Actionable insights — Oil Crisis Not Over Yet, This is the Real Bottleneck

3:23 1. Net a supply loss only against inventory that can actually be mobilised

1. Take the headline global stock number and decompose it by function, not by owner. Ask of each tranche: can this be withdrawn without stopping the system that holds it?
2. Strip out the operating fill — pipelines that must stay liquid to flow at all, cargo permanently in transit to sustain the seaborne trade (voyage days × daily seaborne volume), and tank heels below the suction point.
3. Strip out, or discount, government reserves: they exist but are released politically and at a pace the government chooses, not at the pace the deficit demands.
4. Compare the deficit (barrels/day lost × days lost) against the remainder. If it exceeds it, the market must clear by price, because there is nothing left to release.
5. Use the analogy as a discipline: reported inventory is working capital, not a savings account — "you can't drain oil out of a pipeline, otherwise it stops operating."

• Any inventory series quoted as a single number without separating operating fill from usable stock; hub stocks at operational minimums (Cushing) while national totals look adequate; days-of-cover falling faster than absolute levels.

11:23 2. Bridge supply to demand through the refinery layer before believing a demand number

1. Notice that "supply" and "demand" in a commodity balance are two different physical things: supply is wellhead crude, condensate and NGLs; demand is finished product — gasoline, diesel, jet, petrochemicals. Between them sits a processing industry.
2. Ask how the reported demand figure is actually computed. In oil, "one of the big inputs in your model to estimate demand" is refinery runs, plus GDP — nine times out of ten a safe proxy, because refiners only run when there is a customer.
3. Test the proxy's precondition: is refining capacity the free variable, or the binding constraint? If refineries have stopped for a non-economic reason — war damage, trapped cargoes, an export ban — the proxy inverts and reports a supply-side outage as demand destruction.
4. Rebuild the number from independent end-use series instead: airline miles, vehicle miles travelled, general activity. If they disagree with the model, trust the physical observation.
5. Sanity-check the magnitude against history before accepting it. Demand losses have precedents with known sizes; a claimed loss "2x the global financial crisis" needs GFC-scale evidence.

• A reported demand change that tracks a refining-capacity change almost one-for-one; regional refined-product export bans; refinery utilisation diverging from mobility data.

17:06 3. Use the crack spread as the discriminating price between two stories

1. Write down what each competing explanation predicts about relative prices, not just direction. Demand destruction: crude and product both weak, spread roughly normal. Refining outage: product firm, crude soft, spread blown out.
2. Find the price that separates them — in oil, the crack spread — and check it against its own history rather than against an opinion. Normal is $10–20; anything multiples of that is a regime statement.
3. Test it at the moment of maximum stress for the alternative story: when crude round-tripped to its pre-war lows in June, did products follow? If they refuse to fall with crude, demand is not the binding problem.
4. Convert the answer into a quantity: if throughput is down X mb/d and end demand is flat, the product shortfall is X, and it must be coming out of inventory.
5. Choose the leg you own from where the shortage started. If it began upstream, the spread most likely closes by crude rising, so own crude, not the processing margin.

• Cracks holding while crude sells off; product prices that "never budged" through a crude round-trip; the spread's implied crude price versus the actual quote.

18:38 4. Audit whether a data series is observable before trusting its level

1. For every input in the balance, ask the mechanical question: how is this number physically produced? Reported by a government, sampled by a vendor, or modelled from something else?
2. Learn the measurement technology and its limits. Crude inventories are satellite-estimable because storage tanks have floating roofs that sit on the oil, so the shadow the wall casts encodes the fill level. Refined-product tanks have fixed roofs and give a satellite nothing — "there's really no way from space that you can say how much inventory is inside."
3. Distinguish "estimated well" from "estimated anyway." Client demand for a series does not create the ability to measure it: vendors published product-inventory estimates because customers asked, not because the method existed.
4. Weight your conviction by observability. Put the residual of the balance where the measurement is weakest, and treat the published figure there as an order-of-magnitude guess.
5. Corroborate the unobservable series with second-order evidence — physical shortages, rationing headlines, import scrambles in the regions the data does not cover.

• Which side of a balance the "balancing item" keeps landing on; non-OECD data that moves less than physical events imply; shortage reporting from countries the statistics don't cover.

23:46 5. Enumerate the exits — and size the position when both of them point the same way

1. List the ways the dislocation can end, not just the one the consensus is trading. Here there are two: it breaks (a shortage crisis), or it resolves (the strait reopens).
2. Work each path through to a price. The crisis path: an emerging-market product shortage propagating into a market carrying heavy speculative shorts with the delivery hub at operational minimums — "immediate doesn't happen immediately" in a squeeze.
3. Work the "good news" path just as carefully — it is the one people skip. A reopening restarts ~2.5 mb/d of Gulf refining, brings China's idle export refining back for a $100 crack, and leaves refiners, governments refilling drained reserves, and normal activity all bidding for crude at once.
4. If both branches are bullish, the position is not a forecast, it is an asymmetry — and the timing question ("which path") stops mattering to the sizing.
5. Identify what converts the view into price: usually a data release, not an event. "If that data were to be released to the market tomorrow, that would result in a pretty big panic, certainly a big short covering."

