21 trading days, base rate = unconditional same-horizon median, edge = conditional − base. Restated in r3 per the amended spec (adversarial review, same day): rolling de-clustering (a firing within the gap of the previous firing extends the episode) and the overlap rule (n reported with independent-cluster count at ≥-horizon separation; sub-cluster n never cited alone). Data: 9 series / 45,210 rows pulled via sparrow SQL from the serving node (history), spliced with prod for the current week. Server-side tool still pending — these numbers should be reproducible by it exactly.

5 · Book risk overlay — VaR, stress, and the mark illustrative book

The edition-1 sample book is up ≈ $24.8M in five sessions — the escalation scenario partially paid. That changes the risk question from "where can we lose" to "how much of this do we keep".

Sample position (from ed. 1)	Size	MTM since 17 Jul
Brent M1 flat length	+1.0 mb	+$9.7M
Bull calendar (M1–M6)	2.0 mb	+$15.1M (spread +$7.55)
Total		+$24.8M

Risk measure (1-day, today's vols)	$M
Portfolio σ (M1 σd $3.23, M6 $1.76, ρ 0.968)	6.3
VaR 95% / 99%	10.4 / 14.8
— flat-price line alone (95%)	5.3
— 2-mb calendar alone (95%)	5.2

Stress scenario	M1 / M6 move	P&L $M
Convergence (base case realised)	−18 / −10	−34
Squeeze extension	+12 / +6	+24
Fast normalisation to EIA's $74	−24 / −9	−54
Demand-collapse replay	−35 / −22	−61

The changed read: in edition 1 the calendar was the capital-efficient way in. At M1–M2 $6.02 it is a crowded exit: VaR on the calendar has risen ~35% (front vol 52%, and the M1×M6 correlation slipped 0.987 → 0.968 since mid-June — the decoupling is measured, not assumed), and the convergence stress — the base case — costs the unhedged book $34M. Banking half the spread P&L funds holding the rest with the house's blessing.

6 · Decision dashboard — with the evidence column earned

Expression	Stance	Conviction	Evidence (current)	Evidence (historical) new	Invalidation
Brent bull calendars	Trim half, hold core	HIGH	Earned +$7.55; M1–M2 $6.02 now prices the squeeze richly	—	M1–M2 < $3 (fade) · Cushing > 25 mb
Brent flat-price length	Hold; no adds (rule held from $90)	MEDIUM	All-stocks 0th pctl vs $74 consensus anchor 24% below	gasoline 0th pctl alone: edge −1.0% (n=28 independent, 30 yrs) — combination is the case, not the print	De-escalation confirmed at sea · COT shows crowd arrived
Short prompt structure	Still forbidden	HIGH	The last 5 sessions are the proof: −$4.24 of M1–M2 against shorts	—	Cushing rebuild > 25 mb first
Refining margin	Hedge — trigger fired	HIGH	z +3.25 crossed the z≈3 rule set in ed. 1; refiners at 96.1%, nothing left to give	n=3 (all 2022) — unscoreable (§4); hedge = free tail insurance against the fade tail	Crack z < 1.5 → re-add
Freight length	Stand aside (corrected grounds)	LOW	z +4.32 / 99.9th pctl — no edge either way for us	n=3, 2 independent clusters, one regime (§4) — "mean-revert" withdrawn; not evidence-grade either way	—
US→EU cargo fixtures	Uneconomic	HIGH	arb −1.08; export print confirmed it (−368 kb/d — the tiebreak we named)	—	BWET z < 2 and Brent–WTI > $4

The core position, updated: hold half the calendar, banked the rest

Edition 1's highest conviction earned +$7.55/bbl in five sessions. The trade now is discipline, not conviction. measured

Physical: all three stock classes 0th pctl; Cushing 19.4 and still drawing

Policy: SPR at 5 mb/wk is a bridge, not a fix — 311 mb left of a finite program

Curve: deferred anchored — convergence is the market's own exit plan

Entry economics gone: M1–M2 $6.02 pays you to reduce, not to add

COT tonight may reprice the "uncrowded" assumption

Standing invalidation (any two ⇒ cut to core): Cushing > 25 mb · BWET z < 2 · gasoline seasonal pctl > 20 · M1–M2 < $3.00 (tightened from ed. 1's <$1.00 — the entry moved, so the squeeze-fade tripwire moves with it, set where the spread still banks profit). Ed. 1's four triggers, scored on their original set (incl. the old M1–M2 < $1.00): 0 of 4 fired — the view survived its first week and got paid. judgment

7 · Market structure

The whole story in one picture: the front went vertical, the back never moved. M12 finished the week just $0.90 higher.

