BUILD: COURSE 3 | LESSON 2
Trend-following strategy walkthrough
Learning objectives
Assemble a complete trend-pullback strategy using the P3.1 anatomy: hypothesis, regime filter, setup, trigger, exits, risk.
Describe the characteristic performance profile of trend-following — low win rate, large winners, long flat-to-down stretches — and why most people can't sit through it.
Take the example spec onto a demo account and begin generating your own evidence about it.
This is the first of three strategy walkthroughs, and we start with trend-following because it has the longest public track record of the three ideas (it powered entire fund categories for decades) and the performance profile retail traders find hardest to live with. Everything below follows the P3.1 blueprint. And the framing from that lesson applies in full force: this is a teaching skeleton, not a system we're claiming is profitable as written. The parameters are round, illustrative numbers. Your job is to understand why each part exists, then test your own version on demo.
The hypothesis, and what it predicts
Hypothesis: markets trend more persistently than participants expect — after a directional move establishes itself, hesitation, disbelief and gradual position-building cause continuation more often than immediate full reversal. Buying temporary pullbacks within an established uptrend therefore gets continuation exposure at better prices than chasing.
What this predicts — and you should hold the strategy to its own predictions:
- It should make its money in a minority of trades that run a long way (+3R, +6R), while most trades stop out or scratch.
- It should bleed steadily in ranging, news-chopped markets — the regime where "pullback" and "reversal" are indistinguishable.
- Its equity curve should look like long shallow declines punctuated by sharp climbs. If you need frequent wins to stay motivated, you now know in advance this family will hurt.
Typical honest numbers for trend systems land near 30–45% win rates with average winners 2–4× average losers. Run that through P2.1: at 35% wins × 3R average win, expectancy = (0.35 × 3) − (0.65 × 1) = +0.40R — genuinely good if achieved — but P2.3 says a 35% win rate should plan for 10–12 loss streaks. That is the whole deal of trend-following in two numbers.
The example spec
- Markets & timeframe: EUR/USD and GBP/USD (pick ONE to start — they're one correlation group per P2.5); D1 for regime, H4 for setup and trigger.
- Regime filter: D1 close above the D1 200 SMA, and the H4 50 EMA rising. Both must hold. (Justification: the hypothesis is about established trends; this is the "established" test. Longs only for the walkthrough — mirror everything for shorts.)
- Setup: price pulls back to within 0.5 × ATR(14, H4) of the H4 20 EMA, without closing below the 50 EMA. The pullback must contain at least 2 candles (one red candle isn't a pullback, it's a blink).
- Entry trigger: an H4 candle closes above the high of the previous candle while setup conditions hold. (A close-based trigger, so it's cheap to journal and impossible to argue with.)
- Initial stop: 2 pips below the lowest low of the pullback, and no closer than 1.5 × ATR(14, H4) from entry — whichever is wider (P2.2: structure first, ATR as the sanity floor).
- Exits: no fixed target. Trail the stop below each completed H4 higher low once the trade reaches +1R; exit at market if an H4 candle closes below the 50 EMA. Time exit: if the trade hasn't reached +1R within 15 H4 bars (~2.5 trading days), close it — the continuation thesis has gone stale.
- Filters (the full allowance of three): the regime filter above; a calendar filter — no new entries within 4 hours of high-impact events for the pair's currencies; a volatility floor — skip if H4 ATR is in its lowest 20th percentile of the past 3 months (spread would eat too much of R, per P2.2's cost maths).
- Risk: 1% per trade, max 1 open position for this strategy, bench the strategy for review at −8R from its equity high.
Notice how little of the spec is the entry. The trailing exit is where trend systems earn their big R-multiples; the time exit and volatility floor quietly remove the two most common expectancy leaks (dead trades and untradeably quiet conditions).
Walking through both outcomes
A winner, beat by beat. Regime: EUR/USD D1 above the 200 SMA since last month; H4 50 EMA rising. Price rallies to 1.0920, then pulls back over four H4 candles to 1.0862 — inside 0.5 × ATR (ATR = 22 pips) of the 20 EMA at 1.0870, never closing below the 50 EMA. Setup valid. Next candle closes at 1.0881, above the prior candle's high of 1.0875 → enter long 1.0881. Pullback low 1.0858 → structural stop 1.0856 (25 pips; 1.5 × ATR = 33 pips is wider, so the stop goes at 1.0848, 33 pips). Account $10,000, risk $100 → 0.30 lots (100 ÷ (33 × $10) = 0.303, round down). Price makes higher lows over three days; the trail steps up behind each one; a news spike finally closes an H4 candle below the 50 EMA at 1.0985 → exit. +104 pips ≈ +3.1R. The point to absorb: after entry, every decision was mechanical.
A loser, beat by beat. Same regime two weeks later. Pullback to the 20 EMA, valid trigger at 1.1010, stop at 1.0975 (35 pips). Price drifts up 15 pips, stalls, and a US data surprise drives an H4 close down through the pullback low; stopped at 1.0975. −1R, $100. Nothing was wrong. This is 55–70% of this strategy's trades. The failure mode that kills trend traders is not this loss — it's the response to five of them in a row: skipping the sixth signal (which is the +4R one), tightening stops into the noise band, or "just taking profits at +1R for a while", which deletes exactly the tail that pays for everything (P2.4's arithmetic).
What we honestly don't know — and how you'll find out
We have not told you this spec's expectancy on EUR/USD in 2026, because we'd be inventing it: performance of any fixed rule set drifts across instruments, years and volatility regimes, and a curriculum that printed a glossy backtest here would be teaching you the wrong epistemics. What decades of public research on trend-following do support is weaker and more useful: the family has a real behavioural basis, its profile is exactly the low-win-rate/long-drawdown shape described above, and naive versions have had multi-year flat periods. Whether your version, on your instrument, executed by you, has positive expectancy after costs is an empirical question. Demo is where it gets answered — take the task below seriously, and Lesson 5 will turn your logs into a verdict.
Key takeaways
Trend-pullback anatomy: established-trend regime filter → pullback-to-EMA setup → close-based trigger → structure/ATR stop → trailing exit that lets a minority of trades run to large R-multiples.
The profile is the price of admission: ~30–45% win rates, planned-for 10+ loss streaks, and equity curves that climb in bursts. Expectancy lives in the tail winners the trailing exit preserves.
The entry is the least important component; the trail, time exit and volatility floor do most of the expectancy work.
The most dangerous person to a trend system is its own trader during a normal losing streak — skipped signals and early profit-taking amputate the paying tail.
These rules are a teaching skeleton with illustrative parameters, not a profitable system to copy. Its edge on your market is unknown until you test it — on demo, then in Lesson 5's framework.