The Forward Curve: Contango and Backwardation

The curve, plotted — both benchmarks, five years of delivery months

Settlement prices for trade date Thursday, October 15, 2026. Not a forecast: each point is a price at which someone transacted for delivery in that month.

Source: CME Group Daily Bulletin, Section 61, Energy Futures Products, Bulletin #199, FINAL, settlements for trade date Thursday, October 15, 2026, retrieved October 17, 2026. The chart is generated from the parsed bulletin at build time rather than drawn, which is why it carries no smoothing and no axis break: it is the same data as the two tables below, rendered.

What moved since the last bulletin this page carried

The Institute keeps the bulletin it superseded and parses both, so the reference can print the move and not only the level.

Sources: CME Group Daily Bulletin, Section 61, Energy Futures Products, Bulletin #198, FINAL, settlements for trade date Wednesday, October 14, 2026; and CME Group Daily Bulletin, Section 61, Energy Futures Products, Bulletin #199, FINAL, settlements for trade date Thursday, October 15, 2026. Both artifacts ship with this reference and both are parsed at build time. Every figure in the table is a subtraction between the two documents, so no number here can drift away from either of them. Elapsed: one calendar day, one trading session — and the two bulletins ARE consecutive. They are #198 and #199, with no bulletin number between them and no weekend between the trade dates, so the move below is a true one-session change rather than an aggregate reported as one step. That is the narrowest comparison this reference can carry, and it is the first single-session advance it has made since September 24. The step it replaces was seven sessions wide, which is the direct cost of the FINAL-only rule when the rotating path does not serve a readable FINAL bulletin on consecutive days; it served one here. BOTH legs are like-for-like and neither rolled — CL NOV26 and BZ DEC26 are the front contracts in #198 and in #199 alike, and the month roster is identical on both, 65 WTI rows and 26 Brent rows in each document — so no figure in the comparison carries a contract change or a composition effect, and the spread between the fronts can be read as a spread. One qualification the Institute states rather than absorbs: #199 was read FINAL on ONE retrieval, not two, so the mark is this reference's reading of a single document at a single moment.

The front month moved +$2.88 a barrel, or +3.0%, in one trading session. The November 2031 contract, five years out, moved −$0.31. That is the whole shape of this market in two numbers: the near end is repricing a physical disruption and the far end is not moving at all, because nothing about the long-run cost of producing a barrel changed in a week.

The width of the curve is the figure to watch rather than the level. Backwardation twelve months out went from $14.43 to $17.84 — the market increased what it is paying holders of physical oil to release it, which is a statement about tightness that survives even if one thinks the outright price is wrong.

The full WTI strip — 65 listed contract months

NYMEX Light Sweet Crude Oil (WTI), contract code CL. Physically delivered at Cushing, Oklahoma. Every spread and every share in this table is computed from the settlement column, not transcribed.

Source: CME Group Daily Bulletin, Section 61, Energy Futures Products, Bulletin #199, FINAL, settlements for trade date Thursday, October 15, 2026, retrieved October 17, 2026. The open-interest column sums to 1,880,704 contracts, which is the total the exchange prints for the contract on the same page. That agreement is the Institute's check that the strip above is complete rather than a selection: a missing month would break the total, and the build fails if it does.

Read the open-interest column before the price column. A settlement is published for every listed month out to the end of the strip, and a reader who plots all of them gets a smooth curve that is mostly not a market. The far months carry open interest in the dozens against hundreds of thousands at the front: those settlements are the exchange marking a position for margin purposes, not a price at which anyone is transacting. The curve is informative where the money is, and the open-interest column is how a reader tells the difference.

The concentration is worth a number rather than an adjective. 59.7% of this strip's open interest sits inside the first twelve months and 84.6% inside the first twenty-four, across 65 listed months running to December 2036. The first month with fewer than a thousand contracts open is August 2029, at 493 — against 84,922 in the front month. A settlement of $49.43 for December 2036 is a real published number and it is not a market view of 2036.

The full Brent strip, for comparison

NYMEX Brent Oil Last Day Financial, contract code BZ. Financially settled against the Brent index rather than physically delivered, and listed thinly beyond the front two years.

Source: CME Group Daily Bulletin, Section 61, Energy Futures Products, Bulletin #199, FINAL, settlements for trade date Thursday, October 15, 2026, retrieved October 17, 2026. Open interest sums to 225,071 contracts, matching the exchange's printed total for the contract.

Two things a reader should not mistake. Brent's front listed month here is December 2026 while WTI's is November 2026, because the two contracts expire on different schedules — so the raw gap between the two front settlements, $7.95, is partly a difference of delivery month and not only the Brent-WTI quality and location spread. Compared on the same delivery month, January 2027, Brent settles at $103.17 against WTI's $94.89, a spread of $8.28. That is the comparison worth making, and it is the one most commentary does not make.

And this Brent contract is not the Dated Brent physical assessment quoted in the press. It is the NYMEX financial look-alike, settling against the index. The figures on the hub are a different instrument measured on a different clock, and the two should not be set against each other as though one confirmed the other.

The shape, measured rather than inferred

The same question the rest of this page explains, answered from the strip above by subtraction.

Every figure computed from the settlement column of the two tables above. Sign convention: front minus later, so a positive number is backwardation.

Both benchmarks are in backwardation on this date, and steeply. WTI's front month at $97.97 stands $17.84 above the same benchmark twelve months out, which is 18.2% of the front price. Read that through the third column of the table further down this page and it says the market is paying holders of physical oil to release it, at a rate no storage economics can offset. That is a measurement now, not the inference this page used to carry.

Note where the backwardation stops. The curve falls steeply for about two years and then flattens: from November 2028 at $72.01 to November 2031 at $59.96, it gives up only $12.05 across three further years. The near end is pricing a disruption; the far end is pricing a long-run supply cost, and has barely moved. A reader who wants the market's view of the medium term rather than of this month's shortage should look at the flat part.

The two shapes, and what each one pays someone to do

The third column is the one to read carefully. A futures curve is not a prediction that the market will be at those levels; it is a set of prices at which someone will transact today for later delivery. The trade in that column is available to anyone with a tank, which is why the shape shows up in physical inventory within weeks rather than being merely an opinion about the future.

Why the shape drives the inventory cycle rather than describing it

A commercial oil inventory is not a buffer someone maintains out of prudence. It is a position with a carrying cost: tank rent, insurance, financing, and the loss if the price falls. In contango, the forward market pays that cost for you and a little more, so filling a tank is a low-risk trade rather than an expense. In backwardation the market charges you twice for the same tank, so emptying it is the trade.

That is why inventory data and curve shape must be read together and why reading either alone misleads. A stock build in contango is not evidence of weak demand; it is evidence that the carry trade is on. A stock draw in backwardation is not necessarily evidence of a shortage; it is what the curve is paying holders to do. The mistake is to treat the inventory number as an independent measurement of tightness when it is partly a consequence of the price structure.

There is a second effect that matters to anyone holding an oil fund rather than a barrel. A fund holding front-month futures has to roll its position forward each month. In contango it sells the cheap expiring month and buys a dearer one, losing a little each time; in backwardation the roll gains. Over a year of steep contango that mechanical loss can exceed the change in the spot price itself, which is how an oil price can rise while an oil fund falls. It is not a tracking error. It is the curve.

How often each shape holds

Across the front-month against six-month WTI spread since 1985, the curve has been in contango about 42 per cent of the time and in backwardation about 58 per cent.

CME Group market insight. The Institute has not established a publication date for this statistic and it is printed here as undated. It is carried on this page because the proportion is instructive — backwardation is the more common state, which surprises most readers, who meet the term during a crisis and assume it is the exception. An undated statistic is a weaker citation than anything else on this reference, and it is labelled as one rather than given a date it does not have.

How the table above got here, including the part that was wrong

This page carried no strip until September 9, 2026 and said so at length. The correction is printed rather than quietly removed, because the reasoning that produced the gap is a mistake other researchers are making.