ICE Brent futures — front month vs 12th month

$/bbl, constant-maturity · daily · through 24 Jul 2026 — vintage note: M1–M5 carry 24 Jul settles, M6–M12 carry 23 Jul; intra-curve spreads (incl. M1–M12 $19.91) mix vintages worth ±$1 on this tape

Brent M1Brent M12

ICE Brent forward curve — now vs edition 1

Constant-maturity M1–M12, $/bbl

24 Jul17 Jul

Calendar spread	24 Jul	17 Jul
M1–M2	+6.02	+1.78
M1–M3	+9.82	+3.52
M1–M6	+14.63	+7.08
M1–M12	+19.91	+11.07

Key takeaway: the market pays ~$6/bbl per month at the front — 3.4× last week. Rolling shorts is ruinous; rolling longs is the richest carry of the cycle, and precisely because it is, it will not last. Every $1 of that carry is also $1 of convergence risk.

8 · Physical market — the policy bridge

Inventory (EIA, 17 Jul)	Level	w/w	vs 5-yr avg	Pctl	Read
US crude ex-SPR	411.7 mb	+2.0	−22.3 mb (−5.1%)	0	The build is an SPR transfer — see below
SPR	311.4 mb	−5.1	—	—	172 mb US / 400 mb IEA release program running
Gasoline	211.3 mb	+0.8	−15.8 mb (−7.0%)	0	Still below all of 2021–25, driving season
Distillate	109.6 mb	+1.4	−11.0 mb (−9.1%)	0	new dropped below the 5-yr range this week
Cushing	19.37 mb	−0.67	−10.1 mb (−34.4%)	0	Drawing again; 0.4 mb above the June low

The system balance, honestly stated: headline commercial crude built +2.0 mb — the first weekly print against the draw thesis, and we say so plainly. But the same week the SPR released 5.1 mb: commercial + SPR together drew 3.0 mb. The weekly S&D balance shows −0.8 mb/d current, −1.8 mb/d forecast. The correct frame: policy is renting the market time — commercial stocks are being defended by a finite government program while the physical system still runs a deficit. When the release program ends, either flows have normalised or the squeeze re-arms. measured

Cushing crude stocks — the WTI delivery point

Weekly, mb · Jan 2023 – 17 Jul 2026 · dashed = 5-yr week-29 average (29.5 mb)

Commercial read — arb still shut, and the print proved it. Arb score −1.08 (freight −1.00, curve −1.00, spread +0.37). Edition 1 named the weekly export figure as the tiebreak between our two disagreeing models; it came in at 3,353 kb/d, down 368 — the arb model won, the Theta export forecast lost. Desk action unchanged: no transatlantic fixtures on flat economics until BWET z < 2 and Brent–WTI > $4 (spread now $2.61 on fresh 20 Jul spots — the stale-spot caveat from ed. 1 is retired, fixed upstream).

9 · Refining — the trigger fired

Margin (vs 2020– history)	Current	z	17 Jul z	Status
3-2-1 crack	$65.45	+3.25	+2.99	extreme crossed the trigger
Gasoline crack	$56.70	+3.17	+2.90	extreme new
Diesel crack	$82.95	+2.68	+2.47	elevated

Utilisation 96.1%, no turnaround activity (trend scan clean), ICE Gasoil +12% on the month and now current in the data (staleness fixed). Edition 1 wrote the rule: "hedge incremental margin above crack z ≈ 3." That is no longer advice — it is an instruction the desk set for itself a week ago, now in force.

Why this matters: the backtest (§4) keeps the hedge honest — history offers too few independent crack-extreme episodes to score reversion at all (n=3, all 2022), so no claimed edge is being given up; what the hedge buys is protection against the fade tail, which is exactly the "product-led fade" risk sitting at #3 on this edition's list. Free insurance at an all-time-range margin. Take it.

10 · Positioning — the stale pillar

Managed-money net positioning — WTI (NYMEX) vs Brent (ICE)

Contracts · weekly CFTC · 2 Jun – 14 Jul 2026 — predates the entire squeeze; new COT ~19:30 UTC today

WTI MM netBrent MM net

Everything positioning-shaped in this pack rests on a 14 Jul print — before M1 ran $84.73 → $100.69. The last known state: WTI MM at the 34th percentile, Brent MM net short −24.2k. If tonight's report shows specs piled in during the rip, the "fundamental, not crowded" pillar — the reason pullbacks were expected shallow — breaks, and the convergence scenario gets faster and deeper. This is the single most important data release between now and the FOMC. Risk to monitor, not a conclusion.

