Why Trend Diversifies Better Across Assets

A trend-following strategy on a single market — say, only the S&P 500 — is really a bet on one recurring pattern: markets sometimes drift persistently in one direction, and buying into a rise (or selling into a fall) captures that drift. Run that same strategy simultaneously on forty or fifty different markets across equities, bonds, currencies and commodities, and something changes beyond simply "more markets": the trades themselves stop moving in lockstep with each other, in a way that a trend book confined to one asset class never can.

Independent trends, not independent assets

It is tempting to think cross-asset trend diversification is just standard portfolio diversification — hold uncorrelated assets, get a smoother portfolio. It is a subtler and, in this case, more favorable effect: trend-following isn't a bet on the level of any asset, it's a bet on the persistence of direction in that asset, and persistence in oil prices (driven by OPEC supply decisions, say) has essentially nothing to do with persistence in the Japanese yen (driven by Bank of Japan policy) or in 10-year Treasury yields (driven by the Fed's rate path). Because the underlying catalysts for each trend are unrelated, the correlation between one trend trade's P&L and another's tends to run lower than the correlation between the underlying asset prices themselves.

Drag the correlation down toward zero in the scatter above — that is roughly what a well-built cross-asset trend book is aiming for between its individual market bets: not zero-correlated assets, but close to zero-correlated signals.

Worked example

A trend program runs fifty independent market trades, each targeting the same volatility and each with an assumed Sharpe ratio of 0.4 as a standalone bet, and a rough approximation that the pairwise correlation between any two trend signals is low, around 0.05 given how different the drivers are. Using the standard effective-breadth approximation for the Sharpe ratio of an equal-weighted combination of $N$ signals with average pairwise correlation $\rho$:

$$SR_{portfolio} \approx SR_{single} \times \sqrt{\frac{N}{1 + (N-1)\rho}}$$

Plugging in $N = 50$ and $\rho = 0.05$: the denominator is $1 + 49 \times 0.05 = 3.45$, so $\sqrt{50 / 3.45} \approx \sqrt{14.5} \approx 3.8$. The combined Sharpe ratio is approximately $0.4 \times 3.8 = 1.52$ — nearly four times the standalone Sharpe of any one market, purely from combining many weakly correlated trend bets rather than from any single trade being better.

What this means in practice

This is the core justification for running a managed-futures or CTA-style program across dozens of instruments rather than concentrating in a handful of the most "trendy-looking" markets: breadth itself is the source of return quality, not just diversification of risk. It is also why these programs tend to perform best in exactly the periods equity-only strategies struggle — a sharp equity selloff often coincides with strong trends in bonds, currencies, or commodities that a single-asset trend book would never capture.