Guide

Trend following trading explained

When crude oil rallied 40% in six months during a supply shock, Harbor Capital's discretionary macro desk debated whether the move was “done.” Their systematic trend following futures sleeve had already been long since price cleared a 55-day high with positive 12-month time-series momentum — and stayed long through three false-looking pullbacks because the exit rule had not fired. That single position contributed more to the sleeve's annual return than the next five markets combined. Trend following trading is a rules-based approach that buys assets rising over defined lookbacks and sells (or shorts) assets falling, letting profits run while cutting losers when the trend reverses. It does not forecast turning points; it accepts many small losses and occasional whipsaws in exchange for capturing the fat tails of sustained moves. Managed futures funds and commodity trading advisors (CTAs) have used these systems for decades across equities, bonds, currencies, and commodities. This guide explains time-series versus cross-sectional momentum, common signal rules, volatility-based position sizing, portfolio construction, a Harbor Capital multi-asset futures worked example, a strategy decision table, pitfalls, and a production checklist.

What trend following actually bets on

Markets exhibit persistence: assets that rose recently tend to keep rising (at least for a while), and assets that fell tend to keep falling. Academic literature calls the effect time-series momentum when each asset is compared to its own past, and cross-sectional momentum when you rank assets against each other. Trend following in the CTA sense is primarily time-series: go long soybeans if soybeans are trending up, regardless of whether corn is trending faster.

The economic intuition is behavioral and structural. Underreaction to news, herding, central-bank policy drifts, and slow inventory adjustments can extend moves beyond fair value. Trend followers do not need to know why a trend exists — only that statistically, over hundreds of markets and decades, riding direction has paid, especially during crises when correlations spike and traditional portfolios struggle.

Trend following vs momentum investing

Momentum investing in equity factor parlance usually means cross-sectional ranking: buy last year's winners, sell last year's losers, often in a monthly rebalance with sector constraints. Trend following is more mechanical, more futures-oriented, often daily or weekly, and typically includes explicit stop and position-sizing rules. The philosophies overlap — both bet on persistence — but implementation, instruments, and risk profiles differ. A 12-1 month equity momentum ETF is a cousin of trend following, not a full CTA program.

Common signal rules

Production systems reduce discretion to a small set of testable rules. Three families dominate.

Moving average crossovers

Go long when a fast moving average crosses above a slow one (e.g. 50-day over 200-day “golden cross”), exit on the reverse. Simple and interpretable, but lagging: you enter after a move has started and exit after reversal is underway. Pairs like 20/100 or 50/200 are common; shorter pairs trade more, longer pairs whipsaw less. Moving averages can be simple (SMA), exponential (EMA), or weighted; EMA reacts faster but is noisier.

Breakout / Donchian channels

Popularized by the Turtle Traders: buy when price exceeds the highest high of the last N days (e.g. 20 or 55), sell when price breaks below the lowest low of a shorter window (e.g. 10 days) for a long exit. Breakouts capture acceleration; the trade-off is false breakouts in range-bound markets. Dual Donchian systems (different entry and exit lengths) balance responsiveness and stickiness.

Time-series momentum score

Rank each market by its return over the past 12 months (skipping the most recent month to reduce short-term reversal noise). Go long if the score is positive, short if negative, flat if near zero. This is the backbone of many academic CTA replications and diversifies well across asset classes because signals are independent per market.

Position sizing and risk management

Trend following lives or dies on sizing. A correct signal with oversized leverage blows up on the next gap; undersized positions never pay for decades of small losses.

ATR and volatility targeting

The classic approach sizes each position inversely to volatility, often using Average True Range (ATR): risk a fixed dollar amount (e.g. 0.5% of portfolio equity) per ATR unit. High-vol markets get smaller positions; low-vol markets get larger. Portfolio-level volatility targeting scales gross exposure up or down so realized vol hugs a target (e.g. 10% annualized), similar in spirit to risk parity but applied to directional bets.

Stops, heat, and diversification

Hard stops cap catastrophic loss per trade; trailing stops lock in profits as trends extend. Portfolio heat limits total open risk across all positions (e.g. never more than 6% of equity at risk simultaneously). Diversification across 30–50 uncorrelated-ish markets (equity indices, bonds, FX, energy, metals, agriculturals) is structural: most individual trends lose, but winners arrive unpredictably and pay for the rest.