11 · Qualitative overlay — the story, corrected and current news · outside the audited quant core

Edition 1, written from thin press, said "US–Iran conflict raises fears of disruption." The record is now established and two of its facts change the frame:

The Strait of Hormuz was closed from late February (the platform's signal constellation flagged the supply shock on 9 April, five days before the cause was public — the regime this pack has traded all along). The conflict ended 18 June with a US–Iran MOU; strait traffic is increasing.
EIA, 7 July: raised global production forecasts on the reopening, expects flows near pre-conflict levels by year-end, and cut its Q3 Brent forecast $27 to $74. June Brent averaged $85, down $32 from the April peak.
The SPR release in our weekly data is the announced program: 172 mb US release begun 16 March, within a ~400 mb coordinated IEA action across 31 countries — the policy bridge in §8, sized and dated.
The July squeeze is therefore a post-conflict scarcity event, not new escalation: the damage already done to inventories, restart logistics, and war-priced freight — colliding with a consensus that says this normalises. That collision is the −$19.91 curve.

Sources: EIA press release, 7 Jul 2026 — production forecast raised after Strait of Hormuz reopening · FinancialContent — 172 mb SPR deployment (16 Mar 2026) · FinancialContent — IEA 400 mb coordinated release · Trading Economics — Brent near $91 on reserve release · Hormuz closure tracker

12 · Where we could be wrong

The $74 anchor is 24% below the front

EIA's Q3 average forecast is $74. If normalisation runs fast — flows are already increasing — the fast-normalisation scenario (20%) is worth −$54M on the sample book. The biggest single risk is no longer a shock; it is the consensus being right on schedule.

We cannot see what protection costs

No options dataset — with the front at $97.84 and convergence the base case, the 25Δ put skew is the one number that would price the exit crowd. Still the platform's most valuable missing dial, and this edition is the proof case.

Positioning is 10 days blind + backtest caveats

The COT gap (§10) is the known unknown of the week. And §4's freight result is one regime wearing 23 raw firings (3 episodes, 2 independent) — we corrected our prose with it, we do not trade its edge. Small-n honesty cuts both ways.

13 · House view — updated

Committed view · high conviction on the physical state, disciplined on the exit · analyst judgment on measured evidence

Monetise the squeeze, keep the core, prepare for convergence

Edition 1's view — constructive via the curve, not fresh flat-price — survived its first week with zero invalidation triggers fired and got paid +$7.55 on the calendar. The update is about the exit, not the thesis: the physical system is still in deficit under a finite policy bridge (SPR −5.1 mb/wk), but the market now pays $6/bbl per month at the front while its own deferred prices and the agency consensus vote for normalisation. We trim half the calendar into that richness, hold the core with a raised fade-tripwire (M1–M2 < $3), execute the crack hedge our own z>3 rule triggered, add no flat-price length at these levels, and keep prompt shorts forbidden until Cushing rebuilds. If the squeeze extends, the core participates; if convergence starts, the banked half already won.

Standing invalidation triggers (any two ⇒ cut view to core): Cushing > 25 mb · BWET z < 2 · gasoline seasonal pctl > 20 · M1–M2 < $3.00 (tightened from ed. 1's $1.00 — entry moved; disclosure in §6). Next checks: COT tonight (~19:30 UTC) · API 28 Jul · EIA weekly + FOMC 29 Jul · STEO 11 Aug · IEA OMR + OPEC MOMR + CPI 12 Aug (Q2 production actuals — the outage's true depth, and the agency layer's next refresh).

Report performance — the publication scoring itself standing panel from this edition

Every edition ends by scoring the previous one. Scoring is pre-registered, never retrospective: trigger-based calls score mechanically against thresholds frozen at publication (scoreboards name their rule set); prose claims are editorially scored and flagged as such. From this edition, every new call ships with its scoring rule attached.

Ed. 1 call (23 Jul)	Scoring	Result
Invalidation triggers (4, original set incl. M1–M2 < $1.00)	Mechanical — pre-registered thresholds	0 of 4 fired — view held and paid
Sample book (flat length + bull calendar)	Mechanical — marked to market	+$24.8M in 5 sessions (§5)
Crack-hedge rule (3-2-1 z > 3 ⇒ hedge)	Mechanical — pre-registered in ed. 1	Fired 24 Jul; hedge enforced (§9)
"Freight extremes mean-revert" (prose claim)	Editorial → backtested §4	Withdrawn — not evidence-grade in either direction
"Record-low gasoline stocks are bullish crude" (prose claim)	Editorial → backtested §4	Refuted (edge −1.0%, n=28) — demoted to context
"Extreme cracks mean-revert" (prose claim)	Editorial → backtested §4	Unscoreable (n=3, one regime) — stance restated on tail asymmetry, not history
Scenario prior 50/30/20	Mechanical from this edition — Brier, calibration ledger	Pending — entry 1 resolves 24 Aug 2026

