# Allocator strategy series: chart inputs

Checked on 27 October 2026.
Each CSV records values, units, measurement basis, reporting period, source URL and locator for one chart in the allocator strategy series.
They describe disclosed or published figures, not a harmonised dataset for ranking managers.
13F weights are shares of reported 13F value, which excludes shorts, cash, bonds, futures, currencies, most non-US holdings and private investments.
No chart is a recommended allocation or a forecast.
Chart specifications and source hashes: `docs/research/allocator-series/charts/`.

## Returns, volatility and drawdowns against MSCI World

Checked on 30 October 2026.
Files named `<series>-vs-msci-world-monthly.csv` hold monthly total returns for an investor's fund (or a labelled hypothetical model), MSCI World Net Total Return in the same currency, and the same-currency cash return, from the first full month available to September 2026.
`scripts/allocator-sharpe.py` downloads the inputs and computes the Sharpe and Sortino ratios, CAGRs, volatilities and drawdowns quoted in the articles; the summary is `docs/research/allocator-series/sharpe-vs-msci-world.json`.
Fund returns are daily closing prices with cash distributions reinvested on the ex-date, net of fund expenses but before taxes and trading costs; the index has no costs.
One distribution in the price feed that the issuer's own returns show was never paid (ARKK, 8 September 2023) is excluded.
Cash is the Kenneth French 1-month Treasury bill return (USD) or the Bank of England SONIA monthly average compounded to a monthly rate (GBP).
These are historical, window-dependent statistics, not forecasts or rankings of managers.

## Second wave

Checked on 3 November 2026.
Files prefixed `munger-`, `graham-`, `burry-`, `greenblatt-`, `taleb-`, `lynch-`, `gpif-`, `cpp-`, `adia-` and `harvard-` hold the chart inputs for the ten studies added that day.
Each file's source, locator and basis are repeated in the matching chart specification, and the article's source ledger lists the document hashes.
Labelled scenarios (the Taleb barbell write-off case and the Graham 50/50 backtest) are our calculations from the named public datasets, not forecasts.