Methodology

Every figure Hedgestone prints should be checkable. This page records where the data comes from, exactly what the engine computes, which premium defaults apply, and where the model honestly falls short.

Data sources

Attributions: LBMA gold PM fix and silver fix courtesy of the London Bullion Market Association. FRED series courtesy of the Federal Reserve Bank of St. Louis (U.S. government data). Deep S&P 500 history via Yahoo Finance. The cash sleeve is a synthetic index: idxt = idxt−1 × (1 + DFFt/100/365). Every dataset carries a version stamp (shown as “data vX” beneath results) so runs are reproducible.

Formulas

Value path

Each asset sleeve starts at initialValue × wi and evolves by its simple daily return rt = pt/pt−1 − 1. Under a rebalancing rule (monthly, quarterly, annual) sleeves are reset to target weights on the first trading day of each period. Assets with no data in the window are excluded and remaining weights renormalized, which is always disclosed on the result.

Purchasing power (CPI deflation)

Vreal(t) = V(t) × CPI(t0) / CPI(t). Real, CPI-deflated values are the default everywhere; nominal paths are one toggle away.

Physical-cost layer

For gold and silver held physically: half the round-trip dealer spread is charged on entry and on every rebalancing trade of the physical sleeve; storage and insurance accrue daily at (storage + insurance) / 252 of the sleeve's value; liquidation value subtracts the exit half-spread. Cumulative costs are reported in currency terms.

Block bootstrap (default forward method)

Random contiguous blocks (default 21 trading days) of the joint daily return matrix — all active assets plus a CPI inflation column — are concatenated until the horizon is filled. Sampling whole blocks preserves cross-asset co-movement, short-run autocorrelation, and the monthly cadence of the CPI column.

Correlated GBM (parametric option)

Geometric Brownian motion with drift and covariance estimated from historical daily log returns; correlation is imposed via Cholesky decomposition. Assumption-heavy by construction and labeled as such in the interface.

All randomness flows from a seeded mulberry32 generator — the same seed, inputs, and data version always reproduce identical results.

Physical premium defaults

Defaults are estimates grounded in publicly listed prices from major dealers (APMEX, JM Bullion), vaulted-storage programs, and ETF expense ratios, rounded conservatively. They vary by product, quantity, and market stress — every field is editable in the builder, and the live defaults served by the API may be revised as dealer data changes.

Limitations

• Commodities and CPI are monthly series forward-filled to the daily grid; intramonth moves are invisible to them.
• Single broad indices per asset class — no survivorship adjustments, dividends handling per the underlying index definition, and no spread between index and any investable product except the modeled physical costs.
• No taxes, no transaction costs on non-physical assets, no borrowing, no yield on physical metal.
• Backtests describe the past; the Monte Carlo describes ranges implied by that same past. Neither predicts the future.
• Everything here is an estimate for education. It is not financial, investment, or tax advice.

Affiliate disclosure

Hedgestone may earn a commission when you follow partner links to dealers or storage providers. Partners never influence the model, the data, the defaults, or the results — the ledger does not bend. There are currently no partner placements live.

Educational estimates — not financial advice