THE BARATELLI INSTITUTE
· Mentoring at Scale
Home
THE BARATELLI INSTITUTE
· Mentoring at Scale
Prices, reserves, production, rigs, inventories, refining, breakevens, chokepoints and the currency the barrel is quoted in — 13 ledgers on the numbers that run the oil market, in plain English, in sortable tables, with the date each one was measured printed beside it.
Oil is the most consequential commodity on the planet and its statistics are scattered across a dozen institutions that publish on different schedules, in different units, with different definitions of the same word. A reader who wants the reserve figure for Iraq, the rig count for last week, and the level of the American strategic reserve currently visits three organisations and reconciles three vocabularies. This page exists so that reader does not have to.
It is built the way the Institute builds its acquisition ledgers and its franchise valuation references: as a living reference rather than an article. The tables sort. Every figure names the body that measured it. Every derived ratio is computed here from the figures shown, not carried in from somewhere else. And every number carries the date it was measured, which is not the date this page was published — a distinction that matters more in oil right now than in any other market, because almost everything is moving.
Prepared by The Baratelli Institute · publication date October 24, 2026. Every figure below carries its own measurement date, which is set by whoever measured it and is often older than this page.
Twelve headline figures, each on its own clock. Sort any column.
| Figure | Figure | Measured | Source | What it carries |
|---|---|---|---|---|
| Global oil supply | 102.3 mb/d | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Rose 2.5 mb/d on the month, reversing August's 1.6 mb/d fall and then some. OPEC+ supplied 2.1 mb/d of the gain, led by Saudi Arabia. The IEA's own caution on it is one sentence: renewed attacks in the Gulf leave the recovery fragile. |
| OPEC+ effective spare capacity | 0.27 mb/d | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | 0.26% of world supply, and up from 0.22 mb/d in August — which had itself fallen 80% from July's 1.09 mb/d. So the 2.5 mb/d recovery in supply bought the market almost nothing in the way of a cushion, which is the point of carrying this row beside the supply row. The composition moved the other way from the total: the OPEC-8 subset holds 0.10 mb/d against 0.07 in August, while the non-OPEC half fell to 0.11 from 0.15. This remains the more generous of the two available reads: no other institution publishes a monthly effective spare capacity series, and the EIA's quarterly OPEC surplus-capacity series — a narrower aggregate, and a forecast rather than an observation — sits at 0.0 mb/d for this quarter and the next. The IEA's next edition publishes November 13, 2026. |
| Gulf production shut in | more than 10 mb/d | August 2026 | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | The largest supply disruption since the 1970s. 🔴 This row is deliberately a month behind the two above it: the October 2026 edition does not restate a Gulf shut-in volume in the material read, so the figure and the citation both stay on the September edition rather than being carried forward onto a clock no publisher set. The IEA states a bound rather than a point estimate, against 8.3 mb/d for July. The EIA's read on the same month is 6.72 mb/d against a narrower baseline, up from 4.98 in July; both are carried in the balance table below with the conflict stated. A bound and a point estimate are not the same kind of figure and this reference does not convert one into the other. |
| World proven reserves | 1,572.459 bn bbl | year-end 2025 | OPEC Annual Statistical Bulletin 2026, Table 3.1 (published 2026) | 42.1 years of reserve life at September 2026 supply, computed on this page. |
| United States rigs drilling | 603 rigs | week ended October 23, 2026 | Baker Hughes North America Rig Count, released October 23, 2026, as reported by the trade press on the day of release — see the provenance note, which records that BOTH of the publisher's own surfaces, the report file listing and the summary table, still displayed 10/16/26 when this was read | 460 drilling for oil, 132 for gas, 11 other. |
| United States commercial crude stocks | 412.5 mn bbl | week ended October 16, 2026 | EIA Weekly Petroleum Status Report, Table 4, published October 21, 2026 | At the five-year average for the week — it ran about 1% above it the week before. Excludes the strategic reserve. |
| United States Strategic Petroleum Reserve | 282.7 mn bbl | week ended October 16, 2026 | EIA Weekly Petroleum Status Report, Table 4, published October 21, 2026 | 38.9% of the 727 mn bbl authorised capacity. Lowest since November 1982. |
| Global observed oil stocks | just under 7,900 mn bbl | end-July 2026 | IEA Oil Market Report, published August 12, 2026 | The last LEVEL published, and now two months behind the change row below it by design — see that row and the inventories ledger. |
| Global observed oil stock change | fell 10 mn bbl | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | A 0.3 mb/d draw, and 509 mn bbl end-February to end-September 2026, an average of 2.4 mb/d. The net is the small part: on-land stocks fell 113 mn while oil on water rose 103 mn, non-OECD crude fell 76 mn with 78% of that in China, and OECD stocks fell 37 mn — the opposite sign to August, when the OECD was the one component building. The IEA states that on-land stocks are at their lowest since at least 2016. |
| Brent crude, front-month futures | $113.12 / bbl | opening quote, October 23, 2026 | Forbes Advisor daily crude quote, page dated October 23, 2026 (page stamped 'Audited & Verified: Oct 23, 2026, 1:35pm') | Reached $126 on April 30, 2026, a four-year high. The physical North Sea Dated assessment for the same day is quoted far higher — see the price ledger, where both are printed and the instrument is named on each. |
| WTI crude, front-month futures | $99.36 / bbl | opening quote, October 23, 2026 | Forbes Advisor daily crude quote, page dated October 23, 2026 (page stamped 'Audited & Verified: Oct 23, 2026, 1:35pm') | Posted a 35% weekly gain in early March, the largest since 1983. |
| North Sea Dated (physical assessment) | about $134 / bbl | October 14, 2026, at time of writing | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | The physical assessment, and roughly twenty-seven dollars above the ICE Brent futures quote on the same date — against a fourteen-dollar gap on September 9. A large physical premium to paper is the price signature of prompt scarcity, and it has not closed; it has roughly doubled. Quoted by the IEA “at the time of writing”, which is a publication-clock figure rather than a measured close. |
| Global refinery throughput | 81.4 mb/d | August 2026 | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | Roughly 79% of world capacity. 🔴 Deliberately still on an August clock: the October 2026 edition speaks about refinery runs forward rather than publishing a measured September throughput, and a forecast for the current month is not a substitute for a measurement of the last one. Margins are no longer uniform — US margins at new highs; Europe and Asia crushed as of September 2026 — so a single record-margin claim no longer describes the picture. The utilisation ratio has a dated numerator and an undated denominator; see the caveat on the refining ledger. |
| United States crude production | 13.7 mb/d | calendar year 2025 | EIA Short-Term Energy Outlook, released October 6, 2026 (forecast completed October 1, 2026; next release November 10, 2026) | A record, and the highest annual crude output any country has ever produced. An actual that moved: the August 2026 edition carried 13.6 mb/d for 2025 and the September 2026 edition revised it to 13.7, where the October edition holds it. A completed calendar year is not a settled number while the survey data behind it is still being revised. |
Read the Measured column before quoting anything from this table. Three of these figures are weekly, five are monthly, two are annual, one is a daily quote, and one — world proven reserves — is a year-end 2025 figure published in 2026 and describing a stock that is revised once a year. Quoting them against one another without regard to their dates is the single most common error in energy commentary, and in a period when supply has moved by several million barrels a day inside a quarter it is not a small error.
Each one is a standing page that gets refreshed, not an article that goes stale. The bar at the top of every page reaches all of them from anywhere in the reference.
Where every cent of a $4.48 gallon goes — crude, refining, distribution and tax — with each of the four components on its own clock and three of the four defined differently from how a reader assumes.
Open the ledger →LedgerSeventy-six years of crude prices deflated to constant September 2026 dollars. The 2008 records are still records in real terms and today's are not close.
Open the ledger →LedgerOne central bank, one date, one fixing time. A dollar barrel converted into 29 currencies without mixing thirty sources and thirty clocks.
Open the ledger →LedgerWhat the shape of the curve does to inventory and why it drives storage behaviour rather than describing it — then the full strip: 65 listed WTI months from the exchange's own Daily Bulletin, $97.97 front against $80.13 a year out, $17.84 of backwardation, open interest printed beside every settlement.
Open the ledger →LedgerTwo entirely different questions that share one word. The price a well needs and the price a treasury needs, kept rigorously apart — with the gaps in each named.
Open the ledger →LedgerAll 41 countries the OPEC bulletin itemises, ranked, with share of world total and reserve life computed two ways. Venezuela first at 303.701 billion barrels; the United States ninth with a reserve life of 8.4 years.
Open the ledger →LedgerEvery OPEC+ member's output for September 2026 against its implied target and its sustainable capacity. Effective spare capacity is 0.27 mb/d, or 0.26% of a 102.3 mb/d market.
Open the ledger →LedgerThe seven countries that hold the voluntary cuts — not eight, and not the full membership — their August and September 2026 decisions, and why quota discipline is currently the wrong lens.
Open the ledger →Ledger603 rigs turning in the United States, 460 drilling for oil, 1,919 worldwide. A forty per cent move in price has produced a 0.0 per cent move in the American count.
Open the ledger →LedgerCapacity measured on two clocks five months apart, crack spreads computed from same-week legs in both units the market quotes, and the one undated figure on this reference, labelled as such.
Open the ledger →LedgerDiesel at $6.276 a gallon, below a nominal record of $6.529 that was never a real one. Why distillate is tighter than gasoline, and why distillate fuel oil is not diesel.
Open the ledger →LedgerCommercial crude one per cent above its five-year average in the same week the Strategic Petroleum Reserve reads 38.9% of capacity and its lowest since November 1982. The two facts are causally linked.
Open the ledger →LedgerEight passages, six years of flows each, and bypass capacity that covers under a quarter of Hormuz. Plus what the industry itself expected in April, measured, and how wrong it was.
Open the ledger →The reference is organised in four families. Price and cost answers what a barrel costs and what that means at a pump, in constant dollars, in another currency, and in a later month. Supply and cost answers who holds the oil, who is producing it, what it costs them to produce, and whether anyone is drilling for more. Downstream answers what happens to the barrel after it is bought. Movement answers whether it can get to where it was sold. Almost every misreading of this market comes from answering a question in one family with a number from another.
Taken one at a time these series are a reference. Taken together they describe a single condition, and the Institute's reading of it is below. Each claim links to the ledger that carries the figures behind it.
The market has lost its shock absorber. Effective spare capacity across all of OPEC+ is 0.27 mb/d against world supply of 102.3 mb/d — 0.26%. The subset of countries that manages the voluntary cuts holds 0.10 mb/d between them, and there are 7 of them, not eight: the International Energy Agency still labels the group “OPEC-8” in its August 2026 report, while OPEC's own August and September 2026 releases name Saudi Arabia, Russia, Iraq, Kuwait, Kazakhstan, Algeria and Oman and no longer name the United Arab Emirates. The quota ledger carries both labels rather than choosing between them. For most of the last twenty years the equivalent figure ran between two and four million barrels a day. When cover is three million barrels, an outage is a logistics problem and someone opens a valve. When it is one, the only remaining adjustment mechanism is demand, and demand adjusts through price. See the production ledger.
Reserves and the ability to produce them have come apart. Saudi Arabia, Iran, Iraq and Kuwait hold 45.6% of the world's proven reserves and in September 2026 produced 15.83 mb/d against 23.66 mb/d of sustainable capacity — a gap of 7.83 mb/d. Through August 2026 that bottom-up figure reconciled with the shut-in volume the IEA states from the top down, and this month it stopped: the agency now states more than 10 mb/d and these four countries account for 7.83, so the bottom-up route falls at least 2.17 mb/d short of the top-down floor. The Institute does not close that gap by arithmetic, and reads it as the shut-in having spread past the four largest reserve holders. A reserve ranking now tells a reader who owns the oil and nothing about who can ship it this quarter. See the reserves ledger, which carries both routes and the shortfall between them.
American drilling did not respond, and that retires a decade-old assumption. The rig count went from 603 to 603 across roughly eight weeks at a crude price near a four-year high, and did not move at all in the most recent two. The belief that shale would cap any spike rested on a response time of two to three quarters; this disruption has run since March, and the count has moved by 0. What the last decade built was a supply source with a fast financial response and a slow physical one. See the rig ledger.
The most-watched inventory number is currently the least informative. United States commercial crude reads about one per cent above its five-year average because 172 million barrels of strategic reserve were released into it. The comfortable number is comfortable because the uncomfortable one paid for it, and during a release the two series are not independent. See the inventory ledger.
The constraint is upstream and in transit, not in the plants. Atlantic Basin refining margins set all-time records in July and August at global refinery utilisation of roughly 79%. Refiners earning record margins while running well below capacity is the signature of crude failing to arrive, and July's global stock draw was almost entirely oil on water rather than oil on land. Building refineries does not address this.
