Trend-following expert advisors and indicators

Trend following tries to catch the long moves a market makes a few times a year and stay in them while they last. A trend-following expert advisor usually loses more trades than it wins, so whether it makes money depends on its few large winners paying for many small losses.

How trend-following EAs trade

Every trend system answers two questions: has a trend started, and has it ended? Most EAs use one of two families of rules:

• Moving-average systems. Buy when a fast average crosses above a slow one, or price closes above a long average, and sell on the reverse.
• Breakout-channel systems. Buy when price makes the highest high of the last 20 or 55 bars, and sell at the lowest low. The exit is often a shorter channel the other way, such as a 10-bar low.

Trend indicators mark direction, colour bars by trend or draw channels, and many traders use them as a higher-timeframe filter for an EA's signals. Check any colour-changing indicator for repainting first; our repainting guide shows how.

Why most trades lose

A trend system takes every signal because it can't know which one starts a long move. Most are stopped out for a small loss; the few real trends are trailed and run to several times the risk. Take a system that wins 35% of its trades, with an average win three times its average loss:

• Each trade expects 0.35 × 3R − 0.65 × 1R = 0.40R, where R is the amount risked.
• Over 100 trades risking $100 each, 35 winners make $10,500 and 65 losers cost $6,500: a net $4,000 before costs, and a profit factor of about 1.6.
• At 3:1 the break-even win rate is 1 ÷ (1 + 3) = 25%. If the win rate slips from 35% to 25%, the edge is gone.

Losing streaks are long too. With a 65% chance of losing each trade, a run of ten losses somewhere in a few hundred trades is an ordinary result of chance. At 1% risk per trade, ten losses in a row take about 9.6% off the balance. The risk of ruin calculator simulates them for your own figures.

Markets and timeframes

The style suits markets that make sustained directional moves, such as gold, the yen crosses and stock indices. Many traders run the same rules on several symbols, since trends arrive at different times, but pairs such as EURUSD and GBPUSD often trend, and lose, together.

Most trend EAs work on H1 to D1. Lower timeframes give more signals, but more are noise and costs take a bigger share of each move. Higher timeframes trade rarely and hold for weeks, paying or earning swap every night. Session timing matters less than for intraday styles.

When it works and when it fails

Trend following earns its money in a few strong, persistent moves. Its worst periods are ranges, where price swings back and forth through every average and short channel. Each false signal costs 1R plus spread, and a range lasting a few months can produce a long string of them. That drawdown is the cost of being in the market when the next trend starts.

Sharp reversals hurt too, giving back much of an open profit before the trailing stop is hit. Range filters, such as a minimum trend-strength reading, cut some losers but delay or miss the start of some trends, so test with and without them.

Settings that matter

• Lookback. The average periods or channel length. Shorter means earlier entries and more false signals.
• Initial stop. Often a multiple of the average true range (ATR), so it widens in fast markets and tightens in quiet ones.
• Trailing stop. The exit that decides how much of a trend you keep: a multiple of ATR behind price, a moving average or the opposite channel. A tight trail gets shaken out of long trends; a loose one gives back more at the end.
• Position size. On EURUSD, 0.20 lots is worth $2 per pip, so a 50-pip stop risks $100 and a trailed winner of 150 pips makes $300: the 3:1 above.

How to test it honestly

With profit concentrated in a few trades, a two-year backtest may hold only one or two of the moves that make a trend system worthwhile. Test across many years and several symbols, ranges included:

• Check how much of the net profit comes from the five largest trades. If two trades carry a ten-year test, the sample is too small.
• Look for a plateau, not a peak. If a 50-period average works and 45 and 55 don't, the result is probably luck.
• Hold back the most recent years as an out-of-sample test.

Checks before running one

1. Know the losing streak. Plan for a longer one than the backtest shows, and price it with the drawdown calculator.
2. Check the reward to risk. Enter the average win and loss in the risk reward calculator to see the break-even win rate.
3. Test long and wide. Follow how to backtest an EA over many years and symbols.
4. Size from the stop. Use the position size calculator, and cap risk across correlated symbols as in risk management for EA trading.
5. Decide when you would stop. Write down, in advance, the drawdown that would make you pause the EA.

For the full checklist, read how to evaluate an expert advisor.