The gap this section previously described is closed, and how it closed is worth stating. The reasoning was that EIA's futures series ends April 5, 2024 and that CME's settlements page renders in the browser rather than in the page source, therefore no reproducible public strip existed. Both premises were true. The conclusion did not follow. CME's publication of record is the Daily Bulletin, and Section 61 of it carries the entire listed strip for every energy contract, free, as a dated document, and had done so every business day throughout. The table below is that document. What was missing was not the data.

The archived source text ships with this reference as cme_daily_bulletin_s61_20261015.txt rather than being transcribed into the build, and the tables are parsed from it. The bulletin it superseded ships alongside it as cme_daily_bulletin_s61_20261014.txt, which is what the comparison table higher up this page is computed against. Three things about this citation a reader should know before relying on it. The bulletin publishes at a URL ending /current/, which rotates to the next trade date every business day — so the link that produced this table will not produce this table tomorrow, and the archived artifact rather than the link is what the Institute holds. This table is a FINAL, as every strip this reference has ever carried is: the exchange marks provisional settlements PRELIMINARY and revises them, and provisional settlements are the wrong input to a curve a reader may act on. And this reference's strip history skips a trade date. Friday, September 11 is absent from it, and the reason is worth setting out because it is a working example of a rule this whole build runs on. Bulletin #175 for September 11 was read on September 11 and read again on September 15, and was marked PRELIMINARY on both readings; by September 16 the /current/ path had rotated past it. The Institute states what it read rather than what it infers: the Section 61 extracts available to these builds carried no FINAL for September 11, and whether one exists behind DataMine or elsewhere is not something a reading of the public path can establish. Not in the release we read is a different claim from not published, and only the first is ours to make. 🔴 The split between sections recorded here on September 16 has closed, and how it closed is the part worth keeping. On that date the exchange's Section 00 product index reported Bulletin #177 for trade date Tuesday, September 15 as FINAL while Section 61, on the identical /current/ path and fetched minutes apart, still returned Bulletin #176 for September 14 — one publisher, one URL stem, two different trade dates depending on which section was asked. Read again on September 17, Section 61 serves #177 for September 15, which is the table below, and serves nothing for Wednesday, September 16. So the two sections were not disagreeing about which bulletin was final; they were describing one bulletin at two moments in its publication. 🔴 The reading that followed from that — that Section 61 trails the product index by roughly a business day — is withdrawn. It was an observation drawn from two days, and two days is not a cadence. On September 18 Section 61 served the immediately preceding trade date rather than the one before it, which the claim does not survive. It is labelled rather than deleted because the error in it is the instructive part. This reference follows Section 61 either way, because Section 61 is the document the settlements are actually in. 🔴 This table sat still from September 15 to September 22, and the explanation carried here through five sweeps for why it was sitting still was wrong. That explanation ran: the /current/ path serves one bulletin at a time, a bulletin marked PRELIMINARY rotates away before it is ever marked FINAL, and so a trade date seen only as provisional is lost to this reference permanently. It is set out in those terms rather than deleted because the error in it is the instructive part. On September 19 the path served Bulletin #180 for Friday, September 18, marked PRELIMINARY. On September 22 the path served Bulletin #180 for Friday, September 18, marked FINAL — the same bulletin, the same trade date, the same URL. The exchange finalises a bulletin in place. It does not publish a separate FINAL for the path to rotate to, which is the mechanism five entries here assumed and none of them observed. The table below is that FINAL, and it is what a reading of the event itself produced after five readings of its absence had produced a confident mechanism that was not there. What that leaves standing is narrower and is still worth stating. The path serves one bulletin at a time and does not serve every trade date: Wednesday, September 16 was never served under any mark, and Thursday, September 17 was served only as a PRELIMINARY before rotating away. Friday, September 11 remains absent for the reason given above. So the strip still advances in uneven steps and the gap sentence beside it still has work to do — but it advances, and a date that has stopped changing is no longer evidence that it cannot. 🔴 The same sentence has now been superseded four times in the same way, which is itself the finding. Through the September 22 sweep this note recorded that Monday, September 21 “was a trading session and had not appeared in Section 61 when this sweep read it”; read on September 23 the path served Bulletin #181 for September 21. Through the September 23 sweep it recorded the identical thing about Tuesday, September 22; read on September 24 the path served Bulletin #182 for September 22. Through the September 24 sweep it recorded the identical thing about Wednesday, September 23; read on September 25 the path served Bulletin #183 for September 23. Through the September 25 sweep it recorded the identical thing about Thursday, September 24; read on September 26 the path serves Bulletin #184 for September 24, which is the table below. All four statements were correct about what the path served on the day they were written, and none was ever a claim that the bulletin did not exist. Four times now the missing trade date has arrived one or two days later. A trade date absent from one reading is not an absent trade date — and note that the sentence is written again below about September 25, on the same terms and with the same limit on what it claims. A reader may reasonably ask why the sentence keeps being written if it keeps being overtaken. It is written because it is the only true thing available at the moment of reading: the alternative is to assert that a bulletin exists on the strength of four prior instances, which is a forecast, and this reference does not print forecasts about its own sources. Three consequences of this reading, and none of them is a cadence. First, a bulletin may be served PRELIMINARY and finalised in place, as #180 was, or served FINAL the first time it is seen, as #181, #182, #183 and now #184 were. Both behaviours are on the record and no claim is made here about which the next one will be. Four consecutive FINALs is a run, not a rule — and the September 27 reading settled that point against the run, which is why the refusal to predict is recorded rather than the streak. Bulletin #185 for Friday, September 25 was served on that reading and was marked PRELIMINARY on every one of its ten pages, with no FINAL anywhere in the document. The run of four ended at four. Had this note offered the fifth FINAL as a cadence it would have been wrong one bulletin later. Second, #183 and #184 are consecutive bulletin numbers on consecutive trading sessions, so the move printed against this strip is a true one-session move — the fourth consecutive such step, after a long run in which the uneven multi-session step was the rule. Four in a row is not a pattern and is not offered as one. Third, and this one reaches beyond this page: the October 2026 crude contract is finished, and the strip is what shows it. October settles UNCH at $94.59 for a second consecutive session on this table, and its open interest has gone to zero — from 1,000 on the prior bulletin, which was down 15,738 in one session from 17,075, which was down 43,297 before that. A contract that does not move and that nobody holds at all has stopped being a price in every sense. Every publisher quoting a continuous front-month WTI series is now quoting November, and the Institute's own price ledger records a third-party series making that change between September 22 and September 23 without saying so — roughly $4.07 of a $5.67 printed fall was the contract changing underneath the series. On the step this table measures nothing rolled, so the moves are clean: CL NOV26 settles $94.61, up $2.45, and BZ NOV26 settles $106.60, up $3.52. The front-month Brent-WTI spread widened from $10.92 to $11.99, a second consecutive widening session after the $2.19 widening that preceded it. A strip is what lets a reader tell a roll from a move, which is the argument for carrying one. 