| Benchmark | Figure | Measured | Source | What it carries |
|---|---|---|---|---|
| Brent (futures, front month) | $113.12 / bbl | opening quote, October 23, 2026 | Forbes Advisor daily crude quote, page dated October 23, 2026 (page stamped 'Audited & Verified: Oct 23, 2026, 1:35pm') | Brent is the seaborne light-sweet benchmark most of the world prices against. This is the futures open, not the Dated Brent physical assessment quoted in the trade press — the two are different instruments and do not have to agree. 🔴 THE TWO BENCHMARKS DIVERGED IN SIGN ACROSS THIS STEP AND THE SPREAD WIDENED BY SEVENTY-NINE PER CENT OF ITSELF. This row went $109.13 to $113.12, UP $3.99 or 3.656 per cent, while the WTI row below went $101.44 to $99.36, DOWN $2.08 or 2.050 per cent — opposite signs, over the identical interval, off the identical edition of one publisher’s page. The Brent-less-WTI open spread went $7.69 to $13.76, a widening of $6.07, which is 78.9 per cent of its own prior level in four calendar days. Read with the demoted entry below, which records two consecutive observed sessions of sign divergence followed by one of agreement, this is a fourth divergence on the record. The run is reported, not explained; no cause is offered. 🔴 AND THIS STEP IS NOT DECOMPOSABLE AT A PRINTED FINAL SETTLE, WHICH IS A WEAKER POSITION THAN THE ENTRY BELOW HELD AND IS SAID SO RATHER THAN PAPERED OVER. October 19 was a Monday and October 23 a Friday: four calendar days with three intervening sessions — October 20, 21 and 22 — plus the October 23 session itself. The only bulletins read across that interval are #202, trade date Tuesday October 20, and #205, trade date Friday October 23, and BOTH are marked PRELIMINARY on all ten pages with FINAL on none. So no FINAL settlement brackets any interior session, and the three-leg decomposition the entry below performs cannot be performed here. What this row carries is a level-to-level change between two dated opens: firm at both endpoints, unsplit in the middle. 🔴 AND THE REASON IS STRUCTURAL, NOT A FAILURE OF THE READING. The exchange marks a bulletin FINAL at 10:00 a.m. Central Time on the next business day and finalises it IN PLACE — same bulletin number, same trade date, same URL, PRELIMINARY overwritten by FINAL — so there is no separate FINAL document for the /current/ path to rotate to. This sweep ran on a Saturday. Friday October 23’s FINAL cannot exist before Monday October 26, and no weekend sweep can ever reach a FINAL for the preceding Friday. That is a property of the publisher’s calendar, and it is the reason the forward-curve strip is unchanged on this sweep rather than any doubt about what #205 prints. 🔴 THE PUBLISHER’S OWN INTRADAY LINE LOCATES THE WTI FALL OUTSIDE THE SESSION THE DATE NAMES, AND THAT IS ARITHMETIC ON A ROUNDED PERCENTAGE RATHER THAN A SETTLEMENT. The same page prints an intraday WTI of $101.46 at 8:22 a.m. ET, ‘up 2.06 per cent since yesterday’s close’, which puts the October 22 close at about $99.41 — within five cents of the October 23 open of $99.36 this reference adopts. So the $2.08 fall belongs to October 20, 21 and 22, and essentially none of it to October 23. The reconstruction is limited by the two decimals the publisher printed on that percentage and is not a price anybody settled at; it is recorded as a LOCATED leg, not a measured one. 🔴 THE INTRADAY QUOTES SIT ABOVE THEIR OPENS ON THIS EDITION, WHICH IS THE REVERSE OF THE ENTRY BELOW. Intraday Brent $114.93 and intraday WTI $101.46 against opens of $113.12 and $99.36. On the October 19 edition both intraday figures sat BELOW their opens. A reader taking the intraday line as the day’s figure would record WTI rising where this row records it falling. Neither intraday figure is carried on any Institute page. And the two intraday lines are on different bases again — Brent’s reads ‘in the past 24 hours’, WTI’s ‘since yesterday’s close’ — which is the same asymmetry a demoted entry on this row recorded on October 2 and which no sweep has yet seen the publisher correct. 🔴 THE WRITES ON THIS SWEEP WERE TOOL-SEQUENCED WITH NO BUILD INSIDE THEM, WHICH IS WHAT THE ORDERING REMEDY REQUIRES AND IS NOT THE SAME CLAIM AS ‘ONE PASS’. This row, the WTI row below, FX_SRC and FX_ASOF, the twenty-nine fixing rates and the two FX_CRUDE floats were written as consecutive edits with no build and no verification run between any two of them, so no interval exists in which a build or a reader could observe a disagreeing pair. The editing interface applies one edit per call, so the literal words ‘one pass’ would be false and are not used. The owed two-line assertion — that FX_CRUDE’s floats and measurement strings must equal the Brent and WTI head rows of PRICES — is still owed, is what would make a future green build evidence rather than luck, and remains out of scope for the daily task, which permits edits to oil_data.py only. 🔴 SUPERSEDED AND LABELLED — WHAT FOLLOWS DESCRIBES THE OCTOBER 15 TO OCTOBER 19 STEP, WHICH THIS ENTRY RETIRES. It is retained rather than deleted because it is the other term of a difference a reader must be able to check, because its sign-divergence run is cited above, and because its fully-closed three-leg decomposition is the standard this step could not meet. Its figures are correct as of their own measurement dates and are not the figures this row now prints. 🔴 THE STEP IS FULLY DECOMPOSABLE FOR THE FIRST TIME ON THIS ROW AND IT CLOSES TO THE CENT, BUT ITS MIDDLE TERM IS PROVISIONAL AND ITS ENDPOINTS ARE NOT. The printed number went $105.99 to $109.13, UP $3.14 or 2.963 per cent, and the WTI row below went $95.16 to $101.44, UP $6.28 or 6.599 per cent. October 15 was a Thursday and October 19 a Monday: four calendar days, but only two market sessions inside them — Friday October 16 and Monday October 19 — because October 17 and 18 are the weekend. Three legs bracket it, and two of the three are printed settles rather than inferences. (a) The October 15 session, FINAL: bulletin #199, trade date Thursday October 15, read on October 17 and marked FINAL on all ten pages, prints BZ DEC26 settling $105.92 against this row’s own October 15 open of $105.99 — Brent fell 7 cents intraday. (b) The October 16 session, PRELIMINARY: bulletin #200, trade date Friday October 16, read on October 18 and marked PRELIMINARY on all ten pages, prints BZ DEC26 settling $109.86, a point change of + 3.94. (c) From that settle to the October 19 open: $109.86 to $109.13, DOWN 73 cents, which is the weekend and Monday’s pre-open taken together and is the one leg this reference did not observe at a printed price. The three sum to the printed figure exactly: minus 0.07 plus 3.94 minus 0.73 is plus 3.14. 🔴 AND WHAT IS PROVISIONAL HERE IS THE SPLIT, NOT THE TOTAL. Both endpoints of the step are quoted opens off dated editions of the publisher’s page and neither is provisional, so the $3.14 is firm. Only legs (b) and (c) depend on #200’s preliminary settlement. If the exchange revises that settlement, (b) and (c) move against each other by the same amount and the total does not move at all. Say which of those a figure is before relying on it. 🔴 AND THE PROVISIONAL FIGURE IS NOT MERELY ASSERTED — IT BACKS OUT TO A FINAL PREDECESSOR TO THE CENT, WHICH IS AN IDENTITY TEST ON THE CAPTURE AND NOT A PRICE FINDING. #200 prints BZ DEC26 at $109.86 with + 3.94, and $109.86 less $3.94 is $105.92, which is #199’s FINAL settlement. The same holds on the other leg: #200 prints CL NOV26 at $101.41 with + 3.44, and $101.41 less $3.44 is $97.97, #199’s FINAL figure. A provisional number that reconciles to a FINAL predecessor is a different object from one that does not, and that is what licenses using it as the middle term of a decomposition under a label. It licenses nothing further: NOTHING FROM #200 IS ADOPTED INTO THE FORWARD-CURVE STRIP, which stays on #199 and October 15 for the reason recorded in the curve ledger. 🔴 THE WHOLE OF THE MOVE IS ONE SESSION’S WORK, AND IT IS NOT THE SESSION THE DATE GAP WOULD SUGGEST. Friday October 16 carries + $3.94 of Brent’s + $3.14 total and + $3.44 of WTI’s + $6.28. October 15 contributed minus 7 cents and plus $2.81; the weekend and Monday’s pre-open together contributed minus 73 cents and plus 3 cents. A reader who takes a four-day gap as four days of drift would be wrong about both rows, and would be most wrong about WTI, where the weekend moved the price by three cents. 🔴 THE TWO BENCHMARKS DIVERGED IN SIGN ON OCTOBER 15 AND THEN MOVED TOGETHER ON OCTOBER 16, ON THE EXCHANGE’S OWN CHANGE COLUMN RATHER THAN ON ARITHMETIC. #199 prints BZ DEC26 − 0.14 against CL NOV26 + 2.88 — opposite signs inside one session, confirmed by the source document. #200 prints BZ DEC26 + 3.94 against CL NOV26 + 3.44 — the same sign, within 50 cents of each other. Read with the demoted entry below, which records Brent falling 48 cents while WTI rose 77 on October 13, that is two consecutive observed sessions of sign divergence followed by one of agreement. The run is reported, not explained; no cause is offered for any of the three. 🔴 THE SAME-MONTH DECEMBER SPREAD WIDENED SHARPLY, AND IT IS THE COMPARISON THAT SURVIVES A ROLL. BZ DEC26 less CL DEC26 is $105.92 less $96.50, or $9.42, at #199 (FINAL), against $109.86 less $99.02, or $10.84, at #200 (PRELIMINARY) — a widening of $1.42, which is 15.1 per cent of its own prior level, in one session. Recomputed here off both extracts on disk rather than carried forward, and the second term is provisional with the first. 🔴 THE OPEN AND THE INTRADAY QUOTE ON THIS SAME PAGE ARE DIFFERENT INSTRUMENTS, AND ON THIS EDITION THEY POINT OPPOSITE WAYS. The October 19 page carries, beside the opens adopted here, an intraday Brent of $108.13 (− 1.57 per cent) and an intraday WTI of $100.41 at 8:27 a.m. ET (− 0.97 per cent). Both sit BELOW their own opens. A reader who took the intraday line as the day’s figure would record a falling market where this row records a rise of 2.963 per cent. Neither intraday figure is carried on any Institute page; they are named here so the distinction is on the record rather than available to be rediscovered. 🔴 THE ROLL HAS WIDENED FURTHER AND THE FRONT LABEL STILL HAS NOT MOVED, WHICH IS #199’S OWN PREDICTION ARRIVING. CL NOV26 open interest ran 84,922 to 62,967, down 21,955 lots or 25.9 per cent, while CL DEC26 ran 309,217 to 311,403. December now exceeds November by 248,436 lots, a ratio of 4.95x against 3.64x one session earlier. BZ DEC26 ran 34,132 to 35,002. Both fronts are nonetheless unchanged in #200 — CL NOV26 and BZ DEC26 — so the label has not yet rolled; the next sweep should expect it to, and should expect the apparent price jump that a roll produces to be a change of contract rather than of price. 🔴 THIS SWEEP’S WRITES WERE SEQUENCED, NOT SINGLE-PASS, SO THIS ROW IS NOT EVIDENCE THAT THE ORDERING REMEDY HOLDS. The comment above FX_CRUDE records that the remedy for the barrel-against-currency disagreement is not a reminder but the ORDERING: the Brent head row, the WTI head row, the twenty-nine fixing rates and the two FX_CRUDE floats written in one pass, so that no interval exists in which any pair of them disagrees. On this sweep the mount’s pinned-length truncation fired on an intermediate write and forced a whole-file reconstruction between the rate write and this one, so the writes were SEQUENCED. No build ran in the interval and nothing shipped inconsistent, but the words ‘one pass’ would be false for this sweep. The owed two-line assertion — that FX_CRUDE’s floats and measurement strings must equal the Brent and WTI head rows of PRICES — is still owed, and remains the only thing that would make a future green build evidence rather than luck. 🔴 SUPERSEDED AND LABELLED — WHAT FOLLOWS DESCRIBES THE OCTOBER 13 TO OCTOBER 15 STEP, WHICH THIS ENTRY RETIRES. It is retained rather than deleted because it is the other term of a difference a reader must be able to check, and because its sign-divergence finding is cited above. Its figures are correct as of their own measurement dates and are not the figures this row now prints. 🔴 THE PRINTED STEP IS HALF DECOMPOSABLE AND THE DECOMPOSITION REVERSES ITS SIGN. The printed number went $106.17 to $105.99, DOWN 18 cents or 0.170 per cent, and the WTI row below went $93.37 to $95.16, UP $1.79 or 1.917 per cent. October 13 was a Tuesday and October 15 a Thursday: two calendar days and two trading sessions inside them, October 14 and October 15. What brackets the first part of the step is a printed settle rather than an inference, and for the first time on this row the bracketing bulletin covers the very session whose open the previous sweep carried. Bulletin #197, trade date Tuesday October 13, was read on October 15 and is marked FINAL on all ten of its pages on first reading — it prints BZ DEC26 settling $105.69, which is 48 cents BELOW this row's October 13 open of $106.17. So Brent FELL 48 cents intraday on October 13 and then ROSE 30 cents from that settle to the October 15 open, and the two legs sum to the printed −18 cents exactly. A reader taking the printed −18 cents for a two-day decline has the later and larger window backwards: the fall is entirely in a session that had already closed when the previous sweep published, and everything since has been a rise. This is the same shape as the WTI finding on the October 6-to-9 step recorded further down, arriving on the other benchmark, and it is the plainest argument on this page for carrying the decomposition rather than the printed pair. 🔴 WHAT CANNOT BE DONE IS SPLITTING THE REMAINDER, AND THE REASON IS THE INSTITUTE'S OWN SWEEP CADENCE RATHER THAN THE PUBLISHER'S. The 30-cent remainder spans the full October 14 session plus the October 15 overnight and opening, and nothing here divides it. This reference holds no October 14 open: the sweep of October 14 read this publisher's page and found it still stamped 'Audited & Verified: Oct 13, 2026, 8:59am', and the sweep of October 15 found it stamped October 15. Whether an October 14 edition was ever published is NOT OBSERVED and is not claimed in either direction. The page serves one edition at one URL and overwrites it, so an edition that existed between two readings leaves no trace at that path — 'not in the release we read' and 'not published' are different claims and only the first is the Institute's to make. What the episode does establish is operational and is the Institute's own to fix: this publisher's audit stamp falls mid-morning Eastern, and a sweep that runs before it reads yesterday's edition. A daily series read on a cadence faster than the publisher's own stamp will skip editions, and the skip is a property of the reader. 🔴 THE SAME-MONTH DECEMBER SPREAD PEAKED AND TURNED, AND THE TURN IS A SINGLE SESSION. Brent December against WTI December, every mark a FINAL exchange settle: $12.12 on #191's October 5, $12.65 on #193's October 7, $13.71 on #195's October 9, $13.79 on #196's October 12, and $12.52 on #197's October 13 ($105.69 less $93.17). The series widened $1.67 over five sessions and gave back $1.27 of it in one. Both #196 and #197 are captured and FINAL in this reference's holdings, and NEITHER is the strip on the forward-curve page, which remains October 5 and discloses its own distance on that page — a captured bulletin and an adopted strip are different things and this row does not conflate them. Differencing the two printed opens gives $10.83 against $12.80 last step, and that figure is not a spread: Brent is quoting December and WTI November, and this row has refused that subtraction for weeks on exactly that ground. On this step the synthetic figure happens to agree in SIGN with the exchange's own spread — both narrowed — and disagrees by 78 cents on magnitude, $1.97 against $1.19. A figure that was wrong in sign two steps ago and right in sign on this one is not thereby a measurement; it is a subtraction of two different contract months that sometimes lands the right way round. The November spread remains uncomputable, and that is measured rather than assumed: #197's Brent leg runs DEC26 through DEC30 across 26 rows terminated by its own printed TOTAL, and carries no November row at all. