Carry and Trend Blend

Thesis (edge)

These two strategies have almost opposite personalities, and that is exactly why they belong together.

Carry is the tortoise. It collects a small return most months, it looks wonderfully stable, and then a shock arrives and it gives back years of gains in a few weeks. Its return distribution has many small gains and a few enormous losses.

Trend is the opposite. It bleeds small amounts in quiet, choppy markets and it makes its money in the big sustained moves, which is to say, in the shocks. Its distribution has many small losses and a few enormous gains.

Put them in one book and you get something better than either. When a crisis hits, carry is losing and trend is winning. When markets are calm, trend is bleeding and carry is grinding out gains. Neither is a perfect hedge for the other, but the correlation between their returns is low and often negative in the tails, which is the most valuable kind of diversification you can get.

Where it works (regimes)

The blend does its best work over full cycles rather than in any single regime. It gives up some of carry's smooth returns in the good years and some of trend's spectacular crisis payoffs, and in exchange it gives you a return stream you can actually hold on to without losing your nerve.

The regime where the blend struggles is a choppy market with no direction and no clear carry signal. Trend gets whipsawed, carry offers nothing worth collecting, and the book drifts sideways while paying costs. This is annoying rather than dangerous.

The regime that genuinely hurts is a sharp reversal after a period where both signals agreed. If trend was long crude and carry also liked crude because the curve was backwardated, you were doubly long, and a supply announcement that collapses the price hits you twice.

Signals

• Trend: any reasonable signal, for example a blend of moving average crossovers or a 12-month time-series momentum score, expressed as a number between fully short and fully long.
• Carry: the annualized curve slope, normalized by the market's volatility, expressed on the same scale.
• Both signals must be put in comparable units before combining. If your trend signal ranges from minus two to plus two and your carry signal ranges from minus 0.1 to plus 0.1, a naive sum is just the trend signal with a rounding error attached.
• Combine with fixed weights. Equal weight is the honest starting point and is very hard to beat out of sample.

Portfolio construction

The interesting design choice is what to do when the signals disagree. Three reasonable answers exist.

The simplest is to net them: add the two signals and trade the result. If trend says long and carry says short with equal conviction, you end up flat.

The second is to gate: only take a position when both agree, and stay flat otherwise. This trades less, has a higher hit rate, and gives up a lot of return because it sits out too often.

The third is to use trend as an override on carry only. Carry sets the position, and trend can veto it when it points strongly the other way. This is the classic fix for the value trap problem in carry, where you collect interest on a currency that is quietly devaluing.

We prefer netting with a cap on the combined position, so that agreement does not silently double your risk.

Risk model

The specific danger is correlated conviction. When both signals are confident and pointing the same way across many markets, the book becomes concentrated in a single macro theme, and your position sizing model may not notice because each individual position looks reasonable.

Monitor the realised correlation between the two sleeves. It is normally low. If it starts rising, the diversification you are counting on is disappearing, and the total book risk should come down.

Costs & implementation

The blend is cheaper to run than trend alone, because the carry signal is slow and stable and it damps the trend signal's flip-flopping. That is a real and underappreciated benefit: adding a slow signal to a fast one reduces turnover, not just risk.

Use a no-trade band on the combined target. Because two continuously varying signals are being summed, the target position is always drifting a little, and rebalancing to it exactly would generate pointless turnover.

Failure modes

• Optimizing the blend weights on historical data, which is the most obvious way to ruin a genuinely robust idea.
• Double-sizing when both signals agree.
• Failing to standardize the two signals so that one silently dominates.
• Assuming the diversification benefit is permanent. It is a statistical tendency, not a law.
• Comparing the blend only against carry, which makes it look worse in calm years, or only against trend, which makes it look worse in crisis years. It should be compared against both.

Our Notes & Suggestions

This is, in our view, the most sensible default construction for a systematic futures programme. Neither sleeve is exotic, both have decades of evidence behind them across many markets, and their combination addresses the specific weakness of each.

Keep the blend weights fixed and simple. The temptation to make them dynamic, to lean into carry when volatility is low and into trend when it is high, is very strong and is almost always a route to overfitting. If you must do it, use a rule so simple you could explain it to a client in one sentence, and check that it works with the rule turned off too.

Report all three equity curves: carry alone, trend alone, and the blend. The value of the blend is only visible in the comparison, and it is usually visible as a shallower worst drawdown rather than as a higher return.