Actionable insights — Portfolio Update: One "HAL" of an Opportunity

1. Price a politically-timed headline deal before you trade around it

1. When a government announcement moves a commodity, do not argue about whether it is true. Ask a narrower question: how many barrels (units, tons, megawatts) does it add, and in which calendar year? Everything else is commentary.
2. Attack the timeline first, because it is the cheapest test. Infrastructure that has been neglected or destroyed needs rebuilding before it produces anything, and that is measured in years, not quarters.
3. Check for a hard legal or constitutional obstacle to the specific structure announced — and, in the same breath, name the structure that is permitted. A one-sided debunk is advocacy; naming the loophole keeps it analysis.
4. Examine the counterparty's history with the people it needs. Prior expropriation, unpaid judgements and failed litigation are a behavioural forecast about whether the required capital actually shows up.
5. Ask cui bono on the timing: is there an election, a rate decision, a debt auction? Announcements that lower a politically sensitive price just before a vote should be discounted as messaging until physical volumes appear.
6. Then state the conclusion in price terms: no near-term supply, therefore no near-term price impact — and treat any selloff it causes as a chance to buy the physical thesis, not evidence against it.

• The first evidence that converts messaging into supply: signed production-sharing contracts (not memoranda), rig counts moving, a major booking reserves. Until then, each fresh announcement is a price event without a physical event — and the reflexive response is to check whether the affected equities got cheaper.

2. When the headline's beneficiary is a decade away, move up the supply chain to whoever gets paid doing the work

1. Accept the debunk, then ask the follow-up most readers skip: if this does happen in some form, who is paid first? The owner of the asset is paid last; the contractor is paid during.
2. Identify the step that is required regardless of who wins the deal or who finances it. Indifference to the eventual owner is what makes the supplier a cleaner expression than the producer.
3. Check whether the same supplier has a larger, nearer job already in the queue — one that does not depend on the speculative headline at all. If the headline is the option and the existing job is the base case, you are being paid to wait.
4. Size the base-case job honestly, using the physical damage rather than the political narrative.
5. Confirm the capability is scarce — a short list of firms with the equipment, crews and service network to do the work at scale — otherwise the demand arrives without pricing power.
6. Prefer the supplier when the commodity price itself is being politically suppressed: the service company is paid for activity, which can rise even while the headline price is being talked down.

• Service pricing and utilization, not the oil price — the tell that the rebuild is being booked. The thesis breaks if a genuine supply glut arrives (Hormuz normalizing and Mid-East capacity restored) before the backlog of repair work converts into contracts.

3. Use the P/E-versus-P/S disagreement to tell trough earnings from peak earnings

1. Never accept a single cheapness measure on a cyclical. Chart the name's price/earnings and price/sales over five years side by side and see whether the two agree.
2. If P/E is at a historic low but P/S is only ordinary, the market is paying a normal price for the revenue and a small price for the profit — which means margins are high relative to history. That is the classic peak-earnings value trap.
3. If the disagreement runs the other way — or if industry conditions are demonstrably depressed — the low multiple sits on trough profits, and the same revenue can generate far more earnings later. That is the version worth buying.
4. Settle the ambiguity with an industry-conditions check rather than a ratio: are volumes, day rates, utilization anywhere near boom levels? Say so explicitly, in one sentence.
5. Publish the ratio that disagrees with you. An analyst who volunteers the weaker measure ("it looks cheap, but not as decidedly") is doing analysis; one who quotes only the flattering one is selling.
6. State the resulting asymmetry as the thesis: neither revenues nor profits maxed out means the upside is in the numerator and the multiple.

• Margins recovering toward prior-cycle peaks — the point at which the low P/E stops being a trough signal and starts being a warning. Also watch the reverse tell on SLB: a premium-to-market multiple returning (it currently sits "at a slight discount" after years of "a significant premium") would mean the re-rating half of the trade is done.