Harbor Capital multi-asset futures sleeve (worked example)

Harbor Capital runs a $120M systematic futures program as a diversifier alongside equity and credit sleeves. The design:

Universe: 42 liquid futures — 8 equity indices, 6 bond contracts, 10 currencies, 18 commodities.
Signal: Dual Donchian — enter long on 55-day high breakout, enter short on 55-day low; exit long on 20-day low, exit short on 20-day high. Secondary filter: 12-month time-series momentum must agree with direction.
Sizing: 0.4% equity risk per ATR(20); max 2% notional per single market; portfolio vol target 12%.
Rebalance: Signals evaluated daily; positions adjusted at next session open.

In a five-year backtest with realistic slippage and roll costs, the sleeve returned 6.8% annualized with 0.45 correlation to the S&P 500. The best years (2020 energy dislocation, 2022 bond bear) came from commodities and rates trends, not equities. The worst year (-8%) was a choppy, mean-reverting range market where breakouts failed repeatedly — exactly the expected cost of insurance-like crisis exposure.

Harbor's lesson: trend following is not an alpha engine in isolation but a regime diversifier. It loses in sideways years, pays in persistent moves, and often rallies when long-only equity portfolios draw down.

Strategy decision table

Approach	Best when	Weak when	Typical horizon
Trend following (time-series)	Persistent directional moves, crises, macro regime shifts	Range-bound, mean-reverting chop; low vol grinds	Weeks to months per trade
Mean reversion	Stationary spreads, oversold bounces, pairs near equilibrium	Strong trends, structural breaks, momentum regimes	Days to weeks
Cross-sectional momentum	Equity factor premia, sector rotation, relative strength	Momentum crashes, sharp factor reversals	1–12 month ranks
Discretionary macro	Clear narrative catalysts, policy inflection points	Requires skill; hard to scale and replicate	Variable

Pitfalls

Whipsaw clusters — consecutive small losses in sideways markets erode confidence and capital; size for a 50–60% loss rate on individual trades.
Gap and lock-limit risk — futures can gap through stops overnight; agricultural and energy markets have limit-up days that prevent exit.
Correlation spikes in crises — “diversified” trend portfolios sometimes go long everything up or short everything down simultaneously; diversification fails when you need it unless signals genuinely disagree.
Overfitting signal parameters — optimizing Donchian lengths on one decade fits noise; walk-forward and cross-market robustness tests are mandatory.
Ignoring roll and carry — futures P&L includes roll yield (contango/backwardation); a trend in spot may lose money in the contract series.
Capacity and crowding — large CTAs moving the same breakouts can degrade edge; smaller markets suffer first.
Confusing luck with edge — a few big wins drive decade returns; short track records mislead.

Production checklist

Define a liquid, diverse futures universe with explicit inclusion rules (ADV, open interest floors).
Pick signal family (MA crossover, Donchian, TSMOM) and lock parameters before live trading.
Implement ATR or vol-based sizing with per-market and portfolio caps.
Model roll schedules, margin, slippage, and commission in backtests.
Run walk-forward and out-of-sample tests across multiple decades including 2008, 2014 oil, 2020, and 2022.
Set portfolio heat and max gross exposure limits; define vol-target rebalance frequency.
Monitor rolling win rate and average win/loss ratio — edge shows in asymmetry, not hit rate.
Pair with drawdown and Calmar monitoring; expect multi-year flat periods.
Document regime overlap with equity beta; trend following is often a portfolio hedge, not a standalone bet.
Review capacity quarterly as AUM and peer CTA positioning grow.

Key takeaways

Trend following bets on persistence, not prediction — many small losses fund occasional large wins.
Time-series momentum is the CTA core: each market vs its own past, diversified across asset classes.
Breakout and MA rules are simple to implement but require disciplined sizing and realistic whipsaw expectations.
Volatility-based position sizing is not optional — it equalizes risk across oil, bonds, and yen.
The edge is portfolio-level — crisis diversification and fat-tail capture, not high win rates on single trades.