Actionable insights — The Bull Market Almost No One Sees

1. Distinguish a failed spike from a range expansion that stuck — the retracement's stopping point is the evidence, not the size of the drop

1. Identify the level that capped the asset for years before the breakout — the shelf price repeatedly failed at across separate cycles, not a trendline you drew after the fact. Two or more independent touches, years apart, is the standard.
2. When a violent, event-driven move breaks that level and then corrects, do not judge the correction by its depth. A 40% retracement is meaningless on its own; every blow-off produces one.
3. Ask instead where the correction stopped. If it terminates at the old resistance and holds, the market has converted a ceiling into a floor and the breakout was a genuine regime change. If it slices back through and keeps going, the spike was event-driven noise and the old range is still in force.
4. Give the pattern a stated frequency rather than treating one instance as proof — this behaviour is common after sharp breakouts and after excessive advances, which are two different causes with the same signature.
5. When the same sequence repeats at a higher level with a shallower correction, treat the shallower drawdown as confirmation that the base has migrated up, not as a smaller version of the same failure.

• The invalidation is precise and worth writing down: a decisive break back below the ~500 shelf would retire the whole "regime change" reading and restore the 2010s range. Equally, a second failure at ~740 on lighter breadth is the double-top Hay concedes is possible — so the level to monitor is the one where the two interpretations diverge, not the highs.

2. Let the chart grant permission and the physical balance supply the reason — never the other way round

1. State both technical interpretations of the same pattern explicitly, in the bears' own language, before choosing between them. If you cannot articulate the opposing read in a sentence, you have not looked at the chart.
2. Accept that price action alone cannot settle it — a double top and a mid-bull consolidation look identical until one of them resolves.
3. Go to the physical layer for the tiebreak: inventories, supply deficits, mine closures, recycling rates, demand trends. Ask whether the shortage that justified the first spike is now larger or smaller than it was.
4. Apply the honesty test to the prior episode. If the shortage feared last time did not materialise, say so — that admission is what gives the current claim standing.
5. Support the general claim with one fully worked, checkable example rather than a list of assertions. One deficit you can audit end-to-end beats ten you cannot.

• The single-example trap. Palladium's deficit is idiosyncratic — it rests on a mis-forecast of the hybrid vs EV mix — and does not itself establish a "lengthy list" of shortages. The claim needs a second and third audited example (copper, uranium, oil) before the asset-class conclusion is carried by evidence rather than by one good case.

3. Price one asset class in units of another, over the longest series you can get, and read the extreme as a constraint rather than a forecast

1. Build a ratio, not two charts side by side. Dividing one index by the other removes the currency, the inflation and the denominator-of-money problem in one step, and answers "how much of X does one unit of Y buy" — the only question that matters for relative allocation.
2. Extend it as far back as the data allows. A ratio's usefulness is entirely a function of range: five years tells you nothing, fifty tells you where the boundaries are.
3. Anchor the extremes to named events so the series is interpretable rather than abstract — each peak and trough should correspond to a regime a reader can recognise.
4. Locate the present reading against those anchors, and specifically against the most extreme prior trough, not against the average. The question is whether you are outside the historical range or merely low within it.
5. Draw the conclusion as an identity: a ratio at a boundary resolves through the numerator, the denominator, or both. Do not claim to know which — that is a separate call requiring separate evidence.
6. Take the directional lean from a different piece of work, and say where it comes from, so the two claims can fail independently.

• Composition drift in either leg. The GSCI is heavily oil-weighted and the S&P's sector mix in 2026 bears little resemblance to 1970's, so part of the ratio's decline is structural — the index economy simply consumes fewer barrels per dollar of output. Before treating 0.7 as a mean-reversion signal, ask how much of the fall from 4.3 (2011) to 0.7 is composition rather than valuation. And note the ratio can resolve without commodities rising, if the S&P merely stops.

4. In a volatile bull market, commit to a two-sided discipline in advance — sell into spikes, add on weakness, and say so before either happens

1. Separate the direction call from the path call. Being right on the trend and wrong on the drawdown is the standard way a correct thesis gets abandoned at the worst point.
2. State the volatility expectation up front, as part of the bullish claim rather than as a caveat added after a bad month. "We believe this continues, though we anticipate ongoing volatility" is a commitment, not hedging.
3. Pre-commit to both actions: take profits on excessive advances, add on corrections. Write down what counts as each — in this family of assets, spikes have run ~30–40% above trend and corrections have retraced a third.
4. Apply it across the whole expression — the commodity itself and the producers — because the two legs spike and correct on different schedules and the equity leg is the more volatile.
5. Judge the previous cycle's guidance honestly before extending it, and be explicit that past reward does not carry forward automatically.

• An unfalsifiable version of this policy. "Sell spikes, buy dips" describes every outcome after the fact unless the trigger levels are specified in advance. The archive's usable form is quantitative — a ~⅓ retracement within an intact uptrend was the stated PALL entry — so hold the guidance to that standard rather than to the general statement.

5. Look for the market where the physical balance and the narrative about it have decoupled — that is where the repricing is forced

1. For each candidate commodity, track two things separately: the observable physical state (inventories, days of cover, deficit years, mine and refinery outages) and the prevailing story about it (consensus notes, financial media, social sentiment).
2. Flag markets where the two disagree in a specific direction — physical tightening while the narrative says abundance. Disagreement in the other direction is far less actionable, because a bearish story about a well-supplied market is simply correct.
3. Rank candidates by how undismissable the physical evidence is becoming. A deficit visible only in a forecast can be argued about indefinitely; one visible in drawing inventories or a closed mine cannot.
4. Expect the narrative to hold longest where positioning is largest and shortest where the physical evidence is most public — which is why the repricing tends to arrive first in the most-watched, most-liquid market rather than the most obscure one.
5. Treat the decoupling itself as the trade signal, and the moment the story becomes untenable as the catalyst you are waiting for.

• The measurable version of "increasingly difficult to dismiss": OECD commercial crude inventories and days of cover, the shape of the futures curve (backwardation is the physical market's own vote), and refinery product cracks. If those are not tightening, the decoupling is a claim about sentiment rather than about barrels — and this insight only works when the physical leg is independently checkable.

6. Read a house's notes as a single position, not as separate opinions — the two halves of one trade often arrive on consecutive days

1. When a research house publishes frequently, resist evaluating each note in isolation. Consecutive pieces are usually two views of one allocation, and the argument for each side is left in the other note.
2. Identify the shared object. If one note argues an asset is being held up by something temporary and the next argues a different asset is priced at a historic discount to that same asset, they are the numerator and denominator of a single ratio trade.
3. Check whether the pair is mutually reinforcing or merely correlated. Two calls that fail for the same reason are one call, and should be sized as one.
4. Take the conditional from the note that supplies it. When a piece deliberately declines to give direction — as a well-constructed ratio argument should — do not treat that as agnosticism about the position; look for where the direction was argued.
5. Write down the single falsifier that would break both halves at once, since that is the real risk being carried.

• The common failure mode: both halves depend on the AI capex cycle continuing to absorb capital. If that cycle broke, US equities would fall and industrial commodity demand would weaken — the ratio might not close at all. The pair is a bet on rotation, not on a market decline, and it fails if the money leaves both.