Trend Following: How to Build the System

A trend-following strategy earns from inertia: it enters in the direction of a move already under way and holds while the move continues. It is the most researched class of system and the most awkward to verify on a short history.

What a trend-following system consists of

Three mandatory elements. A definition of trend, usually through the relationship of moving averages or the structure of extremes. An entry rule that fires with the trend rather than against it. And an exit rule that lets the position run far — because all the profit of such a system comes from rare long moves. Remove the third element and take profit early and you destroy the system, leaving it only its losses.

Why a low win rate is normal

Trend systems win in a minority of trades. In our run of an EMA(20/50) stack with an EMA(200) filter the win rate was around 19 percent — expected behaviour rather than a fault. The point is asymmetry: many small losses covered by a few large wins. Hence the requirement most often violated: such a system cannot be judged by win rate and profits cannot be closed early.

What our test showed

We ran the canonical trend-following stack over 518 daily bars across twelve pairs: a median of -1.67%, profitable on 4 of 12 pairs, Sharpe -0.13. A simple Donchian channel breakout on the same data gave a median of -2.56%, profitable on 5 of 12 pairs, Sharpe -0.22. Both sit below simply holding the position (4.79%). But the more important figure is the spread between best and worst pair, which ran to tens of percentage points — and that is the answer to whether the instrument trended.

Why a short history cannot test this

Profit concentrates in rare strong moves. If no pronounced trend occurred on an instrument over two years, the system shows a loss regardless of its quality. That is why classic trend-following approaches are built on portfolios of dozens of markets and evaluated over years: the bet is not that a trend will appear here but that one will appear somewhere. Our single-pair two-year run does not meet that condition, and its negative result is evidence about the sample rather than a verdict on the method.

What breaks the system in practice

Three things, in order of frequency. Taking profit early — it removes precisely the rare large wins the whole construction exists for. Trading in a range without a regime filter: continuation signals there are false by definition. And too tight a stop: a trend needs room for fluctuation inside the move, and a stop sized by intraday logic will remove the position before the move starts.

This material is educational and is not individual investment advice. Trading forex carries the risk of losing capital.