TL;DR: A