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — everything from here to the next superseded marker was measured on the OCTOBER 9-TO-13 step and describes an interval this row no longer prints. It is retained because its half decomposition is the method the paragraphs above extend, and because its October 12 gap was closed the following day: the bulletin it reports as unavailable, #196 for trade date Monday October 12, was read on October 14 and is FINAL. Its own words were 'there is no October 12 settle in this reference because the exchange had not posted one at the path this reference reads, not because a posting was missed' — the narrow claim, and it was right. Had the broader claim been made, this reference would have been wrong in print within a day: 🔴 THE PRINTED STEP SPANS TWO TRADING SESSIONS AND IS HALF DECOMPOSABLE — THE FIRST SESSION IS BRACKETED BY PRINTED EXCHANGE FIGURES AND THE SECOND IS NOT OBSERVED AT ALL. The printed number went $103.73 to $106.17, up $2.44 or 2.352 per cent, and the WTI row below went $91.14 to $93.37, up $2.23 or 2.447 per cent. October 9 was a Friday and October 13 a Tuesday: four calendar days and TWO trading sessions inside them — October 12 and October 13. October 12 was Columbus Day, a United States federal bank holiday, and it is NOT a CME energy holiday, so a full October 12 session traded. What is available on the first half of the step is a printed settle rather than an inference. Bulletin #195 for Friday October 9 is now FINAL on all ten of its pages — see the standing-of-the-document paragraph below, because that is new as of this sweep — and it prints BZ DEC26 settling $104.72, which is 99 cents ABOVE this row's October 9 open of $103.73. So Brent rose 99 cents intraday on October 9 and a further $1.45 from that settle to the October 13 open, and the two legs sum to the printed $2.44 exactly. CL NOV26 settled $91.85 against the WTI row's $91.14 open, up 71 cents intraday, with $1.52 remaining to the October 13 open; that pair also sums exactly. 🔴 WHAT CANNOT BE DONE IS SPLITTING THE REMAINDER, BECAUSE NO OCTOBER 12 BULLETIN WAS AVAILABLE TO READ — AND THAT IS A DIFFERENT KIND OF ABSENCE FROM THE ONE THE SUPERSEDED ENTRY BELOW RECORDS. On the October 6-to-9 step, bulletins #192 and #194 existed and had rotated off the exchange's /current/ path before anybody read them: the figures were published and lost. Here, on October 13, the /current/ Section 61 path still served BULLETIN #195 FOR TRADE DATE FRIDAY OCTOBER 9. There is no October 12 settle in this reference because the exchange had not posted one at the path this reference reads, not because a posting was missed. So the $1.45 Brent and $1.52 WTI remainders each span the October 12 session plus the October 13 overnight and opening, and nothing here divides them. 'Not in the release we read' and 'not published' are different claims; this one is the first, and it is stated as the first. 🔴 AND THE STANDING OF BULLETIN #195 CHANGED BETWEEN SWEEPS, WHICH IS THE THIRD OBSERVED INSTANCE OF A CME BULLETIN FINALISING IN PLACE AT AN UNCHANGED URL. #195 was read on October 10 and marked PRELIMINARY on all ten pages, read again on October 11 and marked PRELIMINARY on all ten pages, and read on October 13 and marked FINAL on all ten pages — same bulletin number, same trade date, same path. The marking was MEASURED on each reading rather than inherited from the last one: ten page headers at identical line offsets, ten occurrences of FINAL, zero of PRELIMINARY. That is why the $104.72 and $91.85 settles above may now be cited as printed exchange figures, where the entry below could only cite them under a PRELIMINARY label. A document's standing is a measurement with its own date, and re-reading a document you have already read is not redundant. 🔴 THE SAME-MONTH DECEMBER SPREAD IS UNCHANGED FROM LAST SWEEP IN VALUE AND UPGRADED IN STANDING. Brent December against WTI December: $12.12 on #191's FINAL October 5 settles, $12.65 on #193's FINAL October 7 settles, and $13.71 on #195's now-FINAL October 9 settles ($104.72 less $91.01). No fourth mark is available, because no bulletin after #195 was readable this sweep. Differencing the two printed opens gives $12.80 and that figure is not a spread: Brent is quoting December and WTI November, and this row has refused that subtraction for weeks on exactly that ground. The November spread remains uncomputable, because Brent has no November leg in any bulletin this reference holds. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — everything from here to the next superseded marker was measured on the OCTOBER 6-TO-9 step and describes a three-session interval this row no longer prints. One of its claims has since been overtaken and is corrected in place rather than left to mislead: where it says #195's October 9 settles 'rest on a PRELIMINARY document', that was true when written on October 10 and was measured false on October 13, per the standing paragraph above. It is retained because the contrast between a step that cannot be decomposed at all and the half decomposition available above is the plainest statement this row holds of what an unread bulletin costs: 🔴 THE PRINTED STEP SPANS THREE TRADING SESSIONS, AND FOR THE FIRST TIME ON THIS ROW IT CANNOT BE DECOMPOSED, BECAUSE TWO OF THE FOUR INTERMEDIATE MARKS WERE NEVER READ. The printed number went $100.25 to $103.73, up $3.48 or 3.471 per cent, and the WTI row below went $89.28 to $91.14, up $1.86 or 2.083 per cent. October 6 was a Tuesday and October 9 a Friday: three calendar days and THREE trading sessions inside them — October 7, 8 and 9. The method the last several sweeps used here takes the publisher's two opens and every exchange settle between them and reconciles the whole step to the cent. It is unavailable on this step. The October 6 settle sits in bulletin #192 and the October 8 settle in #194, and NEITHER BULLETIN WAS READ IN ANY STATE before the exchange's /current/ path rotated past it. No arithmetic recovers them: a point-change column can back out the session preceding a bulletin that was read, and this reference holds no bulletin on either side of October 8. The step is printed as a three-session aggregate and nothing on this row decomposes it. 🔴 WHAT IS AVAILABLE INSTEAD IS NARROWER AND SHARPER THAN THE PRINTED PAIR, AND IT REVERSES THE OBVIOUS READING OF THE WTI LEG. Bulletin #193 for Wednesday October 7 was read and is marked FINAL on all ten of its pages, so its front-month settles are printed exchange figures rather than arithmetic. BZ DEC26 settled $100.20 there, FIVE CENTS below this row's October 6 open of $100.25 — so Brent was flat through the first of the three sessions, and the entire $3.48 rise occurred on October 8 and 9. CL NOV26 settled $88.28, a full DOLLAR below the WTI row's $89.28 open — so WTI FELL a dollar through October 7 and then rose $2.86 to its October 9 open. The printed WTI step of plus $1.86 is the net of a fall and a larger rise, and a reader taking it for a three-session climb would have the first session backwards. Note the standing of the document this rests on: #193 is cited here for two front-month settles and is NOT the strip on the forward-curve page, for a reason set out in full in the curve note — the Institute's extract of it is incomplete on the far tail of the WTI leg. The two rows quoted here sit in the intact part of that extract, and a document may be good enough to answer one question and not another. 🔴 THE SAME-MONTH DECEMBER SPREAD WIDENED ACROSS BOTH HALVES OF THE STEP, AND THE THREE MARKS DO NOT CARRY EQUAL WEIGHT. Brent December against WTI December: $12.12 on #191's FINAL October 5 settles, $12.65 on #193's FINAL October 7 settles, and $13.71 on #195's October 9 settles. The last of the three rests on a PRELIMINARY document and is labelled rather than omitted. Differencing the two printed opens gives $12.59 and that figure is not a spread: Brent is quoting December and WTI November, and this row has refused that subtraction for weeks on exactly that ground. The November spread remains uncomputable, because Brent has no November leg in any bulletin this reference holds. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — everything from here to the end of this note was measured on the OCTOBER 2-TO-6 step and describes a two-session interval this row no longer prints. It is retained because its decomposition is the method the entries above had to decline for want of two bulletins, and the contrast between a step that reconciles to the cent and a step that cannot be decomposed at all is the plainest statement available of what an unread bulletin costs: 🔴 THE PRINTED STEP SPANS TWO TRADING SESSIONS, NOT ONE, AND EVERY PERCENTAGE BELOW IS STATED AGAINST THAT INTERVAL RATHER THAN AGAINST A DAY. The printed number went $102.23 to $100.25, down $1.98 or 1.937 per cent, and the WTI row below went $92.83 to $89.28, down $3.55 or 3.824 per cent. October 2 was a Friday and October 6 a Tuesday: four calendar days, and TWO trading sessions inside them — Monday October 5 and Tuesday October 6. The previous entry on this row predicted the weekend and said so in advance ('October 2 to the following trade date spans a weekend, so one calendar day of movement on the publication clock will carry three on the calendar'). It got the weekend and a skipped session on top of it, because no sweep ran on October 5. The prediction was right about the mechanism and short by one session, and the honest reading is that a sweep cadence the Institute controls is not a property of the market: a gap in the publication clock is the Institute's own doing and must not be presented as an interval the price moved across. 🔴 THE STEP DECOMPOSES EXACTLY, AND THE DECOMPOSITION OVERTURNS WHAT THE PRINTED PAIR APPEARS TO SAY. Taking the publisher's two opens and the exchange's two intervening settles, both legs reconcile to the cent. Brent: $102.23 open, $102.25 at the October 2 settle (up 2 cents intraday), $100.32 at the October 5 settle (down $1.93), $100.25 at the October 6 open (down 7 cents) — total −$1.98. WTI: $92.83 open, $91.11 October 2 settle (down $1.72 intraday), $89.43 October 5 settle (down $1.68), $89.28 October 6 open (down 15 cents) — total −$3.55. Read off the printed pair alone, WTI fell $1.57 further than Brent and the obvious inference is that WTI was the weaker leg. It is wrong. $1.74 of that $1.57 is the October 2 INTRADAY leg — a window that belongs to neither printed quote, since it sits after the open this row carried last sweep and before the settle the exchange struck that evening. On October 5, the only full session both printed quotes bracket, BRENT FELL MORE: $1.93 against WTI's $1.68 on the quoted months, and $1.93 against $1.23 December-against-December. 🔴 AND THE EXCHANGE'S OWN SPREAD AGREES WITH THE SESSION RATHER THAN WITH THE PRINTED PAIR, WHICH IS WHY THE DECOMPOSITION IS WORTH THE SPACE. The same-month December spread went $12.82 to $12.12 — it NARROWED 70 cents — while differencing the two printed opens gives $10.97 against $9.40 last step and reads as a $1.57 WIDENING. The synthetic figure is now wrong in SIGN and not merely in magnitude, which is a sharper statement of the refusal this row has carried for weeks and is the strongest case for it yet recorded. The November spread has stopped being computable altogether: BZ NOV26 has NO ROW in bulletin #191, so there is no November Brent mark of any kind to difference. 🔴 THE DECOMPOSITION'S MIDDLE POINT IS NOW FINAL-CORROBORATED, WHICH IT WAS NOT WHEN THIS METHOD WAS FIRST USED. Bulletin #191, trade date Monday October 5, is marked FINAL, and backing its point-change column off its settles reproduces #190's printed October 2 settles exactly: $100.32 plus $1.93 is $102.25 on BZ DEC26, and $89.43 plus $1.68 is $91.11 on CL NOV26. Both to the cent, 87 of 87 shared contract months reconciling the same way. The October 2 settles entered this ledger off a PRELIMINARY document; they are now confirmed by a FINAL one that was compiled independently of it. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — what follows from here to the end of this note was measured on the OCTOBER 1-TO-2 step and describes a one-session interval with both clocks in agreement; none of it describes the step this row now prints. It is retained because the roll it establishes on each leg is the precedent the clean results above rest on: NEITHER BENCHMARK ROLLED ON THIS STEP, SO FOR THE FIRST TIME IN SEVERAL SWEEPS BOTH PRINTED MOVES ARE PRICE MOVES. The printed number went $97.89 to $102.23, up $4.34 or 4.434 per cent, and the WTI row below went $90.31 to $92.83, up $2.52 or 2.790 per cent. Both steps survive the roll test, and the test is run before the moves are reported rather than after. Bulletin #190, trade date Friday October 2, carries a point-change column, and backing it out reconstructs the October 1 settles this reference never obtained directly: CL NOV26 $92.87 (91.11 + 1.76), CL DEC26 $90.84 (89.43 + 1.41), BZ DEC26 $102.31 (102.25 + 0.06). Against those, this $102.23 Brent open sits EIGHT CENTS beneath the DECEMBER settle — an ordinary overnight gap — and $1.28 beneath the November mark. Brent is still quoting December, so the roll named on the last sweep holds and is not repeated. The WTI open sits FOUR CENTS beneath CL NOV26 and $1.99 ABOVE CL DEC26, which on a session the contract fell $1.76 would be a gap in the wrong direction: WTI is still quoting November. Two legs, two clean results, no roll on either. 🔴 AND THE NOVEMBER BRENT CONTRACT IS NOW EMPTY, WHICH CORRECTS THIS LEDGER'S OWN MECHANISM RATHER THAN MERELY CONFIRMING ITS CALL. BZ NOV26's open interest went 19,421 to ZERO, and #190 attributes the whole −19,421 to its own October 2 session — which means November still held every one of those lots through October 1, the session the roll actually happened on. So the publisher's roll PRECEDED the open-interest drain by one session; it did not follow it. The three sweeps that carried the advance warning rested on open interest as the LEADING indicator, and on the evidence it was the LAGGING one. That is recorded as a correction to the reasoning, not as a vindication of the conclusion: the call was right and the stated mechanism behind it was wrong, and those are two different findings. BZ NOV26 now prints a nominal $103.51 marked UNCH on no volume and no open interest — a price with no position behind it, the same stage CL OCT26 reached in September. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — the October 1 roll narrative this row carried, which is now one step behind: THE BRENT ROLL THIS LEDGER NAMED IN ADVANCE ON THREE CONSECUTIVE SWEEPS HAS FIRED, AND MOST OF THE PRINTED FALL IS THE CONTRACT CHANGING. The printed number went $102.58 to $97.89, down $4.69, and read as a price move that is wrong in magnitude and wrong in SIGN: the barrel rose on this session. The test is the one this ledger used on the WTI roll four weeks ago — compare the new open against the prior session's settle on each candidate month, because the month that produces an ordinary overnight gap is the month being quoted. Bulletin #188, trade date Wednesday September 30, settles BZ DEC26 at $98.03 and BZ NOV26 at $103.53. This $97.89 open sits FOURTEEN CENTS beneath the DECEMBER settle — an ordinary overnight gap — and $5.64 beneath the NOVEMBER settle, which on a session the market spent rising would be implausible. The quote has moved from November to December between September 30 and October 1, and the publisher said nothing, exactly as it said nothing in September. What that leaves of the decline: the November-to-December spread was $5.50 on the September 30 settles, so of the $4.69 printed fall roughly $5.50 is the contract changing underneath the series and roughly $0.81 is a price move — UPWARD. The WTI row below did NOT roll and rose $1.39 on the same step, so the two residuals agree in sign and to 58 cents. That is the check, and it is looser than the September check on the WTI roll, which agreed to about twenty cents; the slack is stated rather than smoothed over. Independent corroboration that October 1 was an up session, from outside this publisher: CNBC reported Brent up more than 4 per cent intraday on a third U.S. aircraft carrier being sent to the Middle East, and the Forbes page's own intraday sentence has Brent rising 2.00 per cent in the past 24 hours. A reader who takes the printed $4.69 as a crude-oil move has the direction of the day backwards. 🔴 WHAT THE PRE-REGISTRATION GOT RIGHT AND WHAT IT GOT WRONG, BECAUSE BOTH BELONG ON THE RECORD. The magnitude landed exactly: the warning carried on the last sweep said the roll would print a fall of roughly five and a half dollars that is not a price move, and the step was $5.50 to the cent. But the open-interest inference carried DIRECTION, not TIMING — the warning stood unfired for four consecutive steps, and this ledger said each time that a prediction quietly carried forward until it happens to land is not a prediction. And the magnitude was right only because it was re-measured every sweep, $7.45 to $6.43 to $5.50: the figure named when the warning was first written would have been wrong by $1.95. An inference from open interest tells a reader that a roll is coming and does not tell them when or how far. 🔴 THE OPEN-BASIS BRENT–WTI SPREAD IS STILL NOT COMPUTABLE FROM THIS PAGE, AND THE LEDGER STILL WILL NOT SYNTHESISE ONE — the two quotes remain on different contract months, Brent December and WTI November, so differencing them gives $9.40 and that number is two legs on two months rather than a spread. Note what it would have told a reader who took it: $7.58 last step to $9.40 now reads as a $1.82 widening, while the exchange's own same-month December spread widened $3.21 over the same ground. What can honestly be quoted is that December figure, on the October 2 settles: BZ DEC26 $102.25 against CL DEC26 $89.43 is $12.82, against $9.61 on the September 30 settles. It widened $3.21 in two sessions, and the decomposition is one-sided — BZ DEC26 rose $4.22 across those two sessions while CL DEC26 rose $1.01. Brent did the work; WTI barely moved. 