🔴 The sentence written here about Friday, September 25 has now been overtaken, and the way it was overtaken is the part worth keeping. Through the September 26 sweep this note recorded that September 25 “was a trading session and no Section 61 bulletin for it was served on the public path at the time of this reading.” Read on September 27, the path serves Bulletin #185 for Friday, September 25. That is the fifth consecutive occasion on which a trade date absent from one reading has arrived on a later one, and it is the fifth time the narrow claim — not served when read — has held while the broader claim it is routinely mistaken for, that the bulletin does not exist, would have been wrong. Only the first of those was ever made here. The strip below does not move to it, and the reason is the standing rule rather than a judgement about this bulletin: #185 is marked PRELIMINARY on all ten of its pages and carries no FINAL anywhere, and provisional settlements are the wrong input to a curve a reader may act on. So this reference now holds a trade date it has seen and has not adopted — the same position it was in over #180, which was read PRELIMINARY on September 19 and served FINAL in place on September 22 and was adopted then. Whether #185 finalises in place, as #180 did, is not predicted here. 🔴 Read on September 29, 2026 the path serves Bulletin #186 for trade date Monday, September 28, and it is marked PRELIMINARY. It is read and not adopted, for the same standing reason #185 was. The strip below therefore remains Bulletin #184 for Thursday, September 24 — the newest FINAL this reference has seen — and it is now three trading sessions behind the newest settlement the Institute has laid eyes on. That distance is the cost of the FINAL-only rule and it is stated rather than smoothed. And the September 25 question above is answered, in the direction the sentence took care not to predict. Bulletin #185 did not finalise in place as #180 did: by this reading the /current/ path has rotated past it to #186, and every reading this reference ever took of #185 showed PRELIMINARY. So Friday, September 25 joins Friday, September 11 as a trade date permanently skipped on the public path — two instances now, against #180's one instance of finalising in place. 🔴 That has a bearing on the withdrawn mechanism above and it is a narrow one. The explanation this note carried through five sweeps — that a bulletin seen only as PRELIMINARY rotates away before it can be marked FINAL and is lost to this reference — was withdrawn on September 22 because #180 was observed finalising in place, and the withdrawal stands. What #185 establishes is that the loss described by that mechanism is a real outcome, not that the mechanism is the rule. Both endings now have instances. A description that is sometimes what happens is not an explanation of what happens, and the error in the withdrawn version was never the outcome it described but the certainty with which it described it. 🔴 One reading of #186 reaches past the strip, and it is an absence rather than a figure. The CL section of #186 opens at NOV26: the October 2026 contract is not in it at all. Bulletin #184 still printed October as a nominal $94.59 on zero open interest, which was the finding recorded above. Four sessions later the exchange has stopped printing the row, and the absence is itself the disclosure — a contract does not fade out of a settlement table, it is removed from it. Read the two bulletins together and the whole end of a contract's life is visible in four steps: open interest collapsing, a settlement that stops moving, open interest at zero with the row still printed, and then no row. 🔴 And a second Brent roll is close, which is an inference from open interest and not a reading of the contract calendar. In #186 the BZ November contract carries 21,074 lots of open interest, down 605 on the session, against 41,236 in December — December now holds nearly twice November. The two months settle $7.45 apart, backwardated. When the continuous front-month series a third-party publisher quotes changes over, roughly $7.45 of whatever fall it prints that day will be the contract changing rather than the price moving. The Institute's own price ledger recorded exactly that happening to WTI between September 22 and September 23 without the publisher saying so. This is the same event arriving for Brent, named in advance rather than reconstructed afterwards, and the reason to carry a strip at all is that it is what makes the naming possible. 🔴 Read on September 30, 2026 the path serves Bulletin #187 for trade date Tuesday, September 29. It is marked PRELIMINARY on all ten of its pages, with no FINAL anywhere in the document, and it is read and not adopted for the same standing reason #185 and #186 were. The strip below therefore remained Bulletin #184 for Thursday, September 24 — the newest FINAL this reference has seen — and on that reading it stood four sessions behind the newest settlement the Institute had laid eyes on, one session further than the entry before it recorded. That distance is the cost of the FINAL-only rule and it is stated rather than smoothed, because the alternative is a curve built on numbers the exchange reserves the right to revise. The live figure is carried in the #188 entry below rather than restated here, because a distance that is derived in one place and typed in another is a figure that can drift from itself. Three bulletins had by then been seen PRELIMINARY on first sight — #185, #186 and #187 — which was one short of the run of four FINALs-on-first-sight that preceded them and was given no more weight than that run was given. Three observations of a publisher's habit are three observations, which is the ground on which the run of four was refused the word cadence, and it applied to this run in the same terms. Nothing was predicted about #188. 🔴 The September 28 question is answered the same way September 25's was, and that makes two consecutive trade dates this reference has seen and may never carry. Bulletin #186 has rotated off the /current/ path without ever being observed FINAL — every reading this reference took of it showed PRELIMINARY — so Monday, September 28 now stands where Friday, September 25 and Friday, September 11 stand. It is worth being exact about what that does and does not establish. It does not establish that no FINAL for September 28 exists: #180 was read PRELIMINARY on September 19 and served FINAL in place on September 22, so a bulletin can still turn behind a reading. What it establishes is that the skipped outcome now has three instances against finalising-in-place's one, and that the gap between the strip and the present is being widened by the publisher's marking rather than by anything this build does. 🔴 One reading of #187 bears on the price ledger rather than on the strip, and it is the resolution of a warning written here a step early. The Brent roll named in the entry above did not occur on the September 28 to September 29 step: the third-party front-month series rose 77 cents rather than falling by the roughly $7.45 the open-interest reading implied. The condition has sharpened rather than gone away. In #187 the BZ November contract carries 19,719 lots, down a further 1,353 on the session, against 42,486 in December, up 1,253 — December now holds more than twice November. But the two months settle $6.43 apart rather than $7.45, so the mechanical step a roll would produce has narrowed by about a dollar. The warning stands with its magnitude re-measured, and it is recorded as not having fired rather than quietly carried forward, because a prediction that is only ever restated until it lands is not a prediction. 🔴 And #187 cross-confirms #186's September 28 settlements from an independent document, which is worth more than either reading alone even though neither is adoptable. #187's point-change column settles BZ NOV26 at $102.59, down $2.69, and CL NOV26 at $89.38, down $3.22; adding the changes back gives September 28 settles of $105.28 and $92.60, which is exactly what #186 printed on both benchmarks. Two bulletins, two different columns, the same two figures to the cent. That does not promote a PRELIMINARY settlement to a FINAL and it does not change what the strip carries. It does mean the provisional settle path the price ledger discloses no longer rests on a single document, and the reason to say so is that a reader who sees a provisional figure is entitled to know how many independent readings stand behind it. 🔴 Read on October 1, 2026 the path serves Bulletin #188 for trade date Wednesday, September 30. It is marked PRELIMINARY on all ten of its pages, with no FINAL anywhere in the document, and it is read and not adopted for the same standing reason #185, #186 and #187 were. The strip below therefore remained Bulletin #184 for Thursday, September 24 — still the newest FINAL this reference has seen — and on that reading it stood five trading sessions behind the newest settlement the Institute had laid eyes on, one session further than the previous entry recorded and the fifth consecutive sweep on which that distance had grown by exactly one. That distance is the cost of the FINAL-only rule and it is stated rather than smoothed. The live figure is carried in the October 2 entry below rather than restated here, following the precedent the #187 entry set: a distance that is derived in one place and typed in another is a figure that can drift from itself, and the derived figure belongs in the newest entry because that is the only one describing the present. 