Correlations worth watching alongside gold

Correlation is context, not confirmation

The temptation with correlated instruments is to treat divergence as a trade signal. In practice correlations are unstable, regime-dependent and frequently resolve in the direction opposite to the one you expected. Their genuine value is in explaining conditions, not in predicting the next move.

The four that earn their screen space

• —Silver — higher beta to the same drivers. Its relative behaviour hints at whether a gold move is monetary or industrial in character.
• —Dollar index — the mechanical denominator. Watch it to separate genuine gold strength from dollar weakness.
• —Real yields — the medium-term anchor discussed elsewhere on this blog, and the most informative of the four.
• —Equity volatility — a fear proxy. Spikes coincide with gold's least persistent rallies.

How we use it

None of these series enter our entry logic. They inform how we interpret a week's results and how we explain conditions to members. Keeping macro context out of the execution path is deliberate: it keeps the system testable, and it keeps our opinions from quietly becoming parameters.

Correlations are conditional, not constant

Every published correlation between gold and another market is an average over a window, and averages conceal exactly the behaviour that matters. Gold's relationship with equities is close to zero on most days and strongly negative during a handful of risk events per year — and it is those days, not the average, that determine whether the relationship is useful.

Measure in regimes, not in aggregate

• —Compute the correlation separately in calm and stressed conditions.
• —Use a rolling window short enough to show the change, long enough to be stable.
• —Treat a relationship that flips sign as information, not as noise to smooth away.
• —Never trade a correlation without knowing which regime produced it.

Using cross-market context without building a second system

The temptation is to promote correlated markets into signal inputs, which usually adds parameters faster than it adds edge. The more durable use is as context: a filter that reduces conviction when the dollar and real yields are both moving hard against the position, rather than a rule that attempts to trade the relationship directly.

Context filters, not signal inputs

Cross-market data is best used to decide how confident to be, not to decide what to do.