🔴 AND THE NOVEMBER SPREAD HAS CEASED TO BE QUOTABLE AT ALL, WHICH IS A LOSS OF INSTRUMENT AND NOT A CHANGE IN PRICE. Last step this row printed $13.11 on November ($103.53 against $90.42) and $9.61 on December, both real, and said the $3.50 difference between them measured how much the month mattered. Only the December leg survives: with BZ NOV26 at zero open interest its $103.51 is nominal, so a November spread computed against CL NOV26's live $91.11 would be a real number differenced against a stale mark. The $13.11 is retained here as what was measured on September 30 and is not carried forward. 🔴 BOTH CLOCKS AGREE ON THIS STEP. October 1 was a Thursday and October 2 a Friday: consecutive calendar days and consecutive trading sessions, with no weekend inside the gap. The agreement is restated rather than left to be assumed, for the same reason a disagreement is stated when it occurs. The next step will not have this property — October 2 to the following trade date spans a weekend, so one calendar day of movement on the publication clock will carry three on the calendar. Against the exchange's own settles, and note carefully which are FINAL: Bulletin #184, trade date Thursday September 24, marked FINAL, settles BZ NOV26 at $106.60. 🔴 THAT SENTENCE USED TO CONTINUE '— that is still the newest settlement this reference carries and still the strip published on the forward-curve page, now [N] trading sessions behind on the convention stated beside PUB_ISO', WITH THE SESSION COUNT SPLICED IN LIVE FROM CURVE_STRIP_SESSIONS_TEXT. BOTH CLAUSES ARE NOW FALSE AND THE SPLICE WAS A DEFECT IN ITS OWN RIGHT. Bulletin #191, trade date Monday October 5, was read marked FINAL, so #184 is no longer the newest FINAL settlement this reference carries and is no longer the strip on the forward-curve page — CURVE_STRIP_TRADE_ISO is now 2026-10-05. And a LIVE derived value inside a paragraph labelled superseded would have recomputed itself on every build, so a 2026-09-24 claim would have gone on reporting its distance from whatever the current publication date happened to be: a stale sentence wearing a fresh number, which is the exact shape the Two-Clock rule exists to stop. The figure is frozen as a bracketed placeholder and the live reference is removed. #184's $106.60 is retained only as the September 24 measurement it always was. 🔴 TWO DIFFERENT KINDS OF ABSENCE SIT BEHIND THAT, AND COLLAPSING THEM WOULD MISSTATE WHAT THE PUBLISHER DID. FIVE later bulletins have been READ AND NOT ADOPTED, every one marked PRELIMINARY on first sight: #185 for Friday September 25, #186 for Monday September 28, #187 for Tuesday September 29, #188 for Wednesday September 30, and now #190 for Friday October 2, which was read across all ten of its pages and carried PRELIMINARY on ten of ten and FINAL on none. #189, for trade date Thursday October 1, was NEVER READ AT ALL — the previous sweep's fetch path refused the URL and the public path ends /current/, which rotates daily and is overwritten without notice. So nothing is claimed about whether #189 existed or what it marked. An unread document is not a PRELIMINARY one, and the PRELIMINARY-on-first-sight count is five observations rather than six. 🔴 THE TWO PROVISIONAL PATHS MUST NOT BE COMPARED WITH EACH OTHER, WHICH IS THE WHOLE POINT OF A ROLL, AND ONE OF THEM HAS NOW ENDED. The NOVEMBER path, provisional after the first figure, ran $106.60 FINAL on September 24, $104.32 derived for September 25, $105.28 for September 28, $102.59 for September 29, $103.53 for September 30, and stops there as a traded series: the $103.51 #190 prints for October 2 is marked UNCH on zero volume and zero open interest and is a mark, not a settlement anybody transacted into. The DECEMBER path — the month this row is quoting — runs $96.16 derived for September 29, $98.03 for September 30, $102.31 DERIVED for October 1 from #190's own point-change column, and $102.25 printed for October 2. This $102.23 open sits eight cents below the October 1 derived settle. Two points about that derivation, stated rather than buried: it is arithmetic on a PRELIMINARY document and inherits that document's provisional status, and it is the only route to an October 1 figure this reference has, because #189 was never read. 🔴 THE PUBLISHER'S INTERNAL INCONSISTENCY REPRODUCES AN ELEVENTH TIME, AND THE DISCRIMINATING TEST RESOLVES TO THE OPEN ON BOTH LEGS FOR A SECOND CONSECUTIVE STEP. The same page records Brent as having fallen 2.49 per cent in the past 24 hours to $99.69/bbl. Back that percentage out and the implied base is $102.2357, which sits 0.57 of a cent ABOVE the figure the same page prints as the October 2 open and 7.43 cents from the October 1 derived settle of $102.31 — the residual picks the open by about thirteen to one. The WTI row below resolves the same way at 0.79 of a cent, so this is again the pair of benchmarks moving together rather than one line's rounding, which remains the single structural observation that has held across all eleven readings. 🔴 AND THE CAPTION DEFECT REPRODUCES WITH IT, WHICH IS A SEPARATE FAULT FROM THE RESIDUAL. The percentage is captioned as a 24-hour change. Measured against the actual prior settle of $102.31, $99.69 is a fall of 2.561 per cent, not the 2.49 per cent printed. So the figure is an open-to-intraday change wearing a close-to-intraday caption, exactly as the WTI line is below, and a reader who takes the caption at face value is reading the wrong interval. 🔴 RETAINED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule, and it is a warning against reading a run of residuals as a rule — including the run that now appears to be re-establishing one. The finding carried by six consecutive sweeps read that the implied base tracks the open and not the close; the seventh reading missed the open by 90.7 cents; the eighth did not restore it; the ninth flipped sign on both benchmarks at 6.5 and 1.5 cents below their opens; the tenth flipped back at 1.2 and 0.5 cents above; this eleventh lands above again, at 0.57 and 0.79 of a cent. A residual that has run 3, 6, 8, 90.7, −6.5 and then +1.2 and +0.57 cents is not a rule with exceptions. TWO CONSECUTIVE READINGS ON THE SAME SIDE ARE TWO READINGS, and the reason this is said out loud is that the six-sweep run which preceded the 90.7-cent miss was longer than this one and still broke. Only the weaker claim is stated: the percentage the page prints is not reliably computed against the open the page prints. 🔴 ALSO SUPERSEDED AND LABELLED, AND NOW TWO GENERATIONS BACK: the entry on this row describing the SEPTEMBER 29-TO-30 step said that the $3.46 fall in Brent and the $4.34 fall in WTI were price moves on unchanged contracts, directly comparable, and that the open-basis spread had widened from $12.78 to $13.66. Every one of those statements was correctly measured for that step and none of them survived to the next. It is flagged here because the paragraph at the head of this note reaches the same verdict — no roll on either leg — about a DIFFERENT interval, the October 1-to-2 step, and the two must not be read as one continuing finding. A verdict that recurs is not a verdict that persisted. Note also that the same page continues to stamp its Brent intraday line with a calendar date and no clock time ('as of October 2, 2026'), while stamping the WTI line with a clock time and no date ('as of 8:28 a.m. ET'). An open is a measurement of one instant and it is not a day. On this edition the asymmetry compounds: the page's own audit stamp reads 10:07am, which is an hour and thirty-nine minutes LATER than the WTI quote it certifies, so the stamp cannot be a verification timestamp for that line. |
| WTI (futures, front month) | $99.36 / bbl | opening quote, October 23, 2026 | Forbes Advisor daily crude quote, page dated October 23, 2026 (page stamped 'Audited & Verified: Oct 23, 2026, 1:35pm') | WTI is priced at Cushing, Oklahoma, inland. Its spread to Brent is a shipping and access number, not a quality number. 🔴 THIS ROW FELL WHILE BRENT ROSE, AND THE SPREAD THAT SEPARATES THEM IS THE MEASUREMENT THAT MOVED. This row went $101.44 to $99.36, DOWN $2.08 or 2.050 per cent, while the Brent row above went $109.13 to $113.12, UP $3.99 or 3.656 per cent — opposite signs over the identical interval off the identical edition of one publisher’s page. The Brent-less-WTI open spread went $7.69 to $13.76. Because this row is the inland leg, a widening spread on a falling WTI is the shipping-and-access number doing the work the lede above describes, and that is a statement about which measurement moved rather than an explanation of why. See the Brent row above, where the interval, the absence of a FINAL settlement to decompose against, the publisher’s finalise-in-place calendar, the intraday reversal and this sweep’s write ordering are all set out once rather than twice. 🔴 AND THE FALL IS LOCATED OUTSIDE THE SESSION THIS ROW’S DATE NAMES, BY ARITHMETIC ON A ROUNDED PERCENTAGE — IT IS A LOCATED LEG, NOT A MEASURED ONE. The same page prints an intraday WTI of $101.46 at 8:22 a.m. ET, ‘up 2.06 per cent since yesterday’s close’, which puts the October 22 close at about $99.41 — within five cents of the $99.36 open this row adopts. So essentially none of the $2.08 belongs to October 23; it belongs to October 20, 21 and 22, which are the three sessions no FINAL bulletin brackets. The reconstruction is limited by the two decimals the publisher printed on that percentage and is not a price anybody settled at. Both the $2.08 endpoints are firm; only its placement inside the interval is inferred, and the two claims must not be read as one. 🔴 SUPERSEDED AND LABELLED — WHAT FOLLOWS DESCRIBES THE OCTOBER 15 TO OCTOBER 19 STEP ON THIS ROW AND IS RETAINED, NOT DELETED. It is kept because it is the other term of a difference a reader must be able to check, and because its fully-closed three-leg decomposition is the standard the step above could not meet. Its figures are correct as of their own measurement dates and are not the figures this row now prints. 🔴 THE STEP IS FULLY DECOMPOSABLE, IT CLOSES TO THE CENT, AND THE WEEKEND MOVED IT BY THREE CENTS. Opened $6.28 ABOVE the October 15 open of $95.16, a rise of 6.599 per cent. October 15 was a Thursday and October 19 a Monday — four calendar days with two market sessions inside them, Friday October 16 and Monday October 19. Three legs: (a) bulletin #199, trade date October 15 and marked FINAL on all ten pages, settles CL NOV26 at $97.97, which is $2.81 above this row’s own October 15 open of $95.16; (b) bulletin #200, trade date October 16 and marked PRELIMINARY on all ten pages, settles CL NOV26 at $101.41, a point change of + 3.44; (c) from that settle to the October 19 open, $101.41 to $101.44, UP 3 cents. Plus 2.81 plus 3.44 plus 0.03 is plus 6.28, the printed figure exactly. See the Brent row above, where the interval, the provisional-middle-term caveat, the identity test that licenses #200’s settle, the December spread, the roll, the intraday distinction and this sweep’s ordering deviation are all set out once rather than twice. 🔴 AND TWO MEASUREMENTS HERE LOOK LIKE ONE AND ARE NOT: OPEN-TO-SETTLE IS NOT THE EXCHANGE’S POINT CHANGE. For October 15 this row’s open-to-settle leg is + $2.81, from its own $95.16 open to #199’s $97.97 settlement, while #199’s printed change column for the same session reads + 2.88, because that column measures settlement against the PRIOR session’s settlement and not against the open. Both numbers are correct and they answer different questions. The decomposition above uses open-to-settle for leg (a), because that leg’s left endpoint has to be the figure this row is retiring; the change column is what makes the identity test in the Brent row possible. A seven-cent discrepancy between two figures that are both right is exactly the kind of thing a later reader repairs by choosing one, so it is recorded here instead. 🔴 SUPERSEDED AND LABELLED — WHAT FOLLOWS DESCRIBES THE OCTOBER 13 TO OCTOBER 15 STEP ON THIS ROW AND IS RETAINED, NOT DELETED. Its figures are correct as of their own measurement dates and are not the figures this row now prints. Opened $1.79 ABOVE the October 13 open of $93.37, a rise of 1.917 per cent. October 13 and October 15 are a Tuesday and a Thursday, with two trading sessions inside the interval — October 14 and October 15 — and the October 14 edition of the quoting page was not read. See the Brent row above, where the interval is set out and where the reason the second half cannot be split is recorded. 🔴 THE STEP IS HALF DECOMPOSABLE AND BOTH HALVES RUN THE SAME WAY, WHICH IS WHY THIS LEG IS SET OUT BESIDE THE BRENT LEG RATHER THAN SUMMARISED WITH IT. Bulletin #197, trade date Tuesday October 13 and marked FINAL on all ten pages on first reading, settles CL NOV26 at $94.14. Against this row's own October 13 open of $93.37 that is 77 cents of intraday October 13 movement, leaving $1.02 — from the settle to the October 15 open — in a window this reference did not observe. The two legs sum to the printed $1.79 exactly. 🔴 AND THE TWO BENCHMARKS MOVED IN OPPOSITE DIRECTIONS INSIDE THE OBSERVED LEG, ON THE PUBLISHER'S OWN CHANGE COLUMN RATHER THAN ON ARITHMETIC. Brent fell 48 cents intraday on October 13 while WTI rose 77; #197 prints the exchange's own point-change column as CL NOV26 + 0.40 and BZ DEC26 − 0.96, which is a disagreement in SIGN confirmed by the source document and not inferred from two quoted opens. That matters for how the Brent row's sign reversal is read: it is a property of this session, visible in the bulletin, and not an artifact of pairing a Forbes open against a CME settle. 🔴 THE ROLL TEST IS RUN PROPERLY — ON THE STEP BEING SUPERSEDED, NOT ON THE CURRENT ONE — AND THE PROMISE THE ENTRY BELOW MADE IS KEPT IN THE ONLY FORM THE HOLDINGS ALLOW. The demoted entry closed with 'the next sweep that can read a bulletin runs the test properly.' A bulletin is now in hand, but it is #197 for October 13, which is the session of the QUOTE BEING RETIRED and two sessions before the quote being adopted. So the test is run retrospectively on the October 13 open, against #197's own same-session Globex opens — the shape used on October 9 and licensed there. CL NOV26 opened $93.21 and CL DEC26 $92.40, and the $93.37 October 13 quote sits SIXTEEN CENTS above November and 97 cents above December: November by a factor of about six. On the Brent leg BZ DEC26 opened $106.09 and BZ JAN27 $102.61, and that row's $106.17 quote sits EIGHT CENTS above December and $3.56 above January: December by a factor of forty-four. Both legs clean, both on the correct month, and the window is printed beside each ratio per the operating rule adopted below. 🔴 THE SETTLE-BASED FORM WOULD HAVE RETURNED THE WRONG MONTH AGAIN, WHICH IS THE SECOND OBSERVED INSTANCE AND RETIRES THE WORD 'FIRST' IN THE ENTRY BELOW. Tested against #197's October 13 settles instead of its opens, the $93.37 quote sits 77 cents BELOW CL NOV26's $94.14 and TWENTY CENTS above CL DEC26's $93.17, so the nearest candidate is DECEMBER and the answer is wrong. The mechanism is the one already recorded: October 13 rose on the front WTI months, so an open sits below its own settle by construction and the comparison measures an intraday move rather than an overnight gap. The Brent leg survives the same error once more, and once more by luck — 48 cents above December's settle against $3.26 above January's. A failure mode that recurs on the same leg by the same mechanism is a property of the test's shape, not an accident of one session. 🔴 THE CURRENT STEP'S ROLL TEST IS UNAVAILABLE AND IS RECORDED AS UNAVAILABLE. The immediately preceding settle for an October 15 open is October 14, for which no bulletin exists in this reference's holdings; the same-session opens would require an October 15 bulletin, which is not posted at the path this reference reads. The nearest exchange figures are #197's, two sessions back. Carrying them forward would place the $95.16 quote $1.02 above November's settle and $1.99 above December's, which points at November — but that is a two-session window and the entries below establish that a margin is only comparable to a margin from the same window. It is an inference and it is not a roll test. 🔴 A ROLL IS NOW PRE-REGISTERED ON THE EXCHANGE'S OWN OPEN-INTEREST COLUMN, SO THE NEXT SWEEPS SHOULD EXPECT THE QUOTED MONTH TO CHANGE. #197 prints CL NOV26 open interest at 140,264 contracts and FALLING by 11,536 on the session, against CL DEC26 at 286,106 and RISING by 11,680. December now carries more than twice November's open interest and the two columns are moving in opposite directions, which is what the days before a front-month expiry look like. This is written down before the fact rather than explained afterwards: when a future sweep's roll test returns December on the WTI leg, that is the expected result and not a defect in the quote. The prior WTI roll, on September 23, was reported only after it had happened. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — what follows was measured on the OCTOBER 9-TO-13 step. Two corrections to it, in place rather than by rewriting it. Its roll test is recorded as unavailable, and it is now available and has been run above, cleanly on both legs; the unavailability was a true statement about that sweep's holdings and is no longer a true statement about this reference's. And its closing section describes the settle-based failure mode as the method's first, which was accurate when written and is superseded by the second instance recorded above: Opened $2.23 ABOVE the October 9 open of $91.14, a rise of 2.447 per cent across TWO trading sessions — October 12 and October 13. See the Brent row above, where the interval is set out, where the step is shown to be HALF decomposable, and where the reason the second half cannot be split is recorded: no October 12 bulletin was available to read, so the /current/ Section 61 path still served #195 for trade date Friday October 9. What can be bracketed is the first session. Bulletin #195 — now marked FINAL on all ten pages, which is what licenses citing it as a printed exchange figure rather than as arithmetic — settles CL NOV26 at $91.85 against this row's own October 9 open of $91.14, so 71 cents of the printed $2.23 is intraday October 9 movement and $1.52 belongs to a window this reference did not observe. The two legs sum to the printed step exactly. 🔴 THE ROLL TEST IS UNAVAILABLE ON THIS STEP, AND IT IS RECORDED AS UNAVAILABLE RATHER THAN SUBSTITUTED. Both forms of the test need a figure from the quoted session or the one before it. The immediately preceding settle is October 12, for which no bulletin exists at the path this reference reads; the same-session Globex opens for October 13 would require an October 13 bulletin, which likewise is not posted. The only exchange figures in this reference's holdings are #195's, for October 9 — FOUR CALENDAR DAYS AND TWO SESSIONS BEFORE THE QUOTE. The entry below sets out, in its own words, why they will not do: 'a margin is only comparable to a margin from the same window', and the failure mode recorded there is precisely that of running the test on figures drawn from the wrong window. So the test is skipped. A roll call that cannot be run is reported as not run. 🔴 A WEAK PLAUSIBILITY CHECK EXISTS AND IS LABELLED AS AN INFERENCE, NOT AS A ROLL TEST. On October 9 the front spread printed CL NOV26 at $91.85 against CL DEC26 at $91.01, an 84-cent November-over-December premium. Carry that curve shape forward unchanged through the roughly $1.5 rise implied above and November lands near $93.4 with December near $92.5, which would place this $93.37 quote on the NOVEMBER contract — the same month the last six clean runs of the real test returned, and the month a reader would expect in mid-October. That is consistent and it is not evidence. It assumes a four-day-old curve shape held, which is the one thing a roll test exists to avoid assuming, and it is stated here only so that the unavailability above is not mistaken for a suspicion about the quote. The next sweep that can read a bulletin runs the test properly. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — what follows was measured on the OCTOBER 6-TO-9 step, and the roll test in it was run against bulletin #195's SAME-SESSION GLOBEX OPENS for October 9, a shape this page had never used before and could not use on the current step because no bulletin for the quoted session exists. One correction to it, in place: where it describes #195 as the source of those opens it does not say how the document was marked, and it was marked PRELIMINARY on both readings then available (October 10, attended; October 11, unattended). Re-read on October 13 it carries FINAL on all ten pages and no occurrence of PRELIMINARY — the THIRD observed instance of a CME bulletin finalising in place at an unchanged URL. The opens cited below are unchanged in value and upgraded in standing: 🔴 THE ROLL TEST IS CLEAN ON BOTH LEGS AND WAS RUN IN A SHAPE THIS PAGE HAS NEVER USED, BECAUSE THE SETTLE IT NORMALLY TESTS AGAINST IS NOT IN THIS REFERENCE'S HOLDINGS. Every prior roll call here compares the publisher's open against the IMMEDIATELY PRECEDING session's settle, on the reasoning that the month producing an ordinary overnight gap is the month being quoted. The immediately preceding settle is October 8, which is bulletin #194, which was never read. So the test is run instead against the exchange's OWN GLOBEX OPENS FOR THE SAME SESSION, printed in bulletin #195 for October 9. CL NOV26 opened $91.29 and CL DEC26 $90.36, and this $91.14 quote sits FIFTEEN CENTS below November and 78 cents ABOVE December — still November. BZ DEC26 opened $103.90 and BZ JAN27 $100.75, and the $103.73 Brent quote sits SEVENTEEN CENTS below December and $2.98 above January — still December. The two residuals agree to TWO CENTS of each other, which is tighter than any roll call this page has recorded. 🔴 AND THE TIGHTNESS IS A PROPERTY OF THE TEST'S NEW SHAPE RATHER THAN OF THE MARKET, WHICH IS WHY THE CHANGE IS STATED RATHER THAN LEFT TO BE NOTICED. An open against the same session's open compares two measurements of one instant and ought to agree closely; an open against the previous evening's settle compares across a night and ought not. A two-cent agreement on the new test is therefore not evidence that these quotes have become more reliable, and it must not be read as continuing the series of margins the entries below report. The two tests answer the same question and do not have the same tolerance, and a margin is only comparable to a margin from the same window. 🔴 ONE WAY OF RUNNING THIS TEST WOULD HAVE RETURNED THE WRONG CONTRACT MONTH, AND IT IS RECORDED BECAUSE IT IS THE FIRST VISIBLE FAILURE MODE THE METHOD HAS HAD. Testing the October 9 open against the October 9 SETTLES — the nearest figures to hand, in the same bulletin — puts this $91.14 quote 71 cents below CL NOV26's $91.85 and THIRTEEN CENTS above CL DEC26's $91.01, so the nearest candidate is DECEMBER and the answer is wrong. The reason is that October 9 rose on every printed WTI month, so an open sits far below its own settle by construction and the comparison measures an intraday move rather than an overnight gap. The Brent leg survives the same error by luck, at 99 cents below December against $2.48 above January. A test whose logic depends on the window it spans has to be run on that window. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — what follows was measured on the OCTOBER 2-TO-6 step against bulletin #191's FINAL October 5 settles, which is the ordinary form of this test and the one the entries above had to substitute for: Opened $3.55 BELOW the October 2 open of $92.83, a fall of 3.824 per cent. October 2 and October 6 are a Friday and a Tuesday — four calendar days and TWO trading sessions, with a skipped sweep inside them; see the Brent row above, where the interval and its exact decomposition are set out and where the apparent gap between the two legs is shown to belong mostly to an October 2 intraday window that neither printed quote contains. 🔴 THE ROLL TEST IS CLEAN ON BOTH LEGS, AND FOR THE FIRST TIME SINCE #184 IT IS RUN AGAINST A FINAL DOCUMENT WITH NO DERIVATION IN IT. Bulletin #191, trade date Monday October 5, was read marked FINAL, so the candidate settles are printed figures rather than arithmetic backed out of a point-change column on a PRELIMINARY sheet — which is what every roll test since September 24 has had to rely on. CL NOV26 settles $89.43 and CL DEC26 $88.20. This $89.28 open sits FIFTEEN CENTS beneath November and $1.08 ABOVE December, so WTI is still quoting November. On the Brent leg, BZ DEC26 settles $100.32 and BZ JAN27 $97.58; the $100.25 open sits SEVEN CENTS beneath December and $2.67 above January, so Brent is still quoting December. Note what has changed in the shape of the Brent test and not only in its result: the alternative month is now JANUARY, because BZ NOV26 has no row in #191 at all, so the November candidate this test used to carry has ceased to exist rather than merely gone stale. 🔴 SEVEN RUNS OF THE TEST NOW — TWO ROLLS AND FIVE CLEAN RESULTS — AND THE MARGINS ARE REPORTED BECAUSE THE VERDICT ALONE WOULD HIDE HOW CLOSE SOME OF THEM ARE. WTI rolled on September 23 (70 cents against November versus $4.77 against October). Brent rolled on October 1 (14 cents against December versus $5.64 against November). WTI cleared on October 1 (11 cents versus $1.89 the wrong way), on October 2 (4 cents versus $1.99 the wrong way), and now on October 6 at 15 cents against $1.08 — which is the SMALLEST alternative-month margin this test has ever recorded, a little over a dollar where the rolls were excluded by four and five. Brent cleared on October 2 (8 cents against a November mark that no longer carried open interest) and now on October 6 at 7 cents against $2.67, the largest margin since the rolls themselves. A detector that only ever fires has demonstrated nothing, which is why the five clean results are reported with the same prominence as the two rolls — and the narrowing WTI margin is stated as a weakening of the instrument rather than left for a reader to notice. 🔴 THE TWELFTH RESIDUAL READING IS THE FIRST ON WHICH THE TWO LEGS DISAGREE, AND IT REFUTES THE SINGLE MOST DURABLE STRUCTURAL CLAIM THIS ROW HAS CARRIED. The page reports WTI trading at $87.14/bbl, 'down 2.40% since yesterday's close'. Back the percentage out and the implied base is $89.2828, which sits 0.28 of a cent above the printed October 6 open and 14.72 cents below the October 5 FINAL settle — the residual picks the open by about fifty-three to one, the strongest resolution of the twelve. The Brent line does NOT follow it. The page reports Brent down 2.75 per cent to $97.53, implying a base of $100.2879: 3.79 cents ABOVE its open and 3.21 cents BELOW its settle, a ratio of 1.18 to one, which is not a pick at all. Eleven previous readings found the two benchmarks moving together, and that co-movement was described on the last sweep as 'the single structural observation that has held across all eleven readings'. It has not held across twelve. This is reported as a NON-REPLICATION rather than dropped, because a claim that survives eleven readings and fails the twelfth is exactly the claim a dated ledger exists to retire in public. 🔴 AND THE PREVIOUS ENTRY PRE-REGISTERED THIS FAILURE, WHICH IS THE ONLY REASON IT CAN BE EXPLAINED RATHER THAN MERELY REPORTED. That entry warned, in as many words, that 'a four-to-one ratio measured across a four-cent window is a weak instrument however clean the arithmetic looks.' WINDOW WIDTH IS THE WHOLE OF THE EXPLANATION. The gap between the two candidates is FIFTEEN CENTS on WTI and SEVEN on Brent; the wide window resolves at fifty-three to one and the narrow one fails to resolve at all. So the legs did not diverge because the publisher changed what it does — they diverged because one leg had enough separation to measure and the other did not. The operating consequence is adopted here and should be kept: PRINT THE WINDOW BESIDE THE RATIO EVERY TIME. A ratio quoted without its window is a confidence figure with its denominator suppressed, and on this step it would have made a 1.18 and a 52.8 look like members of the same series. 🔴 THE CAPTION TEST SPLITS THE SAME WAY, AND ONE LEG IS CONVICTED WHILE THE OTHER CANNOT BE. WTI's printed 2.40 per cent reproduces the open-based change of 2.397 per cent to the digit and misses the close-based 2.561 per cent by sixteen basis points, under a caption that reads 'since yesterday's close' — so on this leg the figure is an open-to-intraday change wearing a close-to-intraday caption, as it has been on every prior reading. Brent's printed 2.75 per cent sits BETWEEN the two candidates: 2.781 per cent against the settle (3.1 basis points away) and 2.713 per cent against the open (3.7 away). It cannot be convicted on this step, and saying so costs nothing. One page, one session, two legs, two verdicts — which is itself the finding, because the caption defect has until now been reported as a property of the page rather than of a line. 🔴 THE HEADLINE QUOTES A THIRD NUMBER AGAIN, AND IT IS NOT THE ONE THIS ROW TABULATES. The page is headed 'Crude Oil Prices Slip: WTI Slides To $87, Brent Weakens Too'. The $87 is the intraday print, not the open, and this row carries the open. A reader who arrives from the headline and reads the table will find two different numbers for the same benchmark on the same morning, both correct and measuring different instants. The clock asymmetry persists as well and has widened: the Brent intraday line is stamped with a calendar date and no clock time, the WTI line with a clock time and no date ('as of 8:29 a.m. ET'), and the page's own audit stamp reads 10:01am — one hour and thirty-two minutes LATER than the WTI quote it certifies. The stamp therefore cannot be a verification timestamp for that line, on this edition as on the last. 🔴 AND ONE SERIES THIS ROW HAS BEEN TRACKING STOPS HERE FOR WANT OF DATA, WHICH IS NOT THE SAME THING AS STOPPING. The heating-oil contract HO OCT26 was followed through #188 and #190 to the nominal-mark-on-zero-open-interest stage. Bulletin #191's extract CARRIES NO HO BLOCK AT ALL — the archive says so in terms and declines to present a partial table as a table — so the correct entry for October 5 is NO READING, not 'no row'. A row that is absent from the exchange's section is a disclosure; a row that is absent from the Institute's extract is a gap in the Institute's reading. Collapsing the two would have made the series look continuous through a session nobody looked at, and would have promoted HO OCT26 to 'gone' on no evidence. 🔴 THE PRELIMINARY RUN HAS ENDED AT FIVE, AND THE LEDGER'S REFUSAL TO CALL IT A CADENCE IS WHY NOTHING HAS TO BE RETRACTED. #181 through #184 were each FINAL on first sight; #185, #186, #187, #188 and #190 were each PRELIMINARY on first sight; #191 is FINAL on first sight. Four, then five, then one. Every sweep that recorded either run declined to generalise from it, and both runs have now ended — which is the demonstration, not the exception. #189 is still NEVER READ and is counted in neither column; absorbing it would be counting an absence as an observation. The separate finalise-in-place behaviour stands on the record unchanged: #180 was read PRELIMINARY on September 19 and FINAL on September 22 at the same number and the same URL. 🔴 THE PERMANENTLY SKIPPED TRADE DATES REACH SIX, AND THE NEWEST ONE IS THE SESSION IMMEDIATELY BEFORE THE STRIP THAT NOW PUBLISHES. #185, #186, #187 and #188 rotated off the public path without ever being observed FINAL, taking Friday September 25, Monday September 28, Tuesday September 29 and Wednesday September 30 out of the strip permanently; September 11 was the first; and #190 has now joined them, so FRIDAY OCTOBER 2 will never enter the strip either. Six disclosed absences against the single finalisation-in-place #180 supplies — stated as the counts observed and not as a probability, because #180 shows the other outcome is available. Thursday October 1 is a SEVENTH missing trade date of a different kind and does not belong in that count: #189 was never read, so whether it was ever FINAL is unknown rather than negative. The rule is not relaxed to close the gap — a provisional settlement is the wrong input to a curve a reader may act on, and six disclosed absences are better than six adopted provisionals. 🔴 THE OCTOBER WTI CONTRACT'S ABSENCE REACHES FIVE OBSERVED BULLETINS, STILL WITH A GAP IN IT. #191's CL section opens at NOV26 and CL OCT26 does not appear, as in #186, #187, #188 and #190. That is five observed absences and not five consecutive ones, because #189 was never read. The exchange prints only months carrying volume or open interest, so the absence is itself the disclosure. 🔴 AND THE ARCHIVE'S OWN RECONCILIATION IS WHAT LICENSES EVERY FIGURE ABOVE, SO IT IS REPORTED RATHER THAN ASSUMED. Across the 87 contract months #191 shares with #190, backing out the FINAL point-change column reproduces the prior settlement 87 TIMES OUT OF 87 TO THE CENT. Open interest reconciles on 79 of 87, with eight revisions the exchange made without comment — CL NOV26 down 169 lots, CL DEC26 down 125, CL FEB27 down 10, BZ APR27 down 4, BZ MAR27 and BZ MAY27 down 2 each, CL DEC30 down 2, and CL MAR27 UP 1. That is a transcription-completeness check and nothing more: it establishes that the two documents were read correctly, not that either is right. 🔴 AND A SECOND INSTANCE OF A MARK THIS LEDGER HAS SEEN ONLY ONCE BEFORE. #191 lists BZ SEP28 with its point-change column reading NEW — a newly listed contract month — closing at 77.17 on 198 lots of open interest and a total-volume field printed as a dash, which is reproduced verbatim rather than explained. The only prior NEW this reference has recorded was CL JUN35 in the September 15 extract, which left zero open interest where SEP28 retains 198. Two observations of a listing event, four months apart in contract terms and three weeks apart in reading terms, are two observations. 