How decisions get made

Three ideas, and how they bound each other

I work with three bodies of ideas, and each has its own job.

Stay in the game. Markets aren’t ergodic: what a strategy earns on average across many investors isn’t what one investor earns over time, because a loss deep enough ends the game for that investor. So a strategy with a real chance of ruin is unacceptable, whatever its expected return. This is the constraint that bounds the other two, through position sizes, leverage and the option structures I use.

Simple rules for an uncertain world. Markets are open-ended: the list of what can happen is never complete, and a precise probability, especially for a financial analyst relying on DCF models with dozens of assumptions, is just a blind belief in the result, driven by a false sense of confidence that comes from the complexity of the process. I don’t want to fall into that trap. In finance and economics, as in every social science, simple, robust rules tend to beat complex models: one good reason, a short decision tree, a clear threshold. I match the tool to the problem, and I treat a problem as calculable only when I can say why it is. This follows Gerd Gigerenzer’s work on ecological rationality.

Checks on how I act. Even with our world as uncertain as it can be, some biases are real: anchoring on a purchase price, holding losers and selling winners too early, loss aversion, being swayed by how a choice is framed. I use them as checks before and during each trade, not as a way to choose trades. This follows Daniel Kahneman’s work on judgment and decision-making.

Complex processes can be there to spare the decision-maker the weight of the blame, rather than to improve decisions. I watch for it.

Mandate

Objective

At least 8.5% a year in absolute terms, with less risk than the equity market

Distribution

About 5% of capital a year, converted to euros for household spending as it’s needed

Exposure

Rarely above 120% of capital; 140–150% at most, and only in special conditions

Instruments

Mainly US stocks and ETFs, listed options on them, and Treasury bills as cash. No market or instrument is ruled out: now and then, European and Canadian stocks and options, and options on futures

Returns are measured against two funds I could hold instead, and the strategy has a rule that would end it: both are on the Method page.

Missing the absolute target does have a consequence. While the comparison with the funds holds, it is not the kill rule but a temporary cut in distributions, driven by market conditions. That is perfectly fine: it is the muscle a privately run portfolio can flex to grow stronger when the tide rises, instead of being crushed like public companies with payout ratios above 80%, or even above 100%, that can’t cut their dividends even temporarily, for fear of losing their status as dividend or high-income payers and being dropped from the passive ETFs that follow those strategies.

From research to decision

1. Research, by hand. Filings first: every figure comes from the company’s filings or from my own calculation on them. Usually I start by taking notes in a simple notebook and digging deeper into the business model, coupled with listening to the latest earnings calls.
2. A starter position, while the research runs. 10–25% of the intended size. If the opportunity seems attractive, I don’t wait to finish the research to start deploying; markets sometimes move fast.
3. The note. Published with the notebook page it came from, a line from that page as its title, and my position on the day.
4. Tests, before the results. Three or four per position, each with its starting point, registered before the results that decide them.
5. Full size, or not. Only once the research is complete and the conditions are met.
6. Every quarter. The tests are read on the results. The reading goes on the note, and the research PDF is re-done as a new version; earlier versions never change or get deleted: they are archived as the history of the process, so the lessons learned don’t get forgotten.
7. Exit by rule, then the closing file. When the thesis is replaced, the exit, its pace and what would pause it are written down. After the exit: what I got right, what was luck, and what changes in the process.

Building a position

I build positions in stages. A starter position of 10–25% of the intended size goes in while the research runs. Full size, the strategic allocation or the maximum position size for the opportunity, comes only once the research is complete and I judge the conditions are met.

The option structures I use

I often hold a position through listed options, alongside the shares or instead of them. These are the structures, drawn at expiry:

The short strangles are never naked on the call leg; every call I write has a long call, the shares or a synthetic behind it, so the loss above the call strike is capped. A short option’s best case is its premium; its worst case grows with the move against it. That is why position sizes and exposure limits come before the premium. How each structure counts in my exposure is on the Method page.

What each structure does is the easy part: the science. When to use it is the art. The combination of both is what makes this job a craft, and it can’t be explained in words: it comes from intuition, which in turn comes from experience, both underpinned by the fundamentals of the stock, the industry and the sector, and by sound judgment.

Tests written before the outcome

Every position carries three or four tests, written before the results that decide them, each with its starting point. A test that fires puts the thesis under review; for a position I’m leaving, the tests say what would pause the exit. The readings go on the note on the day I take them, and a test keeps its wording once registered: a new test is registered as new, with its date.

PepsiCo’s three exit tests were registered on 8 Oct 2026, when that day’s Q3 2026 results brought the exit forward; those results are their starting points, and their first reading comes with the Q4 2026 results, if the exit isn’t complete by then. The tests for Booking Holdings, Intuitive Surgical and Uber were registered on 20 Oct 2026, before their Q3 2026 results. Each note’s tests are on the Notes list.

Exits

An exit starts when the thesis is replaced. I write down why, the pace, and what would pause it, and I keep the record as the exit runs. PepsiCo is the first documented this way: thesis replaced on 5 Oct 2026, with an exit over one to two quarters, brought forward to one quarter on the Q3 results of 8 Oct, with three tests that would pause it.

I don’t have specific loss-limit rules, as I’m driven by fundamentals, not by price action: the maximum loss is constrained by the size of the original investment. Nor do I have a time limit on waiting for the market to recognize the move I expect. However, while I’m not willing to take a loss that isn’t supported by fundamentals, I’m aware of price regimes, market narratives, sector rotation and other market developments. That means that while I continue to hold the position, I usually trade around the strategic asset allocation, sometimes adding a tactical asset allocation overlay, other times with yield-enhancing option strategies, etc. This has the benefit of cutting unrealized losses, improving the yield on income-held positions and/or lowering my cost basis while I wait for the fundamentals to show up.

Risk management during an exit can include a hedge. Each hedge is recorded in the notebook on the day it opens, the page is photographed and its fingerprint published, and the record is opened at the exit, in per-share terms and percentages (the PepsiCo note).

The 72-hour rule

I don’t trade against a published view in the 72 hours after publishing it, except to carry out a pre-stated exit or risk limit. Each note states my position on the day it was published.

Sources

Every figure in a note is marked by where it comes from: i a company filing; ii management, on calls and in presentations; iii a third party; H my own calculation from the filings. I don’t take numbers from data aggregators, which often disagree with each other on something as basic as a P/E.

What I don’t do

• Take outside money, give advice or sell anything.
• Hold a strategy with a real chance of ruin, whatever its expected return.
• Take figures from data aggregators.
• Build financial models that try to predict future financial statements line by line, based on a mountain of assumptions. That is the job of the sell-side analyst, whose aim is to reach a price target to publish in the financial media. On the other hand, what buy-side analysts like me do is focus on a combination of quantitative and qualitative data to make decisions where risk-adjusted returns seem well balanced, without trying to reach a stock’s present value from a five-year DCF forecast model.
• Let a model listen to an earnings call or write a note for me.
• Trade against a published view in the 72 hours after publishing it, except for a pre-stated exit or risk limit.
• Hedge the dollar.
• Write calls without the shares or a synthetic behind them.