🔴 THE PRELIMINARY RUN NOW EQUALS THE FINAL RUN, AND THE SYMMETRY IS GIVEN NO WEIGHT AT ALL. #181, #182, #183 and #184 were each marked FINAL on first sight. #185, #186, #187 and #188 have each now been marked PRELIMINARY on first sight. Four against four. Every sweep that recorded the first run declined to call it a cadence, and that refusal is the reason nothing had to be retracted when the run ended at exactly four — so the second run gets the same refusal in the same terms, and the fact that the two runs are now the same length is arithmetic rather than evidence. Nothing is predicted about #189. 🔴 The September 29 question is answered the way September 25's and September 28's were, and the skipped-trade-date count reaches four. Bulletin #187 has rotated off the /current/ path without ever being observed FINAL — every reading this reference took of it showed PRELIMINARY — so Tuesday, September 29 now stands where Monday, September 28, Friday, September 25 and Friday, September 11 stand. What that does and does not establish has not changed: it does not establish that no FINAL for September 29 exists, because #180 was read PRELIMINARY on September 19 and served FINAL in place on September 22, so a bulletin can still turn behind a reading. What it establishes is that the skipped outcome now has FOUR instances against finalising-in-place's one. Those are the two counts a reader should weigh, and they are stated as counts rather than converted into a probability, because one observation of the other outcome is enough to show it remains available. 🔴 The Brent roll warning has still not fired, and this is the second consecutive entry recording that rather than quietly carrying it forward. The September 29 to September 30 step the third-party front-month series printed was a $3.46 fall, against the roughly $6.43 a roll would have produced on #187's spread; #188 still opens its BZ block at NOV26, so no roll occurred. The condition continues to sharpen on position and to soften on magnitude. In #188 the BZ November contract carries 19,421 lots, down a further 298 on the session, against 43,242 in December, up 756 — December now holds more than twice November and the ratio has widened again. But the two months settle $5.50 apart rather than $6.43, so across #186, #187 and #188 the mechanical step a roll would produce has run $7.45, $6.43, $5.50. Three readings in one direction are not a trend, which is the refusal the SPR row on this same release was forced into after two prints were read as a rate; and the warning is not withdrawn because the number it names has got smaller, because a smaller unexplained step is still an unexplained step. 🔴 And #188 cross-confirms #187's September 29 settlements from an independent document, for the second consecutive sweep on which that check has been available. #188's point-change column settles BZ NOV26 at $103.53, up $0.94, and CL NOV26 at $90.42, up $1.04; subtracting the changes gives September 29 settles of $102.59 and $89.38, which is exactly what #187 printed on both benchmarks. Two bulletins, two different columns, the same two figures to the cent. That does not promote a PRELIMINARY settlement to a FINAL and it does not change what the strip carries. What it does is tell a reader looking at the provisional path in the price ledger how many independent readings stand behind each of its figures, which is the only thing a cross-confirmation can honestly buy. Note also that the check has now worked twice in a row on consecutive sessions, and that this says something about the archive's transcription and nothing whatever about whether the exchange will revise the numbers. 🔴 ON THE OCTOBER 2, 2026 SWEEP THE SECTION 61 PATH WAS NOT RETRIEVABLE, AND THAT IS A FAILURE ON THIS SIDE RATHER THAN A MARK ON THE PUBLISHER'S. Every entry above records what a reading of the public path served. This one records that no reading was obtained: the fetch path available to this sweep refused the bulletin URL, so nothing was served to be read, adopted or declined. The distinction matters because the two outcomes are easy to collapse and only one of them is this reference's to report. Nothing is claimed here about whether a bulletin for Thursday, October 1 exists, what number it carries, or whether it is marked FINAL or PRELIMINARY. The PRELIMINARY-on-first-sight run therefore stands where #188 left it, at four against the earlier four FINALs, and the skipped-trade-date count stands at four against finalising-in-place's one. An unexamined path and an unchanged path are indistinguishable in a table, which is why this sentence exists at all. The strip below consequently remains Bulletin #184 for Thursday, September 24 — still the newest FINAL this reference has seen, and on that sweep not merely unadopted-newer but unread-newer. The live distance is carried in the October 3 entry below rather than restated here, following the precedent the #187 and #188 entries set: the derived figure belongs in the newest entry because that is the only one describing the present, and a distance derived in one place and typed in another is a figure that can drift from itself. Note what the figure does and does not measure, because this sweep is the one where the difference bites: it is the distance to the newest settlement READ, which on the October 2 sweep had not moved since #188 while the publication clock had advanced a session. The distance grew by one for the sixth consecutive sweep, and for the first time the reason was this reference's fetch path rather than the publisher's marking. 🔴 AND THE BRENT ROLL NAMED IN ADVANCE ON THREE CONSECUTIVE ENTRIES HAS FIRED, AT THE MAGNITUDE THE LAST OF THEM MEASURED. The third-party front-month series this reference quotes printed a $4.69 fall from its September 30 Brent open of $102.58 to its October 1 open of $97.89. Measured against #188's settlements the step decomposes: $97.89 sits 14 cents below BZ DEC26's $98.03 and $5.64 below BZ NOV26's $103.53, so the series is now quoting December, and the roll step is the $5.50 the two months settled apart. A printed fall of $4.69 is therefore −$5.50 of contract change and +$0.81 of price movement UPWARD. The barrel rose on a session the series printed a fall of four and a half per cent, and no publisher said so. The strip is what made both halves of that possible and it is the argument for carrying one. The naming came first — #186, #187 and #188 each recorded the condition from open interest and each stated the magnitude a roll would mechanically produce — and the identification came from the same documents, because the only way to tell a roll from a move is to have the two contract months' settlements in hand. A reader without a strip sees a four-and-a-half-per-cent fall and has no way to learn that the barrel went up. 🔴 SCORED HONESTLY, BECAUSE A PRE-REGISTRATION THAT IS ONLY SCORED WHEN IT WINS IS NOT ONE. The condition was right and the timing was wrong: the warning was first written on September 29 and stood unfired across four subsequent steps, and each of those steps was recorded as a non-firing rather than carried forward silently. Open interest carried direction, not timing, and that is the general lesson rather than a detail of this instance. The magnitude is right only because it was re-measured on every sweep: the figure first named was $7.45, then $6.43, then $5.50, and the one that landed is the last. Had the original number been carried forward it would have been wrong by $1.95 — about forty per cent of the step that actually occurred — while looking exactly as confident. 🔴 AND WTI DID NOT ROLL ON THE SAME STEP, WHICH IS PART OF THE EVIDENCE RATHER THAN AN ASIDE. The same series' WTI quote rose $1.39 to $90.31, which sits 11 cents below CL NOV26's $90.42 and $1.89 ABOVE CL DEC26's $88.42 — still November. The same test run on two benchmarks on the same step returns a roll on one and a clean non-roll on the other, and a test that only ever fires has demonstrated nothing. The two residual price moves agree in sign and to 58 cents, which is looser than the twenty cents the September check on the WTI roll achieved; the slack is stated rather than smoothed. One further consequence reaches the price ledger and is recorded there: the two quotes are now different contract months, so the Brent-WTI spread cannot be computed by differencing them. 🔴 THE IDENTIFICATION RESTS ON A PRELIMINARY DOCUMENT, AND THAT IS SAID RATHER THAN ELIDED. #188 is marked PRELIMINARY on all ten pages and is not adopted into the strip below; the exchange reserves the right to revise its settlements. Those are the right terms on which to refuse a provisional figure as the input to a curve a reader may act on, and they are not a reason to refuse it as the discriminator between two candidate contract months. The question asked of #188 was which month a quote belongs to, and the answer turned on a 14-cent match against one candidate and a $5.64 gap against the other — a margin no plausible settlement revision closes. A document may be good enough to answer one question and not another, and the reason this reference reads bulletins it will not adopt is that the two uses are different. 