What a trend actually is
Before any indicator, any pattern, any system — there is one question worth answering first: what's the trend? Almost every dumb trading mistake a beginner makes can be traced back to fighting the answer.
John Murphy's definition is older than most trading textbooks and still the best one:
An uptrend would be defined as a series of successively higher peaks and troughs; a downtrend is just the opposite, a series of declining peaks and troughs; horizontal peaks and troughs would identify a sideways price trend. — Technical Analysis of the Financial Markets
That's it. Peaks and troughs. Higher highs and higher lows = uptrend. Lower highs and lower lows = downtrend. Anything else is sideways. Every more-complex trend definition you'll ever read is an elaboration on this.
Flip between the three above. Notice the H and L markers — those are the peaks and troughs the definition is talking about. In an uptrend, the pattern is L-H-L-H-L-H with each L above the previous L and each H above the previous H. The moment that stops, the trend is in doubt.
Markets move in three directions
The sideways case matters far more than beginners think. Murphy's estimate:
For at least a third of the time, by a conservative estimate, prices move in a flat, horizontal pattern that is referred to as a trading range.
So a third of the time, there is no trend. That's a problem, because most technical systems — moving averages, breakouts, MACD crossovers — are trend-following by construction. In a range they bleed money from whipsaws. Murphy again: "The failure here lies not with the system. Rather, the failure lies with the trader who is attempting to apply a system designed for trending markets into a nontrending market environment."
The practical consequence: "stand aside" is always a valid decision. Long, short, or out — three choices. If you can't name the trend in plain English, you pick "out."
The three nested trends
Dow Theory (Charles Dow, late 1800s — the grandfather of all this) says a market has three trends going at once, at different scales:
| Trend | Duration | Analogy |
|---|---|---|
| Primary | More than a year, often several | Tide |
| Secondary | 3 weeks to 3 months, retracing ~1/3 to 2/3 of the primary | Wave |
| Minor | Less than 3 weeks | Ripple |
They nest. A strong primary uptrend still contains secondary downtrends — the market doesn't climb in a straight line. A 40% drawdown inside a multi-year bull market is a secondary correction, not a reversal. The hard part, and Murphy says this explicitly, is "being able to distinguish between a normal secondary correction in an existing trend and the first leg of a new trend in the opposite direction."
Most beginners can't tell the difference and don't realize that's the question. Old hands know that's the only question.
Why trends persist at all
They shouldn't, if markets were a random walk. But Kaufman names four reasons they do:
- Fundamentals are slow. Interest-rate regimes, commodity cycles, currency-debasement arcs — all take years to play out. Prices drift with them.
- Persistence. "Some stock price moves defy analysis. They continue to rise beyond any normal assessment of value. Only by staying with the trend could you capture the gains of Apple, Amazon, Tesla, and even Bitcoin."
- Fat tails. "Prices are not normally distributed but have a fat tail... an unusually large number of directional price moves — far longer than would be expected if prices were randomly distributed." Trend-following systems earn their living from those tails.
- Flows. Money chases winners. Rising prices attract more buyers, which raises prices further. The flywheel isn't perfect but it's real.
Murphy puts the same idea in one Newtonian sentence: "A trend in motion will continue in the same direction until it reverses."
Detecting a trend mechanically
Rising peaks and troughs is a clean definition — but what counts as a peak? A 2% wiggle? A 10% move? This is the whole game. Too tight: every bar looks like a pivot and nothing is signal. Too loose: you miss real turns.
Kaufman's swing filter is the classical answer: a new peak is confirmed only when price reverses by at least some percentage from the last extreme. 2–4% is common for equities; more for crypto; less for bonds.
The slider is the signal/noise dial, made visible. Every trend-following system — moving averages, breakouts, ATR stops — has some version of this dial baked in. Tuning it is the difference between catching a real trend and getting chopped to pieces.
Kaufman's simpler alternative: the N-day breakout.
Buy when today's high crosses above the high of the past N days. Sell when today's low crosses below the low of the past N days.
That's it. Donchian made this famous in the 1960s with N = 20 (his 4-Week Rule). Kaufman tested it across markets from 2000–2017 and found the average best N was 93 days — closer to quarterly than monthly. The fact that a rule this dumb works at all is the argument for trend-following existing as a strategy.
The timeframe rule
Same price action looks like three different trends at three different timeframes. A stock can be in a primary uptrend on the weekly, a secondary downtrend on the daily, and a minor uptrend on the hourly — all at the same time. None of those readings is wrong. They're answering different questions.
Murphy's rule, lifted verbatim from Long-term vs. Short-term Charts:
The proper order to follow in chart analysis is to begin with the long range and gradually work to the near term. By starting with the big picture, going back as far as 20 years, all data to be considered are already included in the chart and a proper perspective is achieved. Once the analyst knows where the market is from a longer range perspective, he or she gradually 'zeros in' on the shorter term.
The stack he recommends:
- 20-year monthly — macro context. Where is this market in its cycle?
- 5-year weekly — the primary trend. Up, down, or sideways?
- 6-to-9-month daily — the secondary trend. Where are we inside the primary?
- Intraday — timing the entry, if trading that short.
Long-term charts, Murphy warns, "are not meant for trading purposes... not suitable for the timing of entry and exit points." They set the bias. The daily times the trigger.
The heuristic most professionals actually use: the higher timeframe dictates direction; the lower timeframe dictates entry. Don't short a stock on the 15-minute if the weekly is making new highs — you're fighting a bigger clock.
The uncomfortable bit
Trend-following sounds elegant but the track record has a catch. Dow Theory signals, tested from 1920 through 1975, captured 68% of moves in the Industrials and 67% of moves in the S&P 500. But — and this is Murphy reporting on his own framework — Dow Theory "misses 20 to 25% of a move before generating a signal."
Which means the price of waiting for peak-and-trough confirmation is always entering late and exiting late. Trend-followers never catch the top or the bottom; they catch the middle. Beginners expecting indicators to pinpoint reversals are expecting a thing trend-following by definition does not do.
Accept that or stop calling yourself a trend-follower.
Habits worth forming
- Read the higher timeframe first. Weekly before daily. Always.
- Name the trend in plain English before opening any indicator. "Primary uptrend, secondary pullback, minor downtrend forming." If you can't, the market isn't clear enough to trade.
- Set a swing filter and stick to it. Re-tuning it bar-by-bar is curve-fitting in disguise.
- Treat sideways as a decision, not a problem to solve. Stand aside, or use a different system built for ranges.
Quick check
Using Murphy's definition, what qualifies as an uptrend?
What you now know
- A trend is a structure: rising peaks and troughs (up), falling peaks and troughs (down), or horizontal (sideways).
- Markets are sideways at least a third of the time; trend systems lose money there.
- Three nested trends: primary (over 1yr), secondary (3wk–3mo), minor (under 3wk). They all coexist.
- Detecting trends mechanically requires a swing filter — the signal/noise dial, always a tradeoff.
- Higher timeframe sets direction; lower timeframe sets timing. Start monthly, work down.
- Trend-following catches ~68% of the move and misses the first 20–25%. That's not a bug; it's the definition.