Skip to main content

Trend Following

Updated Jun 24, 2026 at 2:35pm

Research Draft Medium 1,262 words

Trend following is the strategy of buying what is already rising (and shorting what is already falling), holding for as long as the move persists, and exiting only when the trend reverses — never trying to call the top or bottom. Its defining tension is that it deliberately sacrifices a high win rate for an asymmetric payoff: most trades are small losers or break-even, and the strategy's entire profit comes from a minority of large, sustained moves. A trend follower is therefore wrong often and rich occasionally — the opposite emotional profile of mean reversion. The style ranges from multi-month CTA/managed-futures programs to multi-day swing breakouts; the core logic (cut losers fast, let winners run, follow price not opinion) is identical across horizons.

The setups

Trend following on a swing/short-term horizon is built from a handful of price-only mechanics:

  • Moving-average regime filter. Trade long only when price is above a longer MA (commonly the 50- or 200-day) and the shorter MA is above the longer; reverse for shorts. MA crossovers (e.g. 50/200, or faster 10/20 for swing) are the classic on/off switch. The filter answers "which direction am I allowed to trade?" before any entry.
  • Channel / breakout entry (Donchian). Richard Donchian's channel — the highest high and lowest low of the last N bars — is the archetypal trend entry. Go long on a close above the upper band, short below the lower. Swing traders typically use 20–30-bar channels (faster than the months-long CTA versions) to balance responsiveness against false signals (StockCharts/Donchian; QuantifiedStrategies, TrendSpider).
  • The Turtle template. Curtis Faith's published Turtle rules: enter on a 20-day breakout, exit on the opposing 10-day breakout, with an additional 55-day/20-day "slow" system for missed trades. Position size is set so that one N (one ATR, average true range) of adverse move equals a fixed fraction of equity — volatility-normalized sizing is core, not optional.
  • ATR trailing stop. The signature exit. Trail a stop at a multiple of ATR (often ~2–3×) below the highest close since entry. This defines the trend mechanically: as long as price hasn't given back N ATRs, the trend is "intact." The trail ratchets up but never down.

A complete swing setup chains these: regime filter (allowed direction) → breakout or pullback-resumption trigger → volatility-based size → ATR trail until stopped out.

How it's used in practice

The decision-useful core is asymmetry and survival, not entry precision. Practical keys a trend trader actually watches:

  • Enter on confirmation, not anticipation. The trigger is a close beyond a level (new N-day high, MA reclaim), ideally on expanding volume/range. Buying strength feels wrong — that discomfort is the edge.
  • Cut losers immediately, scale into winners. Initial risk is a fixed small fraction of equity (commonly cited ~0.5–1% per Van Tharp / Turtle practice). Add to a position only as it moves your way ("pyramiding"), never average down into a loser.
  • Let the trail do the exiting. No profit targets — targets cap the right tail that pays for everything. You exit when the ATR trail or opposing channel breaks, accepting that you give back some open profit on every winner.
  • Filter the chop. Because range-bound markets are where trend systems "bleed to death" via whipsaws, swing practitioners add a regime gate: only take breakouts when a trend/volatility filter (ADX rising, price extended from a long MA, volatility expanding) confirms a directional environment. In tight ranges, stand aside.
  • Diversify the signal. Even single-stock swing traders run the system across many names/sectors so that the few that trend strongly can carry a basket of small losers — the strategy is statistical, not per-trade.

The #1 misuse is overriding the system on the winners: taking quick profits (killing the right tail) while letting losers run (fattening the left tail) — precisely inverting the edge.

Adoption, debate & evidence

Trend following is one of the few discretionary-folklore ideas with strong, peer-reviewed support — at the portfolio/futures level. Moskowitz, Ooi & Pedersen (2012), Time Series Momentum (Journal of Financial Economics), document that a 12-month-lookback, 1-month-hold signal produced significant positive returns across 58 liquid futures (sample 1985–2009), with the diversified TSMOM factor (scaled to ~12% annual vol) reporting a gross annualized Sharpe greater than 1 — roughly 2.5× that of the passive equity market over the period — and profits in nearly every asset class. (The paper itself frames the figure as ">1"/~2.5× equity; the precise level depends on costs and scaling, so treat single-decimal Sharpe quotes as approximate.) Hurst, Ooi & Pedersen (2017), A Century of Evidence on Trend-Following Investing (AQR), extend support back to 1880 across ~67 markets, documenting low correlation to stocks/bonds, positive skew, and notably strong "crisis alpha" (e.g. trend was positive in 2008).

Honest caveats:

  • The robust evidence is futures/cross-asset, not single-stock swing breakouts. The academic time-series momentum factor is months-long and diversified; a discretionary 20-day stock breakout borrows credibility from it but is not the same instrument and may carry far weaker, more crowded edge after costs.
  • Low win rate is real. Trend systems commonly win ~30–40% of trades (widely cited in practitioner literature, e.g. Covel; QuantifiedStrategies) and rely entirely on payoff ratio. Sub-50% hit rates make the style behaviorally brutal.
  • Long droughts happen. Managed-futures trend CTAs are widely documented to have underperformed equities through much of the 2011–2019 bull market (a period practitioners call the trend "winter"), and institutional trend programs commonly see multi-year, double-digit drawdowns — practitioner sources cite drawdowns in the ~15–25% range lasting 12–24 months, though exact figures vary by program. (We could not independently verify a precise "% of rolling 3-year windows that beat equities" statistic against an accessible primary source, so treat any such single figure as folklore unless sourced.) Separately, recent academic work on trend-premia redundancy (arXiv 2510.23150) questions whether faster trend signals add much incremental edge beyond slow ones — note this is contested: the CFA Institute (2026) argues fast and slow horizons are complementary.

So: the mechanism (cut-fast/run-winners producing convex, positively skewed returns) is well established; the magnitude of edge for any given short-term stock implementation is not, and is highly regime-dependent.

Strengths & limitations

Works when markets trend persistently and volatility expands directionally — strong bull legs, sustained sector rotations, crisis sell-offs (for the short side). Its convexity and crisis-alpha make it a genuine diversifier. Fails when markets are range-bound, mean-reverting, or chop sideways: breakouts fail, the ATR trail gets clipped, and the account bleeds via repeated small losses. It is also psychologically the hardest style to follow — most failures are discretionary tampering, not system flaws.

Sources

Dispute flagged: academic robustness is established for diversified, multi-month futures trend following; the edge of discretionary short-term single-stock trend/breakout trading is weaker, more crowded, and not directly evidenced by those studies.