🔴 READ ON OCTOBER 3, 2026 THE PATH IS RETRIEVABLE AGAIN AND SERVES BULLETIN #190 FOR TRADE DATE FRIDAY, OCTOBER 2. It is marked PRELIMINARY on all ten of its pages, with no FINAL anywhere in the document, and it is read and not adopted for the same standing reason #185, #186, #187 and #188 were. The strip below therefore remains Bulletin #184 for Thursday, September 24 — still the newest FINAL this reference has seen — and it is six trading sessions behind the newest settlement the Institute has laid eyes on. That figure is derived from the archived bulletin's own trade date against this page's publication date rather than typed, so it cannot drift from the strip it describes. Note that it did NOT grow on this sweep, for the first time in seven, and the reason is the calendar rather than the publisher: October 3 is a Saturday, the exchange published no bulletin for it, and the convention stated at the head of this module counts trading days only. A distance that stops growing because the week ended is not a distance that has started closing. 🔴 AND THE PREVIOUS ENTRY'S FETCH FAILURE WAS TRANSIENT, WHICH IS WHY NOTHING HAS TO BE RETRACTED. The October 2 entry above recorded that no reading was obtained and said in terms that nothing was being claimed about whether a bulletin existed, what number it carried, or how it was marked. The path answers normally on this sweep, so the failure was on this side and was temporary. The entry is retained exactly as written because the restraint in it is the instructive part: had it reported the path as broken, or inferred an absent bulletin from an absent reading, it would now be wrong on both counts. An unexamined path and an unchanged path are indistinguishable in a table, and the only defence against collapsing them is to say which one was observed. 🔴 BULLETIN #189 FOR THURSDAY, OCTOBER 1 WAS NEVER READ AT ALL, AND THAT IS A THIRD CATEGORY RATHER THAN A VARIANT OF THE OTHER TWO. This reference has until now recorded two kinds of trade date it does not carry: one read and not adopted because it was marked PRELIMINARY, and one read PRELIMINARY and then rotated off the path before any FINAL was observed. October 1 is in neither. The October 2 sweep could not reach the path, the October 3 sweep finds it rotated on to #190, and so no reading of #189 was ever taken. Nothing is claimed here about whether #189 exists, what number it in fact carries, or whether it was marked FINAL or PRELIMINARY at any moment. An unread document is not a PRELIMINARY one, and the distinction is load-bearing twice over: it keeps October 1 out of the PRELIMINARY tally below, and it keeps this reference from reporting its own missed reading as a fact about the exchange. 🔴 SO THE PRELIMINARY RUN NOW EXCEEDS THE FINAL RUN, FIVE TO FOUR, AND THE RUN HAS A HOLE IN IT. #181, #182, #183 and #184 were each marked FINAL on first sight. #185, #186, #187, #188 and now #190 have each been marked PRELIMINARY on first sight. That is five observations against four, and it is not five consecutive bulletins: #189 sits between #188 and #190 unread, and is neither evidence for the run nor against it. The refusal every previous entry made is made again in the same terms — four consecutive FINALs were declined the word cadence, which is the reason nothing had to be retracted when that run ended at exactly four, and a five-observation run with a gap in it earns less than the four did rather than more. Nothing is predicted about the next bulletin. 🔴 THE SKIPPED-TRADE-DATE COUNT REACHES FIVE, PLUS A SIXTH OF UNKNOWN STATUS. Bulletin #188 has rotated off the /current/ path without ever being observed FINAL — every reading this reference took of it showed PRELIMINARY — so Wednesday, September 30 now stands where Tuesday, September 29, Monday, September 28, Friday, September 25 and Friday, September 11 stand. Five instances of the skipped outcome against #180's one instance of finalising in place. What that does and does not establish has not changed: it does not establish that no FINAL for September 30 exists, because a bulletin can still turn behind a reading, and one observation of the other outcome is enough to show it remains available. October 1 is counted separately and not added to either column, because a trade date this reference never read is not a trade date it watched the publisher skip. 🔴 AND NEITHER BENCHMARK ROLLED ON THIS STEP, WHICH MAKES BOTH PRINTED MOVES PRICE MOVES FOR THE FIRST TIME IN SEVERAL SWEEPS. #190's point-change column reconstructs the October 1 settlements it is measured against: CL NOV26 at $91.11 down $1.76 gives $92.87, CL DEC26 at $89.43 down $1.41 gives $90.84, and BZ DEC26 at $102.25 down $0.06 gives $102.31. The third-party series' October 2 WTI quote of $92.83 sits 4 cents below CL NOV26 against $1.99 above CL DEC26 — still November — and its Brent quote of $102.23 sits 8 cents below BZ DEC26 against $1.28 below a BZ NOV26 mark carrying no volume and no open interest at all — still December. So the printed steps are genuine: Brent up $4.34 and WTI up $2.52. The same test on two benchmarks returns a clean non-roll on both, and a test that only ever fires has demonstrated nothing, so the clean result is recorded with the same prominence the roll was. 🔴 BUT THE MECHANISM THREE ENTRIES HERE RELIED ON WAS BACKWARDS, AND THAT IS A CORRECTION TO THE REASONING RATHER THAN A VINDICATION OF THE CONCLUSION. #186, #187 and #188 each read the Brent November contract's declining open interest as the leading indicator of the coming roll, and each stated the magnitude a roll would mechanically produce. The roll fired on the September 30 to October 1 step, at the magnitude the last of them measured. In #190 the BZ November contract's open interest is ZERO, down 19,421 on the session — the whole of the position the September 30 bulletin recorded, drained in one session, and attributed by the exchange's own change column to October 2. Add the change back and October 1 closed at 19,421 lots, which is exactly what #188 printed for September 30 — so the November contract's open interest did not move by a single lot across the unread session. November therefore held every lot it had through October 1, the session on which the publisher's series had already moved to December. The roll PRECEDED the drain by a session. Open interest was the lagging indicator, not the leading one. The condition those three entries named was real and the direction was right; the sequence they assumed was the reverse of the sequence that occurred, and it is set out in those terms because a prediction that lands for the wrong reason is still a wrong model. 🔴 THE END OF A CONTRACT'S LIFE NOW HAS TWO SPEEDS ON THE RECORD AND ONE TERMINUS. The October 2026 WTI contract took the slow path, visible across four bulletins on this page: open interest down 43,297, then 17,075, then 1,000, then zero with the row still printed at a nominal $94.59, and then no row at all. The November 2026 Brent contract has taken the fast one: 19,421 lots to zero in a single session, with its row still printed at a nominal $103.51 marked UNCH on no volume. A third instance sits in the same document and is recorded for the same reason — HO OCT26, heating oil for October, prints 4.9569 UNCH with open interest at zero, down 42 lots, two bulletins after its drain was first noticed. Three contracts, three rates of decline, the same four-stage ending. An earlier entry here declined to put a date on when a drain would complete, and that refusal is the reason nothing has to be retracted now. 🔴 ONE CONSEQUENCE REACHES THE PRICE LEDGER AND A SECOND REMOVES A FIGURE FROM IT ALTOGETHER. The two quotes are once again on the same contract month in neither case — WTI on November and Brent on December — so the Brent-WTI spread still cannot be computed by differencing them, and the $9.40 that differencing produces is not a spread. The exchange's own same-month December spread is $12.82 on these October 2 settlements against $9.61 on the September 30 settlements, a widening of $3.21 in two sessions and an almost entirely one-sided one: BZ DEC26 rose $4.22 across those sessions against CL DEC26's $1.01. And the November spread has stopped being quotable at all, in either direction, because its Brent leg is now a nominal mark on zero open interest. A spread between a traded price and a nominal one is not a spread either, and this reference will not print one. 🔴 THE OCTOBER 1 SETTLEMENTS ARE DERIVED AND UNCORROBORATED, WHICH IS A WEAKER FOOTING THAN THE LAST TWO SWEEPS ENJOYED AND IS SAID RATHER THAN ELIDED. On the #187 and #188 entries above, each bulletin's point-change column reproduced the previous bulletin's printed settlements to the cent, so the provisional path the price ledger discloses rested on two independent readings at each step. That check is unavailable here. #189 was never read, so the October 1 figures quoted above exist only as the arithmetic of #190's own change column — one document, and a derivation rather than a printed number. They are good enough to answer which contract month a quote belongs to, where the margins are 4 cents against $1.99 and 8 cents against $1.28 and no plausible revision closes them. They are not good enough to carry into the strip, and they are not offered as settlements the exchange printed. A document may be good enough to answer one question and not another, which is the whole reason this reference reads bulletins it will not adopt. 