• Delivery-hub stocks at operating minimums into a large short base; strategic reserves drawn to near-minimums across the OECD and China; the first credible non-OECD product-inventory print.

29:54 6. Read the price as a risk limit, not as an opinion

1. Before inferring sentiment from price, reconstruct the positioning that produced it. Track gross and net speculative shorts through the episode, not just the net.
2. Attribute each move to a mechanism. A physical squeeze bids the prompt contract while deferred months lag — "they were forced to cover, and that's what helped bid up the price of the prompt contract, but not the far out months because they weren't playing there."
3. Use institutional behaviour as evidence of prior positioning. Desks fired at the outbreak of a war were almost certainly short into it: "I don't know why else you would fire them all."
4. Ask who is setting the constraint. If gross shorts rebuild past their prior extreme and are then cut again by risk committees rather than by conviction, "the only thing that's really affected the price has been what your risk team is allowing you to short."
5. Value the option that creates. A price that has risen without any bullish capital committed still has that entire buyer base ahead of it.

• Gross (not just net) speculative short interest versus its own year-to-date extremes; prompt-versus-deferred behaviour on a spike; sell-side and pod-shop team changes as a positioning tell.

36:17 7. Use the reaction to your own trade as the sentiment instrument

1. When you make a switch, record the audience's reaction — approval or a groan. It is a free, unbiased read on how crowded each side is.
2. Re-run the thought experiment periodically: if I reversed this trade today, would I be applauded or booed? Applause means the exit is crowded and the entry is not.
3. Treat the frequency of a question as data. "The number one question we get asked all the time is when are we going to buy our gold stocks back — and one of our answers is when everyone stops asking us."
4. Pair the sentiment read with a holdings read so it isn't purely anecdotal: are the marginal speculative holders (ETFs, funds) still holding the position they bought at the top?

• The direction the questions come from; ETF holdings that have "come down some, but not nearly as much as they have in past corrections."

38:06 8. Benchmark a correction on both axes — depth and duration

1. Define the signal that started the correction so the sample is comparable. His is the "silver sell signal": silver lags gold, then stages a violent catch-up rally, which historically marks the end of the leg for both.
2. Build the historical distribution of what followed that signal — here, roughly 40% drawdowns taking one to two years to bottom — and note the extremes as well as the average.
3. Score the current episode on both axes. A 20% fall in six or seven months matches only the single shallowest instance in fifty years on depth, and is far short on duration.
4. Do not stop at the price pattern: require the fundamental that caused the top to reset too. Speculative length that accumulated into the high has to be liquidated before the low.
5. Name what would invalidate the analysis — a source of demand large enough to overwhelm the liquidation.

• ETF holdings versus their pre-rally base; central-bank purchase run-rates; the rate setup flipping from priced-for-cuts to priced-for-hikes (the point at which even a hike can trigger short covering).

42:30 9. In a contract market, price the contract market — not the spot headline

1. Establish where volume actually transacts. In uranium the term contract price "is where 90% of the market transacts"; spot is thin and headline-driven.
2. When spot and term diverge, treat the venue with the volume as the signal and the other as noise — especially when equities are reacting to the noisy one.
3. Verify the divergence has no news behind it. A 30% equity drawdown "on absolutely no news whatsoever" while the term price makes an all-time high is a positioning event, not a fundamental one.
4. Check the two structural sides before acting: buyer coverage (are utilities contracted, or "very under covered in their long-term contract books"?) and the new-supply pipeline (how many years to first production, even for the flagship project?).
5. Adjust nominal records for inflation so you know whether a "record" is a real one — $95.50/lb broke the 2008 nominal high by 50 cents but is still below it in real terms.

• Term-versus-spot divergence; utility contracting coverage; the gap between a takeover headline and the project's actual first-production date.

43:57 10. Screen for the sector that cannot raise capital — then require a dated catalyst

1. Look first for industries that are ideologically un-ownable — "a four-letter word, completely starved for capital, no one cares." No capital means no new supply, so any demand increment goes straight to price.
2. Refuse to buy on neglect alone. Attach at least one specific, dated demand or supply event within a visible horizon — "a couple things in the next two or three months."
3. Prefer catalysts with a precedent you can point at, so the mechanism is not hypothetical: Europe rationed scarce gas by burning more coal in 2022, and is again unlikely to meet its winter gas requirements.
4. Identify the swing supplier and watch its policy, not its geology. Curtailment by the marginal exporter moves a seaborne market faster than any demand change.
5. Ask whether the trade needs a rerating of sentiment to work, or only the physical tightening. Prefer the latter — nobody has to like coal for the price to rise.

• European gas storage trajectories into winter and utility coal-burn data; Indonesian export policy announcements and the domestic market obligation; seaborne thermal freight and stockpile levels.