Tally: 7 calls evaluated — 3 mechanical (triggers 0/4 fired, book +$24.8M, crack rule fired→enforced), 3 editorial via the engine's own backtest (none survived as stated: withdrawn / refuted / unscoreable — the engine grading its own prose), 1 pending calibration. Registered for ed. 5 scoring: the §13 trigger set (mechanical) · trim-half calendars (against M1–M2 at next edition) · crack hedge (against the 3-2-1 level) · scenario prior (Brier at 24 Aug — calibration ledger entry 1).

Appendix A · EnergyScope MCP tools used (this edition)

Tool	Calls	Purpose
latest	1	23-series snapshot; staleness flags all clear (pipeline fix verified)
term_structure	1	Curve M1–M12: −$19.91, roll −73.8% (§7)
seasonal	4	Week-29 context: crude, gasoline, distillate, Cushing — all 0th pctl (§8)
crack_spread	3	3-2-1 z +3.25 (trigger fired), gasoline +3.17, diesel +2.68 (§9)
spread / volatility / arb_signal / snd_balance	1+2+1+1	Brent−WTI on fresh spots; M1/M6 GARCH for VaR; arb −1.08; balance −0.8/−1.8 mb/d
quantile_reg / bubble_test / maintenance_signal	3	Scenario anchors; GSADF still March-only; no turnaround distortion
percentile / correlation / changes / upcoming_events	3+1+1+1	BWET 99.9th, WTI price 88.8th, production 97.5th (stopped firing); VIX corr +0.20; 30-d movers; catalysts
get_data	2	Prod-node chart splice (serving-node lag) + M12 tail
derived_signals	1	39 signals with prev_metrics → 15 active, 3 new / 1 stopped / 3 intensifying, contradiction → extreme (§2, §3)
save_watchlist	1	New: universe pinned as monthly_risk (27 series) — every future edition sweeps identically
save_report_run / compare_runs	1+1	This edition saved once; compare_runs executed live post-save (result quoted in delivery note)
sparrow CLI (non-MCP)	1	One SQL batch: 9 series / 45,210 rows → chart data (no hand transcription) + the §4 backtests
WebSearch (non-EnergyScope)	1	§11 overlay — MOU, EIA forecast revision, SPR program; cited inline, fenced from quant core

Process notes: the monthly_risk report preset is staged in the MCP package but unreleased — this sweep was hand-run one final time; from the next release the data phase is report(watchlist="monthly_risk", preset="monthly_risk"). OPEC MOMR / IEA OMR values carry their July vintages (next refresh 12 Aug); CFTC carries 14 Jul (release cycle, not staleness).

Appendix B · Data series & vintages

Series	Used for	Vintage
ICE.BRENT.M1…M12.D	Curve, squeeze, VaR, charts	23–24 Jul
PET.RWTC.D / PET.RBRTE.D	Spots, Brent−WTI (fresh — caveat retired), backtest targets	20 Jul
PET.EER_* products / ICE.GASOIL.M1.D	Cracks (gasoil staleness fixed: current)	20–24 Jul
PET.WCESTUS1/WGTSTUS1/WDISTUS1/W_EPC0_SAX…/WCSSTUS1/WCRFPUS2/WPULEUS3/WCRIMUS2/WCREXUS2/WCRRIUS2 (.W)	Stocks incl. SPR, S&D, exports tiebreak, utilisation	17 Jul
CFTC.WTI/BRENT.MM_Net	Positioning (predates squeeze — §10)	14 Jul ⚠
SHIP.BWET.D	Freight z +4.32 / 99.9th pctl; backtest	23 Jul
MACRO.VIX/DXY/TNX · NG.RNGWHHD.D	Macro overlay	20–24 Jul
OPEC.MOMR* / IEA.OMR_T1*	Agency balance (unchanged this week; production actuals still Apr)	Jul MOMR/OMR
Backtest set (9 series, 45,210 rows via sparrow)	§4 — WTI/products to 1986, gasoline stocks to 1990, BWET full	serving node (≈10-day lag, history only)

Reproducibility statement: every quantitative claim outside §11 traces to a named tool call or the sparrow SQL batch above; §11 is press-sourced and cited, informing judgment only. §4's backtests are point-in-time with de-clustering and base rates per the signal_backtest spec; §5's book is illustrative with exact arithmetic. Measured values and judgment are tagged throughout. Known gaps unchanged from edition 1 (options/implied vol first among them — §12 is this edition's demonstration of why).