🔴 READ ON OCTOBER 7, 2026 THE PATH SERVES BULLETIN #191 FOR TRADE DATE MONDAY, OCTOBER 5, AND IT IS MARKED FINAL ON ALL TEN OF ITS PAGES, WITH PRELIMINARY ON NONE. IT IS ADOPTED, AND THE TABLE BELOW IS NOW THAT DOCUMENT. This is the first FINAL this reference has seen since #184 for Thursday, September 24, and the strip therefore advances SEVEN TRADING SESSIONS in a single step — September 25, 28, 29, 30, October 1, 2 and 5 — which is the widest step it has ever taken. The step is not a catch-up on all seven of those dates: it carries October 5's settlements and skips the other six entirely. Five of them were read and not adopted because every reading showed PRELIMINARY (#185, #186, #187, #188, #190) and one was never read in any state (#189, for Thursday October 1). Nothing here claims that no FINAL exists for those six trade dates: #180 was read PRELIMINARY on September 19 and served FINAL in place on September 22, and that one instance is enough to keep the outcome available. The claim is the narrow one this reference has made every time — not in the release we read. The distance from this strip to the present is six trading sessions, derived from the archived bulletin's own trade date against this page's publication date rather than typed. 🔴 AND FOR THE FIRST TIME THE COST OF THE FINAL-ONLY RULE CAN BE PRICED RATHER THAN ONLY ASSERTED, BECAUSE A FINAL AND THE PRELIMINARY IT SUPERSEDES ARE BOTH IN HAND. Every entry above states that provisional settlements are the wrong input to a curve a reader may act on, and none of them could say what the exchange's PRELIMINARY mark actually revises. #190 for Friday October 2 and #191 for Monday October 5 share 87 contract months. Backing #191's point-change column out against #190's printed settlements reconstructs the October 2 settles on all 87, and they RECONCILE 87 OF 87 TO THE CENT. Open interest reconciles 79 of 87, with eight figures revised: CL NOV26 258,420 against 258,251 implied, a residual of 169 lots; CL DEC26 236,448 against 236,323, 125 lots; CL FEB27 ten lots; BZ APR27 four; BZ MAR27, BZ MAY27 and CL DEC30 two each; and CL MAR27 one lot the other way. The reading in both directions, because only one of them flatters the rule. The rule is vindicated as a principle: eight open-interest figures a reader could have taken off a provisional document were wrong, and the exchange's reservation of the right to revise is a real reservation and not boilerplate. But the magnitudes are small — the largest revision is 169 lots on 258,420, or 0.065 per cent, the median of the eight is two lots, and the column this strip actually consumes is settlement, which did not move on a single row of 87. So seven sessions of staleness bought the correction of eight open-interest figures and zero prices. That is ONE OBSERVATION of one PRELIMINARY against one FINAL, on one pair of adjacent trade dates, and it is not offered as a rule about what a provisional settlement is worth — this note has refused the word cadence to a run of four and to a run of five, and a single measurement earns less than either. It is recorded because the cost of a discipline should be measured where it can be rather than described where it cannot. 🔴 THE PRELIMINARY RUN ENDS AT FIVE OBSERVATIONS WITH A HOLE IN IT, AND NOTHING HAS TO BE RETRACTED. #185, #186, #187, #188 and #190 were each marked PRELIMINARY on first sight, against #181 through #184's four FINALs, with #189 unread between the last two. Every entry recording that run declined to call it a cadence and declined to predict the next bulletin, and the refusal is why the run ending in a FINAL costs this note nothing. Had the five been offered as a pattern it would have been wrong one bulletin later, exactly as the four would have been. 🔴 AND THE FOUR-STAGE CONTRACT ENDING THIS NOTE DESCRIBED AS COMMON TO THREE CONTRACTS IS CORRECTED: THE STAGE COUNT IS NOT THE SHARED PROPERTY. The entry above reads “Three contracts, three rates of decline, the same four-stage ending.” Traced across the archive set, BZ NOV26's ending took TWO printed stages, not four — a traded row on September 30 at 19,421 lots, a nominal $103.51 marked UNCH on zero open interest in #190 for October 2, and no row at all in #191 for October 5. CL OCT26 took four — 17,075 lots on September 21, 1,337 on September 22, a nominal $94.59 UNCH on 1,000 lots on September 23, the same nominal on zero on September 24, and no row on September 28. What the two share is the TERMINUS and its ORDER — a traded row, then a nominal mark on a draining position, then no row — and not the number of printed steps it takes. A reader who had taken the four-stage framing as a template would have expected two more bulletins of Brent November than the exchange printed. The correction is stated here rather than applied silently to the sentence above, because the sentence was written from two contracts and generalised from them, and the generalisation is the error worth keeping on the record. 🔴 NEITHER BENCHMARK ROLLED ACROSS THIS STEP, AND THE CLEAN RESULT IS RECORDED WITH A ROLL'S PROMINENCE BECAUSE A TEST THAT ONLY EVER FIRES HAS DEMONSTRATED NOTHING. The third-party series' October 6 WTI open of $89.28 sits 15 cents BELOW CL NOV26's $89.43 and $1.08 ABOVE CL DEC26's $88.20 — still November. Its Brent open of $100.25 sits 7 cents BELOW BZ DEC26's $100.32 and $2.67 ABOVE BZ JAN27's $97.58 — still December. Note which way the clocks run on that comparison: the quotes are one session AHEAD of the bulletin, so the test is an open against the immediately preceding settle, which is the same test every roll call on this page has used. 🔴 THE SAME-MONTH DECEMBER BRENT–WTI SPREAD IS $12.12 ON THESE SETTLEMENTS, AND THE WINDOW A READER COMPARES IT AGAINST CHANGES WHAT IT SAYS, SO BOTH ARE PRINTED. Against the strip this table replaces — #184 for September 24, where BZ DEC26 settled $100.22 and CL DEC26 $90.90 — the spread was $9.32, so it has widened $2.80 across seven sessions, and the decomposition is entirely one-sided: Brent's December leg rose ten cents while WTI's fell $2.70. Against #190 for October 2, where it stood at $12.82, the last single session NARROWED it 70 cents. Those are not in conflict; they are a seven-session move and the last session of it, and a note that printed only the first would leave a reader with the direction of the most recent session backwards. The NOVEMBER spread cannot be computed on this strip in either direction, because Brent has no November leg left to difference. 🔴 ONE STRUCTURAL CURIOSITY IN #191, RECORDED WITH ITS PRECEDENT RATHER THAN AS A FIRST. BZ SEP28 — September 2028 Brent — prints a settlement of $77.17 marked NEW, which is the exchange's mark for a contract with no prior settlement to change from, on 198 lots of Globex volume and 198 lots of open interest, up 198. The parser carries a NEW row's point change as absent rather than as zero, because a contract that has never settled before has not failed to move. This is the SECOND such row this reference has observed, not the first: CL JUN35 appeared the same way in the September 15 extract, at $52.65 marked NEW on two lots of volume. The two differ in what they left behind — JUN35 ended its first session with zero open interest, SEP28 retained all 198 — and neither is given any weight beyond being on the record. The archived extract reproduces the row's total-volume dash against its Globex 198 verbatim and does not explain it. 🔴 READ ON OCTOBER 10, 2026 TWO FURTHER BULLETINS WERE OBTAINED AND NEITHER IS ADOPTED, AND FOR THE FIRST TIME ONE OF THEM IS DECLINED ON THE STATE OF THE INSTITUTE'S OWN ARCHIVE RATHER THAN ON THE STATE OF THE DOCUMENT. The table below stays on #191 for Monday, October 5. Bulletin #195 for Friday, October 9 is marked PRELIMINARY on all ten of its pages with FINAL on none, and the standing rule disposes of it as it has disposed of every provisional sheet: read, archived, not adopted. Bulletin #193 for Wednesday, October 7 is the harder case and the instructive one. It is marked FINAL on all ten of its pages, it is the first FINAL this reference has seen since #191, and on the state of the document it is adoptable. IT IS NOT ADOPTED, BECAUSE THE EXTRACT THE INSTITUTE HOLDS OF IT IS INCOMPLETE. That extract carries 41 of the WTI strip's contract months against the 62 that #191 and #195 each print; it is a strict PREFIX of both, ending at MAR30 and missing APR30 through DEC36; and the exchange's own TOTAL CL FUT open-interest line is absent from it altogether. The Brent leg is complete at 26 months with its total present. The /current/ path rotated days ago, so the missing months cannot be re-obtained. A document this reference could have adopted has been lost to the way it was captured, and that is a different failure from every decline recorded above. 