Actionable insights — Backwardation: A Forward Oil-Price Clue?

1. Go looking for the data that argues against your own book — and publish it with the bias declared

1. Name your exposure explicitly, in writing, before presenting evidence — the sector, the direction, the specific sub-positions. A reader (and you) can only discount for a bias that has been stated.
2. Assume the bias is present regardless of effort: it is "inevitable when there are humans involved," not a failure of character. Treat it as a standing condition to be managed, not an accusation to be denied.
3. Then invert the research question. Instead of "what supports the position?", ask "what series would tell me this position is early or crowded?" — and go find it.
4. Prefer a series that is mechanically independent of your thesis. Your thesis here is fundamental (supply, demand, capex); the disconfirming series should therefore be structural or positional, so it cannot simply be your own argument reflected back.
5. Publish it even when — especially when — it says the opposite of what you own. The publication is the commitment device; a signal kept in a drawer gets quietly reinterpreted.
6. Check whether the analysis is genuinely differentiated. If every research provider you subscribe to already runs it, it is priced; if none do, it is worth the work.

• The tell that the exercise has failed: a disconfirming signal that arrives and is immediately explained away as "different this time." Note that Hay does not do that here — he accepts the signal and adjusts the plan around it. Watch also for the opposite failure: treating one contrary indicator as sufficient reason to abandon a fundamental thesis it was never designed to test.

2. When a signal's regime changes, move the reading from the sign to the magnitude

1. Establish the indicator's historical base rate: how often was it "on" in the old regime? Backwardation was once "a rarity," so its mere presence carried information.
2. Check whether that base rate still holds. If the condition is now the default — "for years now, it has been the prevailing condition" — then a binary reading generates a permanent, useless signal.
3. Do not discard the indicator. Re-express it as a continuous variable and look for structure in the distribution: where do the peaks cluster, and what followed them?
4. Set the new decision rule on the magnitude, with an explicit numeric threshold you can check against a live quote.
5. Separate the questions you can answer from the ones you cannot. Why the regime changed may be unresolved — "worthy of a future Daily" — and the indicator can still be used while that stays open.
6. Re-audit the threshold periodically; if the regime shifted once, it can shift again, and a stale threshold is worse than no threshold because it looks precise.

• The spread spending a long stretch parked near the threshold without a correction following — that is the evidence the regime has moved again and the $10 line needs to be re-cut. Watch also for the reverse regime break: a return to contango, the old normal, which would mean the physical tightness that underwrites the whole bullish energy thesis has ended.

3. Calibrate a threshold off the ordinary peaks, not the famous crises

1. List the extreme observations first, with their causes and dates — they are memorable and they anchor the scale.
2. Then explicitly discount them for sample size. Two crisis observations are an illustration, not a distribution; say so rather than letting the drama do the persuading.
3. Go back to the same chart and mark the smaller local peaks, which are "a bit harder to see" precisely because nothing famous caused them.
4. Ask whether the same consequence followed those. If it did, you have many observations instead of two, and a threshold that will actually trigger within an investing lifetime.
5. Set the threshold at the level where the ordinary peaks occurred, not where the crises did — a rule that only fires at the 2022 extreme is a rule you will use once a decade.
6. Match the expected magnitude of the outcome to the magnitude of the trigger. Small peaks preceded corrections "at least to a degree"; extremes preceded "cliff dives." Do not import the crisis outcome into the ordinary signal.

• Threshold creep — quietly widening $10 to $12 or $15 because the signal fired and the position is uncomfortable. Write the number down before it triggers. Watch too for the survivorship trap in "less extreme peaks": the peaks that were not followed by a correction are the ones hardest to see on a chart and easiest to skip.

4. Before trusting a market-derived signal, ask who is on the other side of it

1. Identify the participants who actually set the price in that instrument. A curve dominated by financial speculators is a positioning reading; one populated by physical users is closer to a scarcity reading.
2. Name the commercial hedgers specifically — the airline hedging jet fuel, the refiner, the utility — rather than asserting "real money" in the abstract.
3. Note what those participants are optimising for: hedgers are buying certainty about an input cost, so they will pay up for near-term barrels when supply feels genuinely tight, not when a narrative is fashionable.
4. Use that to grade the signal's reliability, and to decide what it is a signal of. An extreme reading in a hedger-driven curve marks the moment physical anxiety peaks — which is, historically, also the price peak.
5. Apply the same test before importing any curve, spread or ratio into a process: futures term structure, credit spreads, freight rates, forward FX points. The participant mix determines what the number means.

• A shift in the participant mix itself — a curve that becomes financialised (index funds, ETFs, systematic trend followers) stops reading as a scarcity gauge even though the number looks identical. Commitments-of-Traders style positioning data is the check, and it is the same lens the paper-positioning theme applies from the other direction.

5. Let a contrary signal change your entry plan, not your thesis — and size the caution to the fundamentals

1. Classify the signal by horizon before acting on it. A futures-curve extreme is a weeks-to-months timing observation; it says nothing about a multi-year supply thesis and should not be allowed to.
2. Re-check the fundamental set-up independently. If it is unchanged, the correct output is a plan, not a trade.
3. State the caution's magnitude explicitly — "of a modest nature" — and tie it to the reason: the strength of the fundamental backdrop is what caps how much weight the contrary signal gets.
4. Convert the expected move into a pre-committed action on the other side: decide now that a dip is a buy, so the decision is made before the price is falling and the news is bad.
5. Do not chase in the meantime. The same signal that makes the dip a buy makes the current level a poor entry — that is the whole point of a timing indicator on a position you already want.
6. Keep the two statements in one breath when you publish, so no reader can take the caution as a downgrade or the conviction as a licence to chase.

• The dip arriving with a reason attached — a demand-destruction print, a supply restart, a ceasefire — and check whether the reason invalidates the fundamental case or merely delivers the expected correction. That is the one distinction the indicator itself cannot make, and it is where the pre-committed buy plan has to be re-examined rather than executed on autopilot.

6. Require a second, independent read before acting on a single chart — and accept that one signal can be bullish and bearish on different horizons

1. Never let a single series carry a stance. Look for a second observation from a different market — here, the equity, not the commodity — pointing at the same conclusion.
2. Read the equity chart for both facts at once: the structural fact (what a breakout to an all-time high says about the long trend and about capital finally arriving in a neglected sector) and the tactical fact (how far price has travelled from its own moving averages).
3. Resist collapsing the two into one rating. "Confirmation of the long-term uptrend" and "quite extended on a near-term basis" are both true, and forcing a single verdict destroys information.
4. Say which horizon each observation belongs to when you publish it, so the reader knows which one to act on.
5. When two independent reads agree on the near-term risk, treat that as the threshold for flagging it — one is a chart observation, two is a stance.
6. Keep the long-term rating unchanged unless the structural read breaks. Extension is not deterioration; a failed breakout would be.

• XLE falling back below the breakout level and staying there — a failed breakout is a different event from an extended one, and it is the read that would actually challenge the structural case. Watch as well for divergence between the two gauges: the curve normalising while the equity keeps climbing (equity-led, discounting a better future) or the equity rolling over while backwardation persists (physical tight, equity de-rating on something else — costs, capital discipline, taxes).