Methodology

The benchmark

Primary: MSCI World Net Total Return, with its investable proxy SWDA (iShares Core MSCI World UCITS ETF). The Net series deducts dividend withholding taxes and is what SWDA physically tracks, and the proxy line automatically carries SWDA's ~0.20% annual fee — so the comparison is against what a member's passive alternative would actually earn. Secondary reference: the S&P 500 with VOO (CSPX as the UCITS alternative). All comparisons shown in both USD and EUR.

Implementation note: the proxy series used in the calculation is IWDA, the USD-denominated line of the same ETF (ISIN IE00B4L5Y983) — the London SWDA line is quoted in pence, and used unconverted it overstated the 2016 benchmark return by more than twenty percentage points. Every benchmark series is converted to USD before comparison.

The evaluation window and the standard

The primary evaluation window is rolling five years, and the standard is simple: beat the MSCI World over it, net of membership fees. Outperformance of three percentage points a year over that window would be a strong result. This standard was amended on 26 August 2026 — it was previously three percentage points a year over rolling three years — and the amendment is recorded in the 90-question record rather than made quietly.

It is an objective for the period ahead, not a claim about the past, so here is the past: measured from January 2016 to August 2026, Almanac beat the MSCI World in 30.7% of rolling five-year windows and 35.3% of rolling three-year windows. The record's lifetime excess over the MSCI World is 1.10 percentage points a year; over the trailing five years it is 2.41. Judge the objective against those numbers, not instead of them.

The return math

Two return series are always calculated. Money-weighted return (MWR) is the headline figure — the actual experience of the capital, including the timing of external flows. Time-weighted return (TWR) is shown beside every benchmark comparison, because TWR is the standard for judging a strategy against an index. Uninvested cash and short-term government securities are included in performance. Negative cash balances and margin interest are fully included — financing costs reduce the reported return. External withdrawals are recorded as dated cash flows, not losses.

The published record

Real money since January 1, 2016, in a dedicated brokerage account. The record is published as an index set to 100 at inception. All charts, drawdowns and benchmark comparisons are expressed in index points and percentages; absolute currency amounts are never published, and positions appear only as portfolio weights at current market value.

Official prices

Official prices are EODHD end-of-day closes, with Interactive Brokers marks as fallback. Each security is valued at its primary exchange's official close; one daily portfolio valuation is struck after the US close (≈22:30 UTC). Closed markets carry the last official close forward, stamped "as of" the date. Any correction changing published performance by 0.1 percentage points or more receives a visible correction note — records are never silently rewritten.

The official call price

The official model-portfolio price for every call is the end-of-blackout reference price: the first traded price at or after the expiry of the 15-minute members-first blackout (next session's open if the market is closed). It is a price members could actually have obtained — they had the full head start — and it is not the Portfolio Manager's own fill, so the record can never be accused of using a privileged execution. The Portfolio Manager's actual fill is disclosed separately.

How wins and losses are counted

Win rate is measured per investment thesis: one idea is one unit, however many buys, adds or trims it contained. A thesis opens at its first published alert and closes when the target weight reaches zero; whether a re-entry is the same thesis is determined by the investment rationale, not a day count. A thesis is a win if its total realized P&L including dividends is positive in USD; open theses are excluded until closed. Win rate is always published together with average win versus average loss — the approach is asymmetric, and payoff size matters more than hit rate.

The ten gates

No idea becomes a published call without passing all ten gates; any FAIL blocks it:

1. Written differentiated thesis.
2. Perception-vs-reality gap evidenced from primary sources.
3. Quality gate.
4. Survival check — the balance sheet withstands the bear case.
5. Macro fit.
6. Liquidity floor.
7. Downside sized within the 25% portfolio-impact rule and 130% gross cap.
8. AI red-team by The Skeptic, objections logged.
9. Invalidation trigger written down in advance.
10. Portfolio Manager approval logged with timestamp.

The universe

Listed equities, ADRs and conventional ETFs on regulated exchanges (US, Canada, UK, EU/EEA, Switzerland, Japan, South Korea, Hong Kong, Singapore, Australia; others case by case). Minimum US$1 billion market capitalization; minimum US$2 million 20-day average daily traded value; the model position must be exitable within five trading days at no more than 20% of average daily volume. No short selling, no derivatives. Crypto-linked, pre-revenue biotech, IPOs, SPACs and leveraged ETPs only after a documented enhanced review.

Risk limits and the drawdown protocol

Maximum 25% estimated portfolio impact per thesis. Leverage only as margin-financed long exposure, capped at 130% gross. Every live position carries an individually set emergency backstop. Portfolio drawdown protocol:

• −15%: attribution and Skeptic review;
• −20%: re-underwrite everything and publish a drawdown memorandum;
• −25%: cut the weakest theses, no new margin;
• beyond −30%: no new risk until a full strategy review is published.

Accountability

Every losing closed thesis gets a written post-mortem: closure note within two business days, full post-mortem within 30 calendar days. If rolling 3-year excess return goes negative, the annual letter addresses it head-on. If rolling 5-year excess return is negative, members receive a formal Underperformance Letter stating plainly that discontinuing replication is a reasonable choice. Sustained underperformance approaching ten years triggers a shutdown review. The record — including mistakes — stays permanently visible.