🔴 WHAT CAUGHT IT IS THE GATE THAT WOULD HAVE REFUSED THE BUILD, AND WHAT MAKES IT DETECTABLE AT ALL IS THE ABSENCE OF THE VERY LINE WHOSE ABSENCE IS THE DEFECT. The strip on this page is parsed from the archived extract rather than transcribed, and the parser reconciles the open interest it recovers against the exchange's printed total for each contract, refusing to publish when the two disagree. Repointing the strip at this extract would have raised that refusal rather than rendering a short curve, because the total is missing and the parser requires it. Had the capture instead carried a total consistent with its own 41 rows, the reconciliation would have passed and this reference would have published a WTI curve twenty-one contract months short with nothing in the output looking wrong. The 41 rows are correctly formatted, the Brent block beside them is perfect, the extract ends at a clean row boundary and carries its own tail marker, and none of that is evidence of completeness: a capture that stops where a bulletin page ends looks exactly like a capture that stopped where the data did. 🔴 SO THE ADOPTION STATES ARE FOUR AND NOT THREE, AND THE FOURTH IS NOT A VARIANT OF THE OTHERS. This note has carried three: adopted after a FINAL reading; read, archived and declined as PRELIMINARY; and never read at all, because /current/ rotated before a sweep reached it. #193 is in none of them. It was read, it was FINAL, it is archived, and it is unadoptable — and the cause sits in the Institute's capture rather than in the exchange's document or in the sweep's timing. The claim made about it is the narrow one this reference makes about everything: not that no complete FINAL exists for Wednesday, October 7, but that the Institute does not hold one. Across October 5 to 9 the position is now a complete FINAL for the 5th, a truncated FINAL for the 7th, a complete PRELIMINARY for the 9th, and nothing whatever for the 6th or the 8th. 🔴 AND THE DISTANCE THIS STRIP SITS BEHIND IS UNCHANGED BY ANY OF IT, WHICH IS THE DERIVED FIGURE EARNING ITS KEEP. The gap sentence beside this table is computed from the archived bulletin's own trade date against this page's publication date, so it reports the distance without anybody deciding what it ought to say. A typed figure would have had to be revised twice inside this one sweep — once when #193 looked adoptable and once when it proved not to be — and the second revision is the one that would have been forgotten. No forecast is offered about what the next sweep will find. A prior version of this note ended with one — that the September 11 FINAL would be picked up next time — and it was wrong. A claim about a future sweep carries a clock of its own and expires like any other. 🔴 READ ON OCTOBER 16, 2026 THE PATH SERVES BULLETIN #198 FOR TRADE DATE WEDNESDAY, OCTOBER 14, 2026, MARKED FINAL ON EVERY PAGE OF THE CAPTURE, AND IT IS ADOPTED. It replaces #191 for Monday, October 5, which this table carried for nine days. The strip is complete — sixty-five WTI rows against the bulletin's own printed total of 777,379 contracts traded and 1,897,585 open, twenty-six Brent rows against 135,205 and 226,161 — and the open-interest reconciliation the parser performs at import passes on both legs. That reconciliation is a condition this reference places on publishing the strip at all, not a courtesy check: a capture that parses and does not reconcile is withheld. 🔴 AND THIS IS THE FIRST STEP WIDER THAN ONE SESSION THAT IS CLEAN ON BOTH LEGS. CL NOV26 is the front contract in #191 and in #198 alike, 89.43 to 95.09, up $5.66; BZ DEC26 is the front contract in both, 100.32 to 106.06, up $5.74. Neither printed move carries a contract change, so the front-to-front Brent–WTI spread — 10.89 to 10.97, widened eight cents — is like-for-like on each of its own legs. The pair this supersedes was not, and the contrast is the point: BZ NOV26 was the front month in #184 and had no row at all in #191, so that pair's printed $6.28 Brent fall was almost entirely roll. The caveat on this page exists because of pairs like that one, and it is not retired by a pair like this one. 🔴 THE ADOPTED SETTLEMENTS CROSS-CONFIRM AGAINST AN INDEPENDENT DOCUMENT, WHICH IS A CHECK THAT DOES NOT DEPEND ON THIS PAGE'S ARITHMETIC. #197, for Tuesday October 13, settled CL NOV26 at 94.14 and BZ DEC26 at 105.69. #198 reports daily changes of +0.95 and +0.37. 94.14 plus 0.95 is 95.09, and 105.69 plus 0.37 is 106.06 — the adopted figures exactly, on both legs, from the exchange's own change column rather than from a difference this page computed. A settlement that reconciles that way rests on a firmer footing than one reading of one bulletin, and it is the footing the derived October 1 settlements recorded above never had. 🔴 SO THE ADOPTION STATES ARE FIVE AND NOT FOUR, AND THE FIFTH COSTS NOTHING, WHICH IS PRECISELY WHY IT IS NAMED. The four already on this note are: adopted after a FINAL reading; read and declined as PRELIMINARY; never read at all because the path had rotated; and read, FINAL, archived and unadoptable because the capture was truncated. The fifth is READ, FINAL, COMPLETE AND OVERTAKEN. #196 for Monday October 12 and #197 for Tuesday October 13 were each read FINAL and complete, and each would have been adoptable on its own merits; neither was adopted, because a newer FINAL arrived before this page was built. That is not a declination and must not be recorded as one. A state that cost nothing is the state most likely to be mistaken later for one that cost something, and the four-state list above was already an instance of that. 🔴 AND ONE BULLETIN CHANGED STATE BETWEEN READINGS, WHICH RETIRES A READING OF THE PRELIMINARY MARK THAT THIS NOTE HAS CARRIED SINCE SEPTEMBER. #195 for Friday, October 9 was read three times: PRELIMINARY on October 9, PRELIMINARY again on October 11, and FINAL on October 13 — at the same URL, under the same bulletin number, for the same trade date. A BULLETIN CAN FINALISE IN PLACE. So PRELIMINARY is a property of a document at a moment and not of the document, and the only thing a PRELIMINARY reading licenses is a later re-read. Five of this note's earlier entries reasoned about whether a given bulletin would finalise; none of them had grounds to say it could do so without changing its number, and one of them is now shown to have been asking the wrong question. 🔴 ONE FIELD IN THE ADOPTED EXTRACT IS LEFT OPEN RATHER THAN GUESSED. Section 61's rows carry a third numeric column after settlement and change whose heading was not legible in the capture. Nothing this page prints consults it, so the archived extract records it as unlabelled rather than naming it. A field that is not relied on does not need a name; it needs an honest statement that it has none, which is cheaper than a plausible guess a later reader would inherit as a fact. 🔴 READ ON OCTOBER 17, 2026 THE PATH SERVES BULLETIN #199 FOR TRADE DATE THURSDAY, OCTOBER 15, 2026, MARKED FINAL ON ALL TEN PAGES OF THE CAPTURE AND PRELIMINARY ON NONE, AND IT IS ADOPTED. It replaces #198 for Wednesday, October 14, which this table carried for one day. The strip is complete and terminated by the exchange's own TOTAL line on both legs — sixty-five WTI rows against printed totals of 898,585 contracts traded and 1,880,704 open, twenty-six Brent rows against 187,083 and 225,071 — and all four total columns on each leg reconcile exactly, including the signed open-interest change column at minus 16,881 on WTI and minus 1,090 on Brent. The reconciliation the parser performs at import is a condition this reference places on publishing the strip at all, not a courtesy check: a capture that parses and does not reconcile is withheld. 🔴 AND THE STEP IS ONE TRADING SESSION AND ONE CALENDAR DAY, WHICH IS THE NARROWEST THIS REFERENCE CAN CARRY AND ITS FIRST SINGLE-SESSION ADVANCE SINCE SEPTEMBER 24. #198 and #199 are consecutive bulletin numbers on consecutive trading sessions with nothing between them, not even a weekend, so every figure in the comparison beside this table is a true daily change rather than a multi-session aggregate reported as one step. The advance it replaces was seven sessions wide. Neither leg rolled — CL NOV26 and BZ DEC26 are the front contracts in both documents — and the month roster is identical across the pair, so nothing in the move is a composition effect. The full chain predicate holds on 91 of 91 rows on both the settlement and the open-interest chain with zero residuals, which is the second consecutive FINAL-to-FINAL pair this reference has been able to say that about. The predicate is live rather than vacuous: run against the non-consecutive #197 instead of #198 it returns 90 settlement residuals and 56 open-interest residuals, so a check that passes here is a check that was capable of failing. 🔴 THE BRENT–WTI SPREAD COMPRESSED $3.02 IN ONE SESSION AND TWO INDEPENDENT CONSTRUCTIONS AGREE, WHICH IS WHY IT IS REPORTED AS A MOVE RATHER THAN AS AN ARTIFACT. Front-to-front, BZ DEC26 less CL NOV26 went 10.97 to 7.95. Those are different calendar months and the standing caveat on this page applies to the figure. The like-for-like December pair, BZ DEC26 less CL DEC26, went 12.16 to 9.42 — a narrowing of $2.74. The two constructions fail in different ways and moved the same direction by a similar amount. For scale, the same December series moved four cents across the entire preceding seven-session step, 12.12 to 12.16. WTI did the moving: CL NOV26 settled 95.09 to 97.97, up $2.88 or 3.03%, while BZ DEC26 settled 106.06 to 105.92, down fourteen cents. No cause for the compression is offered. The Institute reports the exchange's marks and has no mechanism to put beside them, and a sweep that supplied one would be supplying it from nowhere. 