Cross-Asset Momentum

Ask someone to name an anomaly that survives in equities, commodities, currencies, government bonds, and equity indices, at similar strength, over a century of data, and momentum is close to the only honest answer. A commodity that's risen over the past 12 months tends to keep rising over the next one; so does a currency, a bond future, or a stock index that's done the same. That breadth is the whole case for the strategy — and also the thing that should make you suspicious of it, because a pattern that shows up everywhere is either a fundamental feature of how markets absorb information, or evidence that every asset class is secretly driven by the same handful of trend-following funds.

Why "trend" shows up in unrelated markets

Picture information about an asset's true value as a splash of ink dropped into water that's slightly viscous rather than perfectly still. In a frictionless world the ink (the price) would jump instantly to its new equilibrium and stay there — no trend to follow. But real markets have viscosity: news gets priced in gradually because investors are slow to update, positions get built over days or weeks rather than instantly, and risk limits force gradual entry. A central bank hints at a rate cut; some investors act immediately, others wait for confirmation, others only rebalance quarterly. The price drifts toward its new level rather than teleporting, and that drift is what a trend-follower captures — buying the direction the ink is already spreading, in whichever market it happens to be spreading in.

Measuring it the same way across every asset

Time-series momentum, in Moskowitz, Ooi and Pedersen's formulation, is asset-agnostic by design: for any futures contract, look at its excess return over the past 12 months, and take a position scaled by the sign of that return and inversely by the asset's own volatility:

$w_{t} = \text{sign}(r_{t-12,t}) \times \frac{\text{target vol}}{\sigma_t}$

In words: go long if the asset is up over the trailing year, short if it's down, and size the bet so that a jumpy asset (like crude oil) gets a smaller dollar position than a calm one (like a 10-year Treasury future), so each market contributes roughly the same amount of risk to the book.

Worked example 1 — commodities. Suppose WTI crude has returned +18% over the trailing 12 months, with realized daily volatility corresponding to about 30% annualized. Targeting 10% annualized risk contribution per asset, the position size scales to $10\%/30\% = 0.33$ of the notional a "1x" bet would take — you go long, but at a third of full size specifically because oil is so volatile. If the trend continues and crude adds another 5% over the next month, that position contributes roughly $0.33 \times 5\% = 1.65\%$ to the book from this one leg.

Worked example 2 — bonds, the opposite sign. Suppose the 10-year Treasury future has returned $-6\%$ over the trailing 12 months (rates rising, prices falling), with volatility around 6% annualized. Sized to the same 10% target, the position scales to $10\%/6\% \approx 1.67\times$ notional, and because the trailing return was negative, the position is short the bond future. A continued $-2\%$ move over the next month contributes $1.67 \times 2\% = 3.3\%$ to the book — a much larger swing per unit of trailing signal than the commodity leg, precisely because bonds are calmer and so get levered up more to hit the same risk target.

Increase the drift on these simulated paths and watch how a persistent trend, even a small one, produces a visually obvious separation between paths that isn't there under a pure random walk — that visual separation is exactly what a 12-month lookback is trying to detect, in whatever asset class it's pointed at.

What this means in practice

A managed futures fund typically runs this signal across 40–100 futures contracts spanning equity indices, government bonds, currencies, and commodities simultaneously, each sized to contribute similar risk, summed into one book. The core empirical claim — documented across nearly every asset class and back to the 1800s in some studies — is that this diversification is closer to genuine than an equity-only trend book would be, because a rate-driven bond trend and a supply-shock commodity trend are responding to different fundamental drivers even though the statistical signature (sign of trailing return predicts next month's sign) looks identical.

The strategy's real risk is not any single market's trend reversing — it's every trend reversing at once, because the same signal is telling every desk running it to be long the same handful of positions (long equities, short bonds, long the dollar) simultaneously. That happened in early 2018 and again in 2020: a sharp, cross-asset reversal turned "diversified" trend books into one large, correlated loss, because the diversification that held in normal times was itself a product of every asset trending independently — and independence is exactly what breaks in a crisis.

In interviews

Lead with the empirical fact — trend-following works across essentially every liquid futures market, not just equities — because that breadth is the strongest evidence for a real, priced phenomenon rather than a backtest artifact. Walk through the volatility-scaling mechanic with the bond vs. commodity example, since sizing by inverse volatility (not by dollar notional) is the detail interviewers check for. Close by naming the correlation-collapse risk in 2018 and 2020: a book diversified across asset classes in normal times can still take one large, single-direction hit when a macro shock reverses several trends simultaneously.