4. Keep a dated ledger of every call on one name — and let the ledger, not the story, make the sell decision

1. For a high-volatility name you intend to own for years, keep a written, dated record of every buy and trim recommendation with the price attached. The ledger is the discipline; memory is not.
2. Re-read it before every action on that name. The question it answers is the one conviction cannot: where am I in this security's own range, and what did I do last time it was here?
3. Buy at the valuation trough with a stated absolute anchor ("as cheap as it had ever been at a sub-10 P/E"), not merely on a percentage drawdown.
4. Trim into strength, repeatedly and partially — several harvests across a rally, each one small enough that being early costs little.
5. Publish or at least record the misses alongside the hits; a ledger that only shows the good trades cannot be used as a decision tool.
6. Never let the ledger become a rule to exit. Its output is a size, not a binary.

• The level that would invalidate the ledger's pattern rather than confirm it: a decisive break above $60 and a new five-year high, which he explicitly says "could, and should, happen." A name that stops respecting its historical range is re-rating, and the swing discipline must then give way to holding the core.

5. Fund a new position by trimming its winning sibling — a switch, not a purchase

1. When conviction rises on a name, ask where the money comes from before asking how much to buy. The cleanest source is usually the same theme's winner, not cash and not an unrelated holding.
2. Require the two names to share the same demand driver so the switch keeps the thematic exposure constant and changes only the expression. Theme risk unchanged; relative risk taken deliberately.
3. Compare them on relative grounds — valuation gap, year-to-date performance gap, distance from resistance — rather than on absolute merit. Both can be good; only one can be the better use of the next dollar.
4. Make the trim partial and say so. A full exit converts a relative-value view into a directional call you did not intend to make.
5. Leave the trimmed name's upside case intact and in writing, so the position can be rebuilt without reversing a published opinion.
6. State the time horizon of the preference explicitly — "near-term" is a different claim from "better business," and conflating them is how switches turn into permanent mistakes.

• The spread between the two. If HAL clears its long-standing ~$43 ceiling while SLB stalls below $60, the switch is working and the remaining SLB core can fund more; if SLB breaks out and HAL fails at $43 again, the relative call is wrong even though the sector call may be right — and that is the distinction to check first.

6. Re-underwrite a name whose last call failed — state the miss, then say what has changed

1. Before re-recommending a name you already recommended, quote your prior call and its date in the same paragraph. Anything less is a new pitch pretending to be a first look.
2. Name the specific cause of the failure — a price move, a guidance change, a macro shift — rather than attributing it to "the market."
3. Separate what broke the trade from what would break the thesis. A guided-down quarter inside a multi-year rebuild story is the first, not the second.
4. Check the price paid now against the price paid then. A failed call that has produced a materially lower entry is a better setup, provided the thesis is intact.
5. Look for a small confirming behaviour since the low — relative strength on weak days is the cheapest, most honest of these — and label it as a tell, not a thesis.
6. Re-state the resistance level that has repeatedly rejected the stock, so the next attempt is measured against a number rather than a feeling.

• The fourth failure at ~$43. Two rejections are a level; a third with better fundamentals behind it is a coiled spring; a fourth on improving earnings would say something is wrong with the demand story that the P/E is not capturing. Also watch the next guidance update — the same "guarded outlook" repeated would move this from a trade problem to a thesis problem.

7. Underwrite the earnings; treat the multiple expansion as the free option

1. Split any bull case into its two arithmetic parts: the E going up and the multiple paid for it going up. They compound, which is why the combination is where outsized returns come from.
2. Underwrite only the first. Build the case so it works on earnings recovery alone, at today's multiple.
3. Establish that the multiple has room by comparing it to the name's own history and to the index — an absolute number means little across cycles.
4. Identify what would cause the re-rating (visible multi-year work, a category change, a return to a historic premium) rather than assuming mean reversion happens on its own.
5. Say plainly when both legs are expected, so the reader knows the base case is the smaller one.
6. Take the re-rating gain when it arrives; that is what the ledger in insight 4 is for.

• Earnings revisions turning up while the multiple stays flat — the ideal sequence, and the window in which the position should be built. If the multiple expands first, on unchanged estimates, the free option has already been exercised by other buyers and the remaining return depends entirely on delivery.