Trend Following Strategy: Full Guide

At a glance

Bets that: A move already underway continues further than expected
Typical win rate: 30 to 40 percent
Payoff: Many small losses, few very large wins
Holding period: Weeks to many months
Best markets: Diversified futures, index ETFs, major FX, large-cap crypto
Worst environment: Range-bound, low-volatility markets for months at a time

Key takeaways

Trend following does not predict. It reacts to price, cuts losers quickly, and holds winners for as long as the trend persists.
The return distribution is extremely skewed: typically fewer than 10 percent of trades produce most of the profit, so skipping trades destroys the edge.
Diversification across many uncorrelated markets is not optional; it is the mechanism that makes the strategy tolerable.
Expect flat or losing stretches lasting one to three years. Every documented trend follower has endured them.
The hard part is behavioural: giving back 30 to 50 percent of an open profit at the end of a trend is a normal, planned event.

What trend following actually is

A trend-following strategy buys instruments that have been rising and sells or shorts instruments that have been falling, then stays in the position until the move reverses by a defined amount. There is no forecast, no valuation, and no opinion about whether the price is justified. The only inputs are price and volatility.

The approach is deliberately simple because its edge does not come from cleverness. It comes from an asymmetric structure: losses are cut at a fixed, small size, while winners are allowed to run indefinitely. That structure turns a roughly random-looking series of outcomes into a positive expectancy, provided you take every signal.

Why trends exist at all

Gradual information diffusion. Large institutions cannot buy a full position in one day without moving the price, so accumulation is spread over weeks. That produces persistent one-directional pressure.
Under-reaction and anchoring. Participants adjust their expectations slowly, anchoring on previous prices and previous forecasts, so price drifts toward the new reality rather than jumping to it.
Risk transfer. Hedgers in commodity and currency markets consistently take one side for business reasons, paying speculators to absorb the other side.
Feedback and constraints. Rising prices trigger buying from index funds, risk-parity rebalancing, short covering, and momentum-based allocations, which extends moves beyond fundamental justification.
Herding and career risk. Professional investors face more career damage from missing a large move than from being wrong alongside everyone else, which amplifies existing trends.

None of these mechanisms is a secret, and trend following has been publicly documented since at least the 1970s. It persists because the strategy is psychologically difficult to run, because it requires diversification most individuals cannot maintain, and because its long flat periods drive capital away at exactly the wrong time.

A complete, testable rule set

The following is a full specification in the style of the classic breakout systems. It is an illustration for study, not a recommendation, and the parameters are conventional rather than optimised.

Universe: A diversified basket of 20 to 40 liquid futures markets across equities, rates, currencies, energy, metals, and agriculture. An ETF-only version can use 10 to 15 liquid ETFs covering the same sectors.
Timeframe: Daily bars. Signals evaluated on the close, orders placed at the next open.
Entry long: Close is the highest close of the past 100 trading days.
Entry short: Close is the lowest close of the past 100 trading days. Omit shorts for a long-only equity version.
Initial stop: Entry price minus 3 x ATR(20) for longs, plus 3 x ATR(20) for shorts.
Trailing exit: Exit long when the close falls below the lowest close of the past 50 days, or when the trailing stop at close minus 3 x ATR(20) is breached, whichever comes first.
Position size: Risk 0.3 to 0.5 percent of equity per position. Units = (Risk% x Equity) / (3 x ATR x point value).
Portfolio limits: Maximum total open risk 6 percent of equity. Maximum 4 positions per sector. Reduce all sizes by half after a 15 percent drawdown.
Rebalancing: Recalculate ATR and position risk weekly; trim positions that have grown beyond twice their intended risk contribution.

Worked example: one full trade

Account equity 100,000 USD, risk 0.4 percent per trade, so 400 USD per position. A gold ETF closes at 208.40, its highest close in 100 days. ATR(20) is 3.10.

Over the following four months the ETF rises to 262. The trailing stop, recalculated daily as close minus 3 ATR, has climbed to roughly 246. A correction then takes price to 244 and the position exits at the next open near 243.80.

Metric	Value
Entry	208.40
Exit	243.80
Gain per share	35.40
Total gain	43 x 35.40 = 1,522 USD
Risk taken (1R)	400 USD
Result in R	+3.8R
Peak open profit	43 x (262 - 208.40) = 2,305 USD
Given back at the exit	783 USD, about 34 percent of the peak

That final row is the part beginners cannot tolerate. Giving back a third of the open profit is not a mistake; it is the cost of the mechanism that lets winners run. Tightening the trailing stop to avoid it also truncates the large trends that pay for everything else, which is exactly why most attempts to "improve" trend following reduce its returns.

What the return distribution looks like

Outcome bucket        Trades    Avg result    Contribution
Loss (stopped out)       58        -0.9R          -52.2R
Small win (0 to 1R)      21        +0.5R          +10.5R
Medium win (1 to 3R)     14        +1.9R          +26.6R
Large win (3 to 8R)       5        +5.1R          +25.5R
Outsized win (8R+)        2       +12.4R          +24.8R
                                                  -------
Total over 100 trades                             +35.2R
Win rate 42%, average win 2.6R, average loss 0.9R

Illustrative distribution of 100 trades from a diversified daily trend system. The shape, not the specific numbers, is what matters.