🔴 AND THE TWO ENDINGS THIS LEDGER HAS WATCHED SHARE A TERMINUS AND NOT A SHAPE, WHICH CORRECTS HOW THE SECOND ONE WAS DESCRIBED. BZ NOV26 reached its end in TWO printed stages — traded on 19,421 lots at September 30, a nominal $103.51 marked UNCH on zero open interest at October 2, no row at October 5. CL OCT26 took FOUR — 17,075 lots, then 1,337, then a nominal $94.59 on 1,000, then nominal on zero, then absent. What the two share is the terminus and the ORDER of the stages, not the count of them. A reader told only that both contracts 'bled out and disappeared' would take the stage count for part of the mechanism, and it is not. 🔴 SUPERSEDED AND LABELLED RATHER THAN DELETED, per the Two-Clock rule — everything from here to the end of this note was measured on the OCTOBER 1-TO-2 step, a single session with both clocks in agreement, and none of it describes the two-session step this row now prints. Its figures were correct for the interval they described: NEITHER SERIES ROLLED ON THIS STEP, SO BOTH PRINTED MOVES ARE PRICE MOVES — AND THE TWO QUOTES STILL MAY NOT BE DIFFERENCED, BECAUSE THE MONTHS ARE STILL MISMATCHED. The roll test, run against the October 1 settles reconstructed from bulletin #190's point-change column: CL NOV26 $92.87 and CL DEC26 $90.84. This $92.83 open sits FOUR CENTS beneath the November settle — the tightest ordinary gap this row has recorded — while standing $1.99 ABOVE the December settle, which on a session CL NOV26 fell $1.76 would be a gap in the wrong direction by about two dollars. WTI is still quoting November; Brent is still quoting December, where it rolled on October 1. Subtracting one from the other gives $9.40 and it is not a spread. The Brent row above gives the exchange's own same-month December figure of $12.82, up $3.21 in two sessions, and records that the November spread has stopped being quotable because its Brent leg now sits on zero open interest. 🔴 ONE TEST, TWO BENCHMARKS, FIVE RUNS — TWO ROLLS AND THREE CLEAN RESULTS. The roll-detection method is the one this row was built from in September: measure the new open against the prior session's settle on each candidate month and keep the month that produces an ordinary gap. It found a WTI roll on September 23 at 70 cents against November versus $4.77 against October. It found a Brent roll on October 1 at 14 cents against December versus $5.64 against November. It cleared WTI on October 1 at 11 cents against November versus $1.89 the wrong way against December. It clears WTI again on October 2 at four cents against November versus $1.99 the wrong way against December, and it clears Brent on October 2 at eight cents against December versus $1.28 against a November mark that no longer carries open interest. A detector that only ever fires has demonstrated nothing, which is why the three clean results are reported with the same prominence as the two rolls. 🔴 AND THIS STEP SUPPLIES THE REASON THE TEST WORKS, WHICH IS NOT THE REASON THE LEDGER HAS BEEN GIVING. The phrase carried on every prior run was 'an ordinary overnight gap'. Bulletin #190 is for trade date October 2 — the SAME trade date as these quotes — so for the first time the publisher's open can be set against the exchange's own open rather than only against the prior settle. The exchange opened CL NOV26 at $93.46 and BZ DEC26 at $102.75. The publisher's opens are $92.83 and $102.23, which miss the exchange's opens by 63 and 52 cents while sitting 4 and 8 cents from the prior SETTLES. Both legs, same direction, roughly an order of magnitude. So what this publisher labels an opening quote tracks the previous session's settlement far more closely than it tracks any opening print, and the small numbers the roll test keeps finding are near-identities rather than overnight gaps. The test's discriminating power is unaffected — it identifies WHICH MONTH'S settlement a quote echoes, and a dollar-and-a-half alternative is still excluded either way — but the explanation beneath it was wrong and is corrected here. One step, two legs: a consistent reading and not yet a property of the feed. Against the exchange's own settles: Bulletin #184, trade date Thursday September 24, marked FINAL, settles CL NOV26 at $94.61. 🔴 THIS CLAUSE USED TO ADD 'and that FINAL is still the newest settlement this reference carries'; IT IS NOW FALSE, corrected in place rather than left standing inside a superseded block. Bulletin #191, trade date Monday October 5, was read marked FINAL, and it is the newest FINAL settlement this reference carries and the strip now published on the forward-curve page. #184's $94.61 is retained as the September 24 measurement it always was. Bulletins #185 through #188 and #190 were read and not adopted, all five being PRELIMINARY; #189, for trade date Thursday October 1, was never read at all, because the public path rotated before it could be fetched. The two absences are different and the Brent row above keeps them apart. The provisional path is $94.61 FINAL on September 24, $92.41 derived for September 25, $92.60 for September 28, $89.38 for September 29, $90.42 for September 30, $92.87 DERIVED for October 1 from #190's point-change column, and $91.11 printed for October 2. This $92.83 open sits four cents below the October 1 derived settle and $1.72 above the October 2 settle printed in the same bulletin — which is the ordinary shape of a quote taken at the start of a session the market then spent falling, and not a discrepancy. 🔴 THE OCTOBER CONTRACT HAS LEFT THE SECTION ENTIRELY AND HAS NOT RETURNED, THROUGH THE FIFTH AND NEWEST BULLETIN READ. #184 still printed CL OCT26 as a nominal mark at $94.59 on zero open interest — a price with no position behind it. From #186 onward the CL section opens at NOV26 and October does not appear at all; #187, #188 and now #190 are the same. #189 was never read, so the run of observed absences is four bulletins with a gap in it rather than five consecutive ones. The exchange prints only contract months carrying volume or open interest, so the absence is itself the disclosure, and the contract this series was quoting three weeks ago is now gone rather than merely empty. 🔴 THE SECOND INSTANCE OF THAT LIFE CYCLE HAS ADVANCED A STAGE, AND IT IS REPORTED AS AN OBSERVATION RATHER THAN AS A FORECAST FULFILLED. The #188 reading recorded HO OCT26 settling 4.9569 on 94 lots of open interest, down 1,536 lots in a single session, and said in as many words that this was the collapsing-while-still-listed stage and NOT a prediction of when the contract would disappear. In #190, HO OCT26 prints the same 4.9569, now marked UNCH on no volume, with open interest down a final 42 lots to ZERO. So it has reached the nominal-mark stage CL OCT26 occupied in #184, two bulletins after the drain was noticed — and BZ NOV26, described in the Brent row above, reached the same stage on the same day by a different route, losing 19,421 lots in one session rather than bleeding out over several. Two contracts, two speeds, same terminus. That the earlier note declined to put a date on it is the reason nothing has to be retracted now. 🔴 THE PUBLISHER'S INTERNAL INCONSISTENCY REPRODUCES AN ELEVENTH TIME, AND THE RESIDUAL STAYS ON THE HIGH SIDE. The same page reports WTI trading at $89.70/bbl, 'down 3.38% since yesterday's close'. Back that percentage out and the implied base is $92.8379, which sits 0.79 of a cent above the figure the same page prints as the October 2 open and 3.21 cents from the October 1 derived settle of $92.87. The Brent line on the same page lands the same way at 0.57 of a cent above its own open, so this is again the pair of benchmarks moving together and not one line's rounding. 🔴 BUT THE DISCRIMINATOR IS WEAKER THIS STEP THAN THE RATIOS SUGGEST, AND SAYING SO IS THE POINT OF REPORTING IT. The two candidates — open and prior settle — are only FOUR CENTS apart on this leg and EIGHT on Brent, against fourteen last step and over a dollar on the steps before the quotes converged. A four-to-one ratio measured across a four-cent window is a weak instrument however clean the arithmetic looks, and the paragraph above on the exchange's own opens explains why the window closed: the publisher's quote and the prior settlement have very nearly the same value, so distinguishing which one a derived base is tracking is close to asking which of two identical numbers was used. The reading is recorded with its confidence and not with the ratio alone. 🔴 AND THE PAGE'S OWN LABEL IS WRONG, WHICH IS A SEPARATE DEFECT FROM THE RESIDUAL. The percentage is captioned 'since yesterday's close'. Yesterday's close on the contract this row is quoting is the October 1 CL NOV26 settle of $92.87, and $89.70 against $92.87 is a fall of 3.413 per cent, not the 3.38 per cent printed. The implied base sits 3.21 cents from that settle and 0.79 of a cent from the open. So the number is an open-to-intraday change wearing a close-to-intraday caption, and a reader who takes the caption at face value is reading the wrong interval. What survives across all eleven readings is the weaker claim that has never failed: the percentage the page prints is not computed against the close the page says it is, so one number on that page is doing duty as two different things. 🔴 THE ARCHIVE'S SEVENTH FINDING HAS BEEN RE-RUN AND IT DOES NOT REPRODUCE — WHICH IS REPORTED AS A NON-REPLICATION AND NOT QUIETLY DROPPED. That finding compared each benchmark's quote against the bulletin's own Globex opening prices for the same trade date, and found ONE LEG TRACKING AN OPEN AND THE OTHER A SETTLE. It could not be run on the October 1 step because #189 was never read. #190 restores the conditions exactly — trade date October 2, the same date as these quotes — and on this step BOTH legs track the settle: 63 cents from the exchange's open against four cents from the settle on WTI, 52 against eight on Brent, worked in full in the paragraph above. So the split result is superseded by a symmetric one. It is retained and labelled rather than deleted because a finding measured on one step and contradicted on a later one is the ordinary case for a dated ledger, and because two observations of a publisher's sourcing habit, taken a week apart and disagreeing with each other, are grounds for weaker claims rather than for picking the more recent. 🔴 RETAINED FROM THE SEPTEMBER 24 SWEEP AND LABELLED RATHER THAN DELETED, because a roll that is explained once and then dropped leaves the series looking continuous when it is not — and because it is the precedent the Brent finding in the row above rests on. It describes the September 22 to September 23 step, which is now six steps behind this row: THE SERIES ROLLED, AND MOST OF THE $5.67 IS THE ROLL. The entry before that one ended by warning, in as many words, that “a reader should expect this publisher's series to roll without saying so, and a roll is not a price move.” It rolled on the next step, and the publisher said nothing. Bulletin #182, trade date Tuesday September 22, marked FINAL, settles NYMEX CL OCT26 at $94.59 and CL NOV26 at $90.52. The September 23 open of $89.82 sits 70 cents beneath the NOVEMBER settle — an ordinary overnight gap — and $4.77 beneath the October settle, which would be an implausible one. The October contract's open interest collapsed on that same session from 17,075 to 1,337, down 15,738 lots, so October was no longer a traded contract in any meaningful sense. The quote has moved from October to November between September 22 and September 23. What that leaves of the decline: the October-to-November spread was $4.07 on the September 22 settles, so of the $5.67 fall in the printed number roughly $4.07 is the contract changing underneath the series and roughly $1.60 is a price move. Brent, which was already on November and did not roll, fell $1.40 on the same step. Those two residuals agree to about twenty cents, which is the check. A reader who takes the printed $5.67 as a crude-oil move is overstating it by about four dollars. That entry also recorded the publisher's internal inconsistency for a fourth time, on the September 23 figures: WTI quoted at $90.76/bbl up 1.01% implied a base of about $89.85, which missed the $90.52 settle by 67 cents and landed within three cents of that day's printed open. It is superseded by the tenth-time paragraph above only in its claim of exactness, not in its substance — the implied base tracked the open on that step as it tracks the open on this one, and it has never once resolved to a close. That the tenth and eleventh readings' residuals are the smallest of the eleven does not strengthen the cross-reference: a residual that has run 3, 6, 8, 90.7, −1.5, +0.5 and now +0.79 cents, in both signs, is not a rule with exceptions — and the paragraph above on candidate separation explains why the two smallest readings are also the least informative. 🔴 AND THE MECHANISM THIS PARAGRAPH DESCRIBES HAS NOW HAPPENED TWICE, ON TWO BENCHMARKS, FOUR WEEKS APART. The Brent row above records the October 1 Brent roll, found by the same test and decomposed the same way. Two instances are two instances and are not a cadence; what they do establish is that this publisher's silent-roll behaviour is a property of the feed rather than a one-off, and that the ledger's standing warning — expect this series to roll without saying so — has now been borne out on both legs. 🔴 THE PRELIMINARY RUN NOW EXCEEDS THE FINAL RUN, AND STILL NEITHER IS CALLED A CADENCE. #181, #182, #183 and #184 were each marked FINAL on first sight, and every sweep that recorded that run declined to generalise from it — which is the reason nothing had to be retracted when it ended. #185, #186, #187, #188 and now #190 have all been seen PRELIMINARY on first sight. Five against four. A run that has just become the longer of the two is exactly the moment to repeat that the symmetry was arithmetic and the asymmetry is arithmetic too: five observations of a publisher's habit are five observations, and the first run's own end is the demonstration. Note also what is NOT in the five — #189 was never read, so it is neither evidence for the run nor against it, and a count that absorbed it would be counting an absence as an observation. The separate FINAL-in-place behaviour remains on the record: #180 was read PRELIMINARY on September 19 and FINAL on September 22 at the same number and the same URL. Two behaviours, neither predictive. 🔴 THE PRACTICAL CONSEQUENCE HAS REACHED FIVE TRADE DATES PLUS ONE OF A DIFFERENT KIND, AND IT LANDS ON THE FORWARD-CURVE PAGE RATHER THAN HERE. #185 rotated off the public path without ever being observed FINAL, so Friday September 25 will never enter the strip; #186, #187 and #188 have each done the same for Monday September 28, Tuesday September 29 and Wednesday September 30. With September 11 that is FIVE permanently skipped trade dates, against the single finalisation-in-place #180 supplies — five to one, stated as the counts observed and not as a probability, because #180 shows the other outcome is available. Thursday October 1 is a SIXTH missing trade date and does not belong in that count: #189 was never read, so whether it was ever marked FINAL is unknown rather than negative. Friday October 2 is not in either column yet, because #190 is the current bulletin and may still finalise in place. The rule is not being relaxed to close the gap: a provisional settlement is the wrong input to a curve a reader may act on, and five disclosed absences are better than five adopted provisionals. |
| Brent, end-July 2026 | $96.80 / bbl | July 31, 2026 | IEA Oil Market Report, published August 12, 2026 | Rose $25.67 over the month of July and traded a range of nearly $40 inside it. |
| Brent, intra-month peak | $105 / bbl | July 23, 2026 | IEA Oil Market Report, published August 12, 2026 | Reached on the breakdown of the mid-June ceasefire. |
| Brent, 2026 high | $126 / bbl | April 30, 2026 | CNBC, April 30, 2026 | A four-year high, reached during the blockade escalation. |
| Brent, August 2026 monthly average | $91 / bbl | August 2026 monthly average | EIA Short-Term Energy Outlook, released September 9, 2026 | Seven dollars above the July average, which the same release therefore puts near $84. A monthly average is not a spot quote and the two are routinely quoted against each other: July ended at $96.80 and averaged some twelve dollars below its own closing print. The EIA's forecast from here is about $90/bbl through the second half of 2026, $77 in 2Q27 and $67 in the second half of 2027 — a forecast, not an observation. |
| North Sea Dated, August 2026 monthly average | $91.00 / bbl | August 2026 monthly average | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | Up $7.61 on the July average. This is the Platts physical assessment, not a futures quote. Note that it lands on the same figure the EIA's separate August Brent average gives to the dollar — two publishers, two methods, one number, which is the closest thing to corroboration this reference gets on a price. |
