Gold
A weekly model on gold that holds the metal when conditions favor it and steps aside into cash to avoid its deepest drawdowns.
For this audience, drawdown carries equal weight to return. The study’s stated maximum drawdown is −15.31%, measured on daily closes; the chart above plots the month-end series, whose deepest decline is −10.66%. The benchmark’s worst case was far deeper.
Buy-and-hold gold fell close to 45% during the 2012–2015 decline. By stepping aside into cash during sustained downtrends, the model held its worst case to −15.31% — roughly half.
A drawdown curve shows the shape of risk; this shows its worst individual moments. These are the ten months the benchmark fell hardest — the months a client remembers — and what the model did while it happened. Because the decision is weekly rather than monthly, the model can finish a falling month higher than the benchmark.
Gold's upside, without the deepest of its falls.
This model applies the same systematic discipline to gold. Each week it decides whether to hold gold (via the GLD ETF) or move to cash, staying invested when momentum and conditions are favorable and stepping aside during sustained declines.
Gold can deliver powerful returns but is prone to long, painful drawdowns; the model's aim is to participate in the upside while avoiding the deepest of those losses. It offers exposure to an asset that often moves independently of equities and bonds — a diversifying complement within the collection.
Note: because the GLD ETF launched in late 2004, this model's track record begins then — a 21-year hypothetical history rather than the longer record of the equity and fixed-income models.
Everything an auditor would ask for.
Sharpe and Sortino are annualized on a single return frequency — monthly mean and monthly standard deviation, both scaled to annual — net of a 3% risk-free rate. Because the numerator is the arithmetic annualized return, these ratios are not reproducible from the geometric annualized return shown above. Batting average is the share of positive months and is not annualized. Ratios are computed on the record through 31 August 2026; the annualized return and maximum drawdown above are the figures stated in the source study, which ends 29 May 2026.
All performance shown is hypothetical and back-tested — it does not reflect actual trading with client assets and has inherent limitations (designed with the benefit of hindsight; may not reflect the impact of real market conditions). Past performance is not indicative of future results.
Results are gross of fees; transaction / custodial fees and taxes are not reflected and would reduce results. This is not an offer to sell or a solicitation to buy any investment. Charts use representative illustrative data; final disclaimer wording to be confirmed with counsel / compliance.
The strategy is live. The conversation is private.
For qualified investors and institutions. Tell us about your mandate and we'll be in touch.
Get in touch →