Two properties follow directly. First, the seven largest trades produced more than the other 93 combined, so skipping signals because one "looks bad" is the fastest way to destroy the strategy. Second, long stretches without a large winner are normal, which is why annual returns are lumpy and why the strategy can lose for a year or more while remaining entirely healthy.

Risk management specifics

Volatility-based sizing is essential. Using ATR means a quiet market and a violent one receive the same risk, which keeps the portfolio balanced as conditions change. See volatility targeting.
Correlation is the hidden exposure. Ten long positions in energy, metals, and index futures during a reflation move are one trade. Cap sector exposure and monitor realised correlation of open positions.
Reduce size in drawdown. Halving risk after a defined drawdown extends survival at a modest cost in recovery speed, and it is far better than the alternative of stopping entirely at the bottom.
Gap risk is real. Stops do not protect against overnight gaps or limit moves in futures. Keep individual position notional modest even when the stop distance is small.
Do not filter signals discretionarily. The moment you start skipping "obviously bad" trends you are trading a different, untested strategy, and you will most likely skip the one that pays for the year.

Variations that are worth testing

Variation	Change	Trade-off
Moving average crossover	Enter when fast MA crosses slow MA	Smoother signals, later entries and exits
Donchian channel	Classic N-day high breakout	Simple and robust; more whipsaws in ranges
Time-series momentum	Long if 12-month return is positive, rebalanced monthly	Far fewer trades, lower costs, less responsive
Dual momentum	Combine absolute and relative strength	Fewer positions, concentration risk
Volatility filter	Skip entries when ATR is in the lowest decile	Fewer false starts, occasional missed large trend
Trend plus carry	Overlay a carry signal on trend direction	Better selection among markets, more complexity

How to backtest trend following honestly

1
Use decades of data across many markets
Trend following makes its return in rare regimes. A ten-year test on five markets is not enough to observe them. Test 20 to 40 markets over 20 to 40 years where possible.
2
Handle futures rolls correctly
Back-adjusted continuous contracts change historical price levels. Percentage-based rules must be applied to the adjustment method you intend to trade. See market data quality.
3
Model costs and slippage on the open
Signals generated on the close and executed at the next open must include the gap between them, plus spread and commission. Optimistic same-bar fills inflate results substantially.
4
Report the full drawdown profile
Maximum drawdown, time to recovery, and longest flat period matter more than annual return. Most people quit because of duration, not depth.
5
Test parameter neighbourhoods
Plot performance across lookback values. Require a plateau. If results collapse when you change 100 days to 90, discard the result.
6
Run a Monte Carlo on trade order
Reshuffling the sequence shows the range of drawdowns the same edge could produce. Size the strategy for the bad draws, not the historical one. See Monte Carlo simulation.

Common mistakes

Trading one market. A single instrument can be trendless for years. Diversification across uncorrelated markets is the strategy, not an enhancement to it.
Taking profits early. Capping winners at 2R while losses stay at 1R converts a positive expectancy into a negative one, because the rare large trends are the entire edge.
Adding filters after a losing streak. Each filter fits the recent past. The strategy that emerges is optimised for the period that already happened.
Sizing by dollars rather than volatility. Equal dollar positions give a volatile market three times the risk of a quiet one, which quietly concentrates the portfolio.
Quitting during a flat period. Historical trend records include multi-year stretches with little progress. Capital that leaves during those periods misses the recovery that follows.

Frequently asked questions

Does trend following still work?

Published trend-following indices and manager records continue to show positive long-run returns with periods of poor performance, notably in the low-volatility years of the 2010s and strong results during sharp regime shifts such as 2008 and 2022. Returns have compressed relative to earlier decades, which is consistent with more capital pursuing the same effect, but the mechanism, gradual information diffusion and risk transfer, has not disappeared.

What is the best moving average for trend following?

There is no best one, and searching for it is a mistake. Lookbacks between roughly 50 and 200 days all capture medium-term trends, and robust systems perform acceptably across that whole range. If your results depend critically on a specific value, the result is noise rather than edge.

How much capital do I need for a diversified trend strategy?

With micro futures contracts, a meaningful diversified portfolio becomes feasible from roughly 25,000 to 50,000 USD. With full-size contracts it is considerably more. An ETF-based version can be run with less, at the cost of fewer uncorrelated markets and no access to some sectors.

Can I trend follow individual stocks?

Yes, and it works reasonably in the long direction with sufficient breadth, typically 20 or more positions. Single stocks carry earnings gaps and idiosyncratic shocks that futures indices do not, so position sizes must be smaller and an index-level trend filter is often used to switch the strategy off in bear markets.

Why not just exit when the trend "obviously" ends?

Because that judgement is only obvious afterwards. Every mechanical exit gives back part of the open profit, and every discretionary attempt to avoid that also exits early from the trends that produce the year’s returns. Backtests of discretionary overrides almost always show lower expectancy than the mechanical rule.

How long are the flat periods?

Historically, one to three years is common for diversified trend followers, and individual markets can go much longer without a usable trend. Your position sizing and your personal finances should both assume that a two-year stretch with no progress will happen at some point.

Test this idea before you trade it

Describe the rules in plain language and AlgoTrader AI turns them into a structured strategy blueprint with a configurable historical backtest, cost assumptions, and exportable code.

Build a backtest

Keep reading

Referenced by

Educational use only. This guide explains how a strategy works. It is not investment advice, not a recommendation, and no result described here is a forecast. Test any approach on historical and out-of-sample data, size positions conservatively, and never risk money you cannot afford to lose.