🔴 THE MARK ON THIS BULLETIN RESTS ON ONE RETRIEVAL, NOT TWO, AND THAT IS WEAKER THAN WHAT ITS PREDECESSOR CARRIED. #198 was read FINAL on two separate retrievals. #199 was read once, on October 17, and was FINAL on every page of that reading. The strength of the predecessor's reading is not inherited by its successor, and the distinction is recorded because a reader comparing the two entries would otherwise have no way to see it. This is the seventh bulletin this reference has found FINAL on first sight, and seven instances are not offered as a cadence — the reading that Section 61 trails the product index by a business day was withdrawn earlier in this note for exactly that reason. 🔴 THE WTI ROLL IS WIDENING AND THE NEXT SWEEP SHOULD EXPECT THE FRONT MONTH TO CHANGE. CL NOV26 open interest fell 113,474 to 84,922, down 25.2% in one session, while DEC26 rose 302,291 to 309,217. DEC26 now carries 224,295 lots more than the front month, a ratio of 3.64 times against 2.66 the session before. Brent is moving the same way more slowly, DEC26 37,295 to 34,132 against JAN27 29,560 to 28,735. A front-month change is a contract change, and the counterexample retained on this page is there because a printed front-to-front move across one can be almost entirely roll. 🔴 SUPERSEDED BY THE OCTOBER 18 READING, RETAINED AND LABELLED RATHER THAN DELETED, BECAUSE WHAT ONE READING FOUND IS NOT UNMADE BY WHAT A LATER ONE FINDS. On October 17 this path served nothing for Friday, October 16, and what this page printed then read: “Bulletin #200 is not in the release this reference read. Whether it has been posted elsewhere, or behind DataMine, or at this path at some other hour, is NOT OBSERVED — not in the release we read is a different claim from not published, and only the first is ours to make.” That was an accurate statement about a reading, and it is no longer the position, because the October 18 sweep read bulletin #200. What that bulletin carried is recorded in the October 18 entry at the foot of this note. The strip still sits one trading session behind the newest settlement the exchange has published, and the gap sentence beside it computes that distance rather than asserting it. 🔴 THE THIRD NUMERIC COLUMN IS STILL UNLABELLED, AND IS STILL LEFT THAT WAY. The legend page was not returned on this reading either. Nothing this page prints consults the column, so the archived extract continues to record it as unlabelled rather than naming it. A field that is not relied on needs an honest statement that it has no name, which remains cheaper than a plausible guess a later reader would inherit as a fact. 🔴 OCTOBER 18, 2026 — BULLETIN #200 WAS READ, IS MARKED PRELIMINARY, AND IS THEREFORE NOT ADOPTED. The /current/ path served bulletin #200 for trade date Friday, October 16, 2026, and all ten of its page headers are stamped PRELIMINARY; none is stamped FINAL. The curve on this reference carries FINALs only, for the reason it has carried since bulletin #185: the exchange revises provisional settlements, and a provisional settlement is the wrong input to a curve a reader may act on. So the strip below remains bulletin #199, FINAL, for Thursday, October 15, and nothing in #200 has moved a figure anywhere on this reference. The reading was kept anyway, because it is evidence about the publisher and about the contracts even where it is not evidence this page will print. The capture is complete on both legs — 64 WTI months and 26 Brent months, each terminated by the exchange's own printed TOTAL line rather than by a judgement that the extract looked long enough — and all four numeric columns reconcile against those printed totals exactly: volume 913,100 and open interest 1,871,039 on WTI against 159,077 and 229,442 on Brent, with the signed open-interest change summing to minus 9,665 and plus 4,371 as printed. The extract is archived at _ops/cme_daily_bulletin_s61_20261016.txt. 🔴 WHAT THE READING ESTABLISHES WITHOUT BEING ADOPTED, AND THE FIGURES ARE PROVISIONAL. The provisional November WTI settlement is 3.44 above, and the provisional December Brent settlement 3.94 above, the FINALs this page carries. Those two moves close the arithmetic exactly — 101.41 less 3.44 is 97.97, and 109.86 less 3.94 is 105.92, which are bulletin #199's settles — so no trade date is missing between the two bulletins, which is the thing a reader of a curve most needs to know and is checkable without adopting either number. Both figures are provisional, are named as such here, and appear in no table on this reference. 🔴 AND A MONTH LEAVING THE SECTION IS NOT A CAPTURE DEFECT. CL JUN35 is printed in bulletin #199 and is absent from #200. #199 carried it at 51.78 on zero volume and zero open interest, and the exchange prints only the months carrying one or the other — the same behaviour this page already records for CL OCT26 between bulletins #184 and #186. A row that stops being printed because nothing traded in it is the publisher's convention and not a short read, and the completeness gate above is what makes those two cases distinguishable from one another. 🔴 OCTOBER 20, 2026 — BULLETIN #201 WAS READ, IS MARKED PRELIMINARY ON TEN OF TEN PAGES, AND IS THEREFORE NOT ADOPTED. The /current/ path served bulletin #201 for trade date Monday, October 19, 2026. Every one of its ten page headers is stamped PRELIMINARY and none is stamped FINAL, so the strip below remains bulletin #199, FINAL, for Thursday, October 15, and nothing in #201 has moved a figure anywhere on this reference. The document is intact and it is the MARKING that disqualifies it, which is worth keeping separate from a short read: both completeness gates pass on it, each leg terminating at the exchange’s own printed TOTAL line rather than at a judgement that the extract looked long enough. A bulletin that is complete and provisional and a bulletin that is truncated both fail to reach the table, and only one of them is evidence about the publisher. 🔴 AND THIS READING RESTS ON ONE INSTRUMENT FEWER THAN ITS PREDECESSOR’S: #201 WAS NOT ARCHIVED. The October 18 entry above can point a reader at _ops/cme_daily_bulletin_s61_20261016.txt and let them check bulletin #200 for themselves. This reading has no such extract — the archive on this reference ends at that October 16 file — so what is recorded here about #201 rests on the reading alone and cannot be re-checked once the /current/ path rotates, which it does every business day. That is stated rather than smoothed. A sweep that wrote this entry in the same voice as the one above it would be claiming an instrument it does not have. 🔴 THE STRIP IS NOW BEHIND TWO PUBLISHED TRADE DATES RATHER THAN ONE, AND BOTH ARE BEHIND IT FOR THE SAME REASON. Friday, October 16 (#200) and Monday, October 19 (#201) have both been read and both are PRELIMINARY, so the distance between this table and the newest settlement the exchange has published is neither a gap in the publisher’s record nor a gap in this reference’s reading — it is the FINAL-only rule doing what it was adopted to do. The gap sentence beside the table computes that distance from the two dates rather than asserting it, so it advances without anybody re-typing it. The bulletins this reference has now seen PRELIMINARY on first sight are #185, #186, #187, #188, #190, #200 and #201; they are enumerated rather than totalled, because #189 was never read at all and an unread document is not a PRELIMINARY one. Two consecutive provisional bulletins are not a cadence either, and nothing is claimed here about what #202 will carry.

The curve against the inventory data, and what each independently says

Two measurements of the same market, from different publishers on different clocks. They agree, and the agreement is the point.

Global observed oil stocks fell about 509 million barrels from end-February to end-September 2026, an average draw of 2.4 million barrels a day sustained across seven months, and the most recent month's decline was concentrated in oil on water rather than on land. Sustained draws of that size are the behaviour associated with backwardation: holders being paid to release inventory rather than to keep it.

Until this page carried the strip, that was all it could say, and it said so as an inference from a consequence — because a physical shortage draws stocks and also produces backwardation, so the stock data alone is consistent with either ordering. The strip now settles the first half of it as a measurement. WTI is backwardated by $17.84 over twelve months, or 18.2% of the front price, and Brent by $20.84. No storage economics offsets a carry that negative, so every holder of a tank is being paid to empty it.

What that does not settle is causation, and the improvement should not be overstated. Two independent measurements now agree on the state of the market where before there was one measurement and one inference. Neither establishes which came first, and the Institute is not going to claim it does on the strength of a better citation.

The rest of the reference