| North Sea Dated | $113.48 / bbl | September 9, 2026 | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | 🔴 Read this against the front-month Brent futures open on that same date, September 9, which was $99.32 — the figure the rows above carried before they rolled forward through the September 10, 11, 14, 15 and 16 opens to the September 17 open they carry now. That is a spread of more than fourteen dollars between the physical assessment and the front-month futures open on one date, and it is not a reporting error in either one: a physical assessment carrying a large premium to paper is the price signature of prompt physical scarcity, which is the same thing the inventory and refining ledgers show from the volume side. The Institute prints both and names the instrument on each, because a reader who takes one of these as “the oil price” on September 9 will be fourteen dollars wrong depending on which one they took. |
| North Sea Dated, September 2026 monthly average | $116.89 / bbl | September 2026 monthly average | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Up $26.17 on the August average — a 29% move in one month, and the largest monthly change in any price on this reference. The August average is retained in the row above under its own citation rather than being overwritten. Same instrument: the Platts physical assessment, not a futures quote. |
| North Sea Dated | about $134 / bbl | October 14, 2026, at time of writing | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Up about 43% and quoted by the IEA as “at the time of writing”, which is a publication-clock figure rather than a measured close; it is carried here as such. Read it against the ICE Brent row below on the same date: the physical assessment is running roughly twenty-seven dollars above the paper benchmark, against a fourteen-dollar gap on September 9. The physical premium that the September row calls the price signature of prompt scarcity has not closed — it has roughly doubled. The IEA also notes other North Sea grades assessed as high as $160/bbl, which is the widest quality spread this reference has carried. |
| ICE Brent futures | around $107 / bbl | October 14, 2026, at time of writing | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Up about 13% since the start of September. The September edition's “about $105 / bbl” on September 11 is retained in the row below under its own citation; note that the two figures are two dollars apart over five weeks in which the physical assessment moved more than twenty, which is the whole argument for naming the instrument on every price this reference prints. The IEA's own phrase is again “at the time of writing”, a publication-clock figure rather than a measured close. |
| ICE Brent futures | about $105 / bbl | September 11, 2026, at time of writing | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | Up about $21 since August 1 and roughly 45% above the pre-war level. The IEA's own phrase is “at the time of writing”, which is a publication-clock figure rather than a measured close, and it is carried here as such. Superseded as the current reading by the October row above, retained because a dated quote is a measurement of its own date. |
| Diesel, cracked-product highs | above $200 / bbl in Singapore; records near $220 / bbl in Europe and the United States | mid-September 2026, since eased | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | 🔴 THE PEAK IS NOW BEHIND THE READER AND THE ROW SAYS SO. The September edition had US diesel “past $200 / bbl” in early September and rising; the October edition dates the all-time highs to mid-September in Singapore and to records near $220/bbl in Europe and the United States, and says prices have since eased. A row that reported the high without reporting that it had passed would be accurate about the measurement and wrong about the market. The easing is not relief in the margin, which is the part a reader will get backwards: European diesel cracks fell from about $100/bbl to close to $60/bbl by early October because freight and crude costs rose, not because the product got cheap. See the refining ledger. The superseded September reading — about 94% above the pre-war level, roughly double the percentage move in crude — is retained in the row below under its own citation. |
| US diesel | past $200 / bbl | early September 2026 | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | About 94% above the pre-war level — roughly double the percentage move in crude over the same period. Diesel and gasoil are about 30% of global oil demand, and the Gulf's net exports of them are running at just over a quarter of pre-war volumes. This row is the price consequence of the export figures on the refining ledger. Superseded as the current reading by the row above, retained under its own clock. |
A price is the fastest-moving figure on this page and the one most quickly stale. The two opening quotes above were accurate at the moment stated and should be treated as historical the following day; they are included to fix the level, not to serve as a live feed. The dated highs are the more durable entries. West Texas Intermediate traded above Brent by the widest margin since 2009 during this episode, which inverts the normal relationship: when seaborne transit is disrupted, an inland benchmark can command a premium over a seaborne one.
| Series | Figure | Measured | Source | What it carries |
|---|---|---|---|---|
| Global oil supply | 102.3 mb/d | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Rose 2.5 mb/d on the month, reversing August's 1.6 mb/d fall and then some. OPEC+ supplied 2.1 mb/d of the gain, led by Saudi Arabia; the United Arab Emirates drove what growth came from outside the group. The October edition also raised its own 3Q26 supply estimate by 190 kb/d against the September Report, and it expects the second half of 2026 to average 101.4 mb/d against 96.4 mb/d in 2Q26 — which is the scale of the restoration now assumed. The IEA's own caution on it is one sentence: “Renewed attacks in the Gulf leave the recovery fragile.” |
| Gulf production shut in | more than 10 mb/d | August 2026 | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | The IEA states a bound rather than a point estimate this month — “more than 10 mb/d of Gulf output remained shut in amid heightened security risks” — against 8.3 mb/d for July. The EIA's read on the same month is 6.72 mb/d against a different baseline; see the row below and the conflict note beneath this table. A bound and a point estimate are not the same kind of figure and the Institute does not convert one into the other. 🔴 RETAINED ON AN AUGUST CLOCK, DELIBERATELY. The October 2026 edition, published October 14, 2026 and read for this sweep, does not restate a Gulf shut-in volume in the material the Institute read. What it gives instead is the export side: Gulf crude exports up 4.4 mb/d to 13.9 mb/d, over 75% of the pre-war level; total oil flows through Hormuz up 3.7 mb/d to 10.9 mb/d; bypass-route exports up 970 kb/d to 5.6 mb/d. Those are a different quantity and are not a substitute for this one, so this row keeps the September edition's figure and the September edition's citation rather than being re-stamped forward. “Not in the release we read” is the claim here; the Institute is not in a position to say the publisher has stopped measuring it. |
| Gulf production shut in — the EIA's read | 6.72 mb/d | August 2026 | EIA Short-Term Energy Outlook, released September 9, 2026 (forecast completed September 3, 2026; next release October 6, 2026) | The second of the two reads, now on the SAME month as the IEA row above rather than a month behind it. Up from 4.98 mb/d in July. Measured as closure-related shut-ins against each producer's February 2026 output, which totalled 25.2 mb/d across the seven affected countries. August by country: Saudi Arabia 3.55, Iraq 1.16, Iran 1.00, Kuwait 0.68, Qatar 0.20, Bahrain 0.13, United Arab Emirates nil. |
| Gulf shut-ins, forward path | 6.23 mb/d in 3Q26, 5.66 in 4Q26, 2.72 in 1Q27 | quarterly forecast, completed September 3, 2026 | EIA Short-Term Energy Outlook, released September 9, 2026 (forecast completed September 3, 2026; next release October 6, 2026) | A forecast, not an observation, and the only figure in this group that is not a measurement. It assumes partial restoration and no further escalation. |
| OPEC+ effective spare capacity | 0.27 mb/d | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Up from 0.22 mb/d in August, which had itself fallen 80% from July's 1.09 mb/d. Against 102.3 mb/d of world supply it is 0.26% of the market — so the one-month recovery of 2.5 mb/d in supply bought the market almost nothing in the way of a cushion, which is the point of carrying this row next to the supply row rather than on its own page. Reachable within 90 days and sustainable, excluding shut-in Iranian and Russian crude. The split is 0.10 mb/d for the OPEC-8 cutting group against 0.07 in August, and 0.11 mb/d for the non-OPEC half against 0.15 — so the group total rose while the non-OPEC half it had all been sitting in actually fell. Read the figure against the production table rather than alone: Saudi Arabia produced 8.30 mb/d in September against a sustainable capacity of 12.11, and the IEA still credits it with no effective spare at all. Unused capacity and effective spare capacity are different quantities — capacity that cannot reach a buyer is not a shock absorber — and the whole of this reference's argument about spare capacity turns on that distinction. This remains the more generous of the two available reads: the EIA's own surplus-capacity series, in the row below, is at zero for this quarter. One transcription note, because a reader checking the arithmetic will find it: the IEA's own effective-spare column does not foot against its own country rows at the roundings it prints, and the figures on the production table are carried as printed rather than reconciled. |
| OPEC surplus crude oil production capacity — the EIA's read | 0.0 mb/d in 3Q26 and 4Q26 | quarterly forecast, made August 11, 2026 | EIA Short-Term Energy Outlook, August 11, 2026 release, current/previous forecast comparison table (International Crude Oil and Liquid Fuels) | Against 3.4 mb/d for calendar 2025 and 1.7 mb/d in 1Q26. The annual averages are 0.5 mb/d for 2026 and 1.8 mb/d for 2027. This is not the same quantity as the IEA row above — it covers OPEC rather than OPEC+, and it is a quarterly forecast rather than a monthly observation — so the two are not interchangeable. What matters is the direction of the disagreement: the figure a reader is most likely to see quoted, the IEA's 1.09 mb/d, is the higher of the two. |
| Global oil demand, 2026 | declining 2.6 mb/d to 102.4 mb/d | 2026 full-year forecast, made October 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | A forecast, not an observation, and 100 kb/d steeper than the 2.5 mb/d decline the September edition carried — which was itself 940 kb/d steeper than August's 1.6 mb/d. Three editions, three successive downgrades, and this is the first of them to print a level alongside the change. Middle distillates carry about a quarter of the whole decline. The quarterly path now reads a contraction of 5 mb/d in 2Q26, 3.7 mb/d in 3Q26 and 2.5 mb/d in 4Q26, so 4Q26 is still contracting and is contracting half a million barrels a day faster than the September edition expected. On 2027 the October edition is more explicit and less reassuring than its predecessor: a recovery of about 2 mb/d, which still leaves 2027 consumption 620 kb/d — 0.6% — below the 2025 level. The September edition's 2.6 mb/d rebound had been framed as narrowly offsetting this year's losses; on this edition's numbers it does not offset them. |
| Global oil supply, 2026 | declining 5.8 mb/d to 100.7 mb/d | 2026 full-year forecast, made October 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | A forecast. The level is unchanged from the September edition at 100.7 mb/d and the decline is 100 kb/d deeper, which is a restatement of the 2025 base rather than a change in the outlook for this year. The 2027 rebound is raised to 7.7 mb/d from the September edition's 8 mb/d — read the two together rather than as a reversal, since a deeper 2026 trough and a shallower 2027 recovery land in much the same place. |
| Global observed stocks | fell 10 mb | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | A small net number sitting on top of two very large offsetting ones, and the composition is the finding rather than the total. On-land inventories fell 113 mb while oil on water rose 103 mb — which is barrels in transit standing in for barrels in tanks, not a market in balance. Non-OECD crude stocks fell 76 mb, 78% of it in China; OECD stocks fell 37 mb. Cumulative draws since end-February now stand at 509 mb, an average of 2.4 mb/d, and the IEA states that on-land stocks are at their lowest since at least 2016. Note the asymmetry a reader should take away: the headline monthly change understates the position by an order of magnitude, because the water leg is a float that can reverse in weeks and the on-land leg is the one that sets prompt scarcity. |
| 3Q26 balance | not obtained | September and October 2026 editions — figure not in the public highlights of either | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | GAP, stated rather than carried forward, and now checked against two successive editions rather than one. The August edition put the 3Q26 balance at a deficit of 1.8 mb/d, more than double the roughly 800 kb/d it had estimated a month before that. Neither the September nor the October public highlights restate a quarterly balance figure, and the Institute will not carry an August-clock deficit under a later source line or infer one from the supply and demand revisions above. What the October edition does say qualitatively is that “oil market balances are forecast to remain in deficit through the end of the year” — a direction, not a magnitude, and it is carried here as a direction. The number is available in the full subscriber Report and is not on this reference until it has been read there. |
| Global inventory draw, 2026 to date | about 400 mn bbl | August 2026 | EIA Short-Term Energy Outlook, released September 9, 2026 (forecast completed September 3, 2026; next release October 6, 2026) | An independent read. The IEA's own figure for cumulative draws since end-February is 509 mn bbl on its October 2026 edition, against 507 mn bbl on the September edition, so the two agencies have separated by roughly a hundred million barrels and stayed there — having agreed within ten million on the same quantity two months ago, when the IEA had 410 against this 400. The EIA figure in this row is still the September STEO's and carries that clock; it is the IEA side of the comparison that has moved. Where the two diverge the Institute prints both rather than choosing. The EIA puts the draw at 3.9 mb/d in 2Q26 and forecasts 3.0 mb/d in 3Q26 and 1.7 mb/d in 4Q26. Two agencies converging on a stock change is worth more than either alone; on the size of the production shut-in, immediately above, they do not converge. |
The two agencies that measure this disruption do not agree on its size, and the gap is not small. For August 2026 the IEA states that more than 10 mb/d of Gulf production remained shut in; the EIA's Table 1 totals 6.72 mb/d for the same month. Two things are worth noticing before the numbers are compared at all. The first is that they are now measurements of the same month — until September 11, 2026 the freshest IEA figure was a month behind the EIA's, and part of the apparent disagreement was simply that. It was not all of it: on July 2026, the last month both agencies now describe, the IEA had 8.3 mb/d against the EIA's 4.98, so the gap survives the alignment of the clocks. The second is that the two figures are not the same kind of object. The EIA publishes a point estimate built from named countries: closure-related shut-ins measured against each producer's February 2026 output, seven countries totalling 25.2 mb/d before the closure. The IEA publishes a floor — 'more than 10' — with no stated baseline and no country breakdown in the public highlights. A floor cannot be subtracted from a point estimate to yield a difference, and the Institute does not compute one. What can be said is the direction and the rough size: the IEA's floor sits about half again above the EIA's estimate, and the reason is almost certainly the baseline, because the IEA's figure has consistently implied a wider pre-war Gulf aggregate than the EIA's 25.2 mb/d and therefore carries losses the EIA does not attribute to the closure. Neither figure is wrong. They answer different questions, and a reader who takes one as a correction of the other has been misled by a definitional problem dressed as a factual dispute. The Institute prints both, dates both, and names what each one is — which is the only treatment that leaves the reader able to choose.
