Allocation · macro theses

Why these choices

The portfolio's design choices, and why.

The allocation is fixed by macro theses, not by performance. It is revised only when a thesis is structurally falsified — never because of a temporary drawdown. Each thesis below carries explicit refutation markers, and a hard review date for the page.

Anti-bonds

Long-duration sovereign debt is structurally unattractive. Refuted if: G7 debt-to-GDP enters a durable decline, or 10-year TIPS yields hold above 2% for 24 months, or core inflation holds below 2% for 24 months.

Gold is insurance, not an engine

Gold holds up in inflationary and monetary regimes and gives way in others — its record depends on the decade. That makes it a regime bet, not a robust diversifier, so it no longer sits in the strategy's defensive sleeve, which is now plain cash. Gold earns its place only as static, self-custodied catastrophe insurance held outside the momentum system — a hedge against the rupture the engine cannot trade through, not a lever the engine pulls. Refuted if: cash proves a worse defensive asset than gold across a full cycle.

Pro US-style (secular)

Growth/value rotation reflects invariant investor psychology, not a passing regime. Validated by the century test — the rotation premium is distributed across nearly every decade since the 1920s (on Fama-French book-to-market: the logic, not the exact IWF/IWD/IWB trio), though it deepens the drawdown in a generalised crash with no style refuge, as in the 1930s. Refuted if: the rotation premium vanishes for 15 years.

Pro real assets (the AI build-out)

AI is deflationary for services, not for real assets: data centres run on electricity and metals, so the build-out puts a structural bid under energy and materials even as software costs collapse. This is a worldview, not a tilt — the position is never hard-coded; it is left to momentum to take or leave. Refuted if: AI capital spending contracts for 24 months, or real-asset demand visibly decouples from compute growth.

The wrapper is part of the strategy

A momentum strategy rotates, and rotation realises gains. In an ordinary brokerage account (compte-titres), every sale of a position that has risen is a taxable event — so the capital that keeps compounding is trimmed at every rotation. In any single month the drag is tiny; across decades it compounds against you.

Inside a French life-insurance contract (assurance-vie), switching between funds is not a taxable event. The strategy rotates in full exemption; tax falls due only when money is withdrawn. So the capital compounds on its gross balance the entire way, instead of one the tax authority trims at every winning trade. The edge grows with how often you rotate — it is worth far more to an active strategy than to a fund left untouched. Deferral also turns into a discount: once the contract passes its eighth year, withdrawals get an annual allowance and a reduced rate, so a yearly partial exit sized to the allowance comes out almost untaxed.

The eight-year clock runs from the day a contract is opened, not from when money enters it. An old contract is therefore an asset in its own right — and it is why money never moves between insurers here: a transfer counts as a withdrawal, which both triggers tax and resets the clock to zero. We adapt the strategy to the funds each contract already lists rather than chase a better fund elsewhere. That constraint is the whole reason the backtest runs on faithful proxies of the funds actually held, not on whatever ticker would score best.

None of this is free. The wrapper levies an annual fee on top of each fund's own cost; the investable universe is limited to the funds the contract lists (hence the proxies — and the gaps, such as commodity-sector futures, which no contract offers); and a switch settles a day or two later, at a price you cannot see when you place it. Most consequentially, in a systemic crisis the regulator can temporarily suspend redemptions and arbitrages — the shelter can be bolted shut precisely when you would want to act. The euro general account is the first support frozen, which is one reason it is not used as the cash refuge (see the “Why not the euro fund?” note under Methodology).

The wider lesson echoes the one about the parachute: the structure that holds a strategy is not an afterthought. For something that trades, the tax wrapper moves the compounded outcome as much as a point or so of annual return would — as the chart below shows — and it deserves to be chosen as deliberately as the assets inside it.

Mandatory review: May 2028 at the latest.