What Is Trend Following?
Trend following means entering in the direction a market is already moving and holding until that stops. It rests on the premise that price moves persist, which is a measurable property of a series rather than an assumption you have to take on faith.
The oldest systematic approach there is, and one of the few whose central assumption you can actually go and measure.
How it works
Three rules and no forecast:
Enter when the market is already moving — price above a rising average, or a break of a recent high. Hold while it continues. Exit when it stops.
Nothing in that predicts anything. It is a rule for responding to what has happened, which is why it can be written down completely — the test the algorithmic trading page applies to any strategy.
The premise, measured
Trend following bets that a move up is more likely to be followed by another move up. That is a statement about the autocorrelation of price changes, and it is a number you can compute.
Measure it with care, because a rising market can fake it. The S&P 500 closed up on 53.1% of trading days from 1950 to 2026 and finished higher over 59% of ten-day windows.
That gap is drift, not necessarily persistence. A market that rises a little more often than it falls ends most longer windows higher even if each day ignores the one before it.
Read that carefully, because it cuts both ways. A long-only rule on the index profits from the drift with no persistence at all, and a test that ignores this credits the rule for what the index did. It does not say no market ever trends — and the whole industry of trend following exists because on some instruments over some periods it has.
What it does say is that the premise is not obvious, and anybody presenting persistence as self-evident has not measured it.
The same market disagrees with itself
The answer also changes with the timeframe. A market can persist across months and reverse across days, or the other way round.
A negative reading means moves tend to be followed by moves against them, which is the mean-reversion premise, not this one.
Two honest cautions. A slow timeframe gives few observations, so its reading carries a wide error: with 46 of them the ordinary error is about 0.15, and a figure has to sit well beyond that to mean anything. And any one market is one series.
The useful conclusion is modest and real: the regime is a property of the timeframe as much as the market, so “does this market trend” is not a question with one answer.
What the rules cost
You cannot enter before the move exists, because the move existing is the signal.
And you cannot exit at the top, because the top is only visible afterwards. The last stretch of every move is given back before the exit rule can see it. S&P 500 holders saw the long version: from the 9 Oct 2007 close the index fell 56.8% over 17 months, the kind of move an exit rule is built to step aside from part-way down.
Both costs are structural. Any version that entered earlier or exited sooner would do so on weaker evidence and be wrong more often — the trade every page in this glossary keeps arriving at.
Why it might work even if moves do not persist
Worth setting out, because the measurement above looks like it undercuts the whole approach and does not quite.
A strategy does not need to be right often. It needs its wins to be bigger than its losses.
Trend following is built to produce exactly that shape. The exit rule cuts a move that stops quickly and leaves a move that continues alone, so losses are truncated by the rule and wins are not. That asymmetry does not require price changes to be correlated at all — it only requires that the distribution of moves has some large ones in it, which every market has.
So the honest version of the case is not “trends persist”. It is: a rule that exits small moves and holds large ones will, over enough attempts, be carried by the large ones — provided each loss is small enough that you survive to take the next attempt.
Which is why every serious account of this style spends most of its length on position sizing and almost none on entries. The entry is close to arbitrary; the exit rule and the size are the strategy.
A worked example
Define the entry as a rule. Close above a 20-bar average, or a break of a 20-bar high — the Donchian construction.
Define the exit as a rule too, and accept that it will be late.
Size from the stop distance, via risk management, because the next section is about how many of these do not work.
Then take every signal. Selecting among them is where the rule stops being a rule, and the whole approach depends on catching the few that run.
The original data
Across our study of 24,971 trading videos, 132 have trend following in the title. The median one gets 3,608 views, 86% never pass 50,000, and the median length is 15.2 minutes.
The corpus carries description text for only 3 of those 132, which is far too thin to say anything about how the topic is written.
15.3 minutes is among the longer medians here, which suits a subject that is more about holding than about entering.
When it fails
Sideways it is punished
Every crossing is an entry by the rule, and none of them go anywhere.
This is where trend following loses, and it loses slowly and constantly rather than dramatically.
Most signals do nothing
The approach is not built on being right often. It is built on a small number of moves being large enough to cover many small losses, which means a high failure rate is the design, not a fault.
That is only survivable with position sizing that keeps each failure small, and it is the reason this style is inseparable from risk management.
You skipped the signals that looked bad
The few winners cannot be identified in advance. Filtering by how a setup feels removes some of the losers and, on the evidence available at the time, an unknown number of the winners.
You looked once the trend had run
Price above a rising average is the entry condition and also what the top of a range looks like.
Related
Mean reversion is the opposite bet on the same property, and the pair is easier to understand than either alone.
Moving average is the usual way the rule is expressed.
And Supertrend is a trailing exit built for exactly this style.
Whether short runs predict the next day is tested on the market trend page, and on the S&P 500 a three-day run did not.
The analysis is not the hard part: price is above the average, the lows are rising, stay in. The hard part is sitting through the fourth pullback that feels like the top. I accept up front that I will give back the last part of every move, because then giving it back stops feeling like a mistake.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.