Four of these rows are forecasts and say so. A forecast made in August 2026 about calendar 2026 is a different kind of statement from an observation of July 2026, and the Institute does not print them in the same voice. The demand rows in particular are worth reading carefully: demand is not measured, it is inferred from supply and stock changes, and in a period of large stock movements the inference carries the error.
| Series | Figure | Measured | Source | What it carries |
|---|---|---|---|---|
| Global refinery crude throughput | 81.4 mb/d | August 2026 | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | A summer peak. Up 960 kb/d on the month from July's 80.9 mb/d, and still 4.2 mb/d below August 2025, with the losses spread across the Middle East, Russia and crude-importing economies in Asia. 🔴 RETAINED ON AN AUGUST CLOCK. This is the freshest MEASURED month the Institute has read. The October 2026 edition, published October 14, 2026, speaks about refinery runs in the forward direction — an October seasonal low and a year-end rebound, both in the row below — and the material read for this sweep does not carry a measured September throughput. A forecast for the current month is not a substitute for a measurement of the last one, so this row keeps the September edition's figure and the September edition's citation. The annual figure that used to sit in this note has moved and both clocks are given: the September edition forecast global runs declining 2.6 mb/d to 81.5 mb/d across 2026, and the October edition puts the same quantity at a decline of 2.3 mb/d to 81.7 mb/d, with a rebound of 2.8 mb/d in 2027. On either edition's number, August ran at about the level the full year is expected to average. |
| Global refinery runs, forward path | 81.3 mb/d in October 2026, rebounding to 84.6 mb/d by year-end | monthly forecast, made October 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | A forecast, not an observation, and the only throughput figure on this ledger that is not a measurement. October is the seasonal low — autumn maintenance — and the 3.3 mb/d climb the IEA expects from it into December is the steepest two-month move in runs this reference has carried. Read it against the margin row below rather than on its own: a run recovery of that size is a forecast about refiners choosing to process crude at margins that have just been cut in half outside the United States. |
| Global refining capacity | about 103.3 mb/d | 2026, across 650+ refineries | Industry compilation | A capacity figure, not a throughput figure. Asia-Pacific holds roughly 36%. |
| US operable atmospheric distillation capacity | 18.2 mb/cd | January 1, 2026 | EIA Refinery Capacity Report | Down more than 250,000 b/cd, about 1%, from January 1, 2025. |
| US operable refineries | 131 | early 2026 | EIA Refinery Capacity Report | Down from 132 at the start of 2024 and 2025. |
| Refining margins | US margins at new highs; Europe and Asia crushed | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | 🔴 THE DIRECTION REVERSED AND THE GEOGRAPHY INVERTED. The September edition had Atlantic Basin margins at record levels on sharply higher diesel cracks, with Singapore the laggard. One month later the IEA's own sentence is that “record freight and crude costs crushed margins in Europe and Asia, while discounted feedstocks propelled US margins to new highs.” So the Atlantic Basin has split in two: the United States is still setting records and Europe is no longer in the same trade. The mechanism named is not demand — it is the cost of getting a barrel to a refinery, and a US refiner buying a discounted domestic or near-shore feedstock does not pay it. The crack itself puts a number on the European half: diesel cracks there fell from about $100/bbl to close to $60/bbl by early October, a 40% compression in roughly a month, which the IEA discusses under the heading “Surging Freight Rates Cut into Refining Profitability.” The superseded August record-margin reading is retained in the September edition's own citation and is not re-asserted here. |
| Middle East and Russian net diesel and gasoil exports | 820 kb/d | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Down 1.3 mb/d and more than 60% below a year earlier. Note the scope: this is the Middle East and Russia COMBINED, which is the aggregate the October edition publishes, and it is not the same quantity as the Gulf-only 390 kb/d the row below carries for August. The two are not comparable and the Institute does not net one against the other. What the combined figure adds is where the replacement barrels came from: United States and Chinese net exports were up 610 kb/d year-on-year, but already 150 kb/d off their own August peak — so the substitution is real, partial, and has stopped growing. Global gasoil exports ran 4.7 mb/d in September, 700 kb/d below a year earlier, and 4.7 mb/d across the first nine months of 2026, about 10% below the year-earlier period. |
| Gulf net diesel and gasoil exports | 390 kb/d | August 2026 | IEA Oil Market Report, September 2026 edition, published September 11, 2026 (reporting month August 2026) | Just over a quarter of pre-war levels. Combined Gulf and Russian net exports of diesel and gasoil were 1.6 mb/d below February 2026, when the two together accounted for almost 45% of global seaborne trade in the product. 🔴 RETAINED ON AN AUGUST CLOCK, DELIBERATELY. The October edition publishes the Middle East and Russia as one combined figure, in the row above, rather than restating the Gulf alone. A combined aggregate cannot be decomposed back into its Gulf component, and this reference will not re-stamp a Gulf-only figure with an October source line it was not read from. |
| Gulf crude exports | 13.9 mb/d | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Up 4.4 mb/d on the month and now above 75% of the pre-war level — the single largest monthly recovery in any flow on this reference. Total oil movements through the Strait of Hormuz rose 3.7 mb/d to 10.9 mb/d, and exports routed around the Strait rose a further 970 kb/d to 5.6 mb/d, so both the direct and the bypass channel widened at once. This is a crude figure. The product channels are in the row below and they have not recovered with it. |
| Gulf refined product and LPG exports | 1.9 mb/d of products, 670 kb/d of LPG | September 2026 | IEA Oil Market Report, October 2026 edition, published October 14, 2026 (reporting month September 2026) | Products up 210 kb/d on the month against a pre-war 4.7 mb/d; LPG up 90 kb/d against a pre-war 1.5 mb/d. So products are running at about 40% of pre-war and LPG at about 45%, while crude in the row above is past 75%. The crude channel is recovering and the product channel is not, which is the whole of the diesel story on this reference — and the gap has widened rather than closed since the September edition, because the crude recovery is the part that moved. Two figures are printed here rather than one total, because the IEA publishes them separately and adding them would be the Institute's arithmetic rather than the publisher's measurement. The superseded August reading — Gulf TOTAL oil exports of about 13 mb/d, nearly half the pre-war level — is retained in the September edition's own citation. It is a different scope from either figure in this row and is not re-asserted here. |
Refining capacity and refinery throughput are different quantities and are routinely conflated. Capacity is what the plants could process; throughput is what they did process. The ratio between them — roughly 79% in July 2026 — is the number that matters, and it is computed on this page from the two figures above rather than quoted. It is quoted to the whole per cent for a reason: It is an industry compilation, not a figure the Institute has traced to a dated primary publication, and global refinery utilisation is computed against it. The numerator — August 2026 throughput — is dated and primary. The denominator is not. A fresher numerator over the same undated denominator does not improve this figure; it makes it look newer than it is, which is worse. Every utilisation figure quoted on this reference therefore carries a dated top and an undated bottom, and that is why it is quoted to the nearest whole percentage point and never to a decimal. The refining ledger carries the capacity clocks, the American plant count, and the crack spreads that say what a refiner earns per barrel.
There are always two clocks running on a figure. The publication clock is when the page was prepared, and it covers the whole document. The measurement clock is when the figure itself was measured, by whoever measured it — and it is set by the source, not by the author.
A date at the top of a page does not protect a figure whose own measurement date is older. That is the failure that document dating misses entirely, and it is why every table here carries a Measured column instead of a single date in the footer.
In oil the highest-risk categories are: reserve figures, which are annual, self-reported and published months after the year they describe; sustainable capacity, which is a standing judgement with no monthly date of its own; any figure carried forward from an earlier piece; and any price older than a day. The reserve total on this page describes year-end 2025 and was published in 2026. The capacity figures behind the spare-capacity ratio may predate the damage that made them unreachable. Both are stated, and neither is smoothed over.
One practical example from building this page. The publicly posted current summary of the American weekly petroleum status report was found to be serving a frozen copy from a week in December 2025 — fourteen weeks stale relative to the live series, at a URL that presents itself as current. Nothing on the file said so. A reader who trusted the URL rather than the date inside the document would have published a fourteen-week-old inventory figure as this week's. The rule caught it, which is what the rule is for.
Every term used across the 13 ledgers, defined once. Sortable.
| Term | What it means, plainly |
|---|---|
| Barrel (bbl) | 42 United States gallons, or about 159 litres. The unit survives from the wooden barrels of the 1860s Pennsylvania fields and has no physical justification beyond custom. |
| mb/d | Million barrels per day, the standard flow unit. The world produced roughly 102.3 of them in September 2026. Some sources write kb/d for thousand barrels per day; 1,000 kb/d is 1 mb/d. |
| b/cd | Barrels per calendar day, used for refinery capacity. It differs from barrels per stream day because no refinery runs 365 days a year, and quoting one where the other is meant overstates capacity by several per cent. |
| Proven reserves | The volume a country's reporting says is recoverable under existing economic and operating conditions. A bookkeeping category, not a geological one: it rises when prices rise, with no new discovery required. Mostly self-reported and not independently audited. |
| Reserve life, or the R/P ratio | Proven reserves divided by one year's production. A units conversion that answers how many years the current reserve base would last at the current rate. No country produces at a flat rate to its last barrel, so it is a scale indicator and never a forecast. |
| Sustainable production capacity | The maximum output the International Energy Agency judges a country could reach within ninety days and hold for an extended period. A judgement rather than a measurement, and it carries no monthly date of its own. |
| Effective spare capacity | Sustainable capacity that is also actually available — so it excludes crude shut in by damage or sanction. This is the number that determines whether an outage is a logistics problem or a price problem. The IEA measured it at 0.27 mb/d in September 2026 and is the only institution that publishes a monthly series of it. |
| OPEC and OPEC+ | OPEC is the twelve-member organisation itself. OPEC+ adds non-member producers, most significantly Russia and Kazakhstan, that coordinate on output. Nearly all production management now happens at the OPEC+ level, and the OPEC-8 is the subset holding the current voluntary cuts. |
| Brent and WTI | The two dominant crude benchmarks. Brent is a seaborne North Sea light sweet grade that most of the world prices against; West Texas Intermediate is priced inland at Cushing, Oklahoma. The spread between them is mostly a shipping and access number, not a quality one, and it inverts when transit is disrupted. |
| Commercial inventories | Crude and product held by companies for operating and trading purposes. Reported weekly in the United States and compared against a five-year average for the same week of the year, which is how seasonality is removed. |
| Strategic Petroleum Reserve | Government-owned crude held in salt-dome caverns on the United States Gulf Coast for release during a supply emergency. Authorised storage capacity is 727 million barrels. A release is sold to refiners, so it moves into commercial inventories rather than disappearing. |
| Oil on water | Floating inventory: cargoes at sea, in transit, or waiting to discharge. It is counted in global observed stocks, and a fall concentrated in this pool points to a constraint on movement rather than on production or storage. |
| Observed stocks | Inventory in locations that report their levels. A large share of world inventory sits in locations that do not — China's strategic stockpile most consequentially — so a change in observed stocks is a change in the observed part and not a census. |
| Crack spread | The margin between the price of crude and the prices of the products refined from it. Record cracks with refineries running below capacity indicate a shortage of crude arriving, not a shortage of capacity to process it. |
| Refinery utilisation | Throughput divided by capacity. It is not a measure of how hard refiners are trying; a plant with no crude to run reports low utilisation and high margins at the same time. |
| Rig count | The number of rotary rigs actively drilling, counted by Baker Hughes weekly in North America and monthly worldwide. It is a current measurement of a future quantity: a rig turning today is a well producing in six to nine months. |
| Contango and backwardation | The shape of the futures curve. In contango later months cost more than the front month, which pays traders to store oil; in backwardation the front month costs more, which pays them to sell inventory. The shape therefore drives stock behaviour and is not merely a description of it. |
| Measurement date | The Institute's own convention, applied on every page here: the date a figure was measured, by whoever measured it, as distinct from the date the page was published. A page dated today can carry a figure measured fourteen months ago, and only the measurement date reveals it. |
Provenance. Reserve figures are transcribed from the OPEC Annual Statistical Bulletin; production, spare-capacity, stock and refining figures from the International Energy Agency Oil Market Report; United States production, refinery capacity and inventory figures from the Energy Information Administration; rig counts from the Baker Hughes North America and Worldwide rig counts; benchmark prices from published daily quotes. Each row names its source and the date the figure was measured. Where a figure was not published, or where a published series does not itemise a country, the cell reads n/d rather than carrying a fabricated estimate. Derived quantities — reserve life, shares of world total, capacity utilisation, spare capacity as a share of supply — are computed from the figures shown and are not separately sourced; they inherit the measurement dates of their inputs, and where the two inputs run on different clocks the page says so. The Baratelli Institute is a publisher operating under the Lowe v. SEC publisher exception. Nothing here is investment advice, a recommendation, or an offer to buy or sell any security or commodity. Figures change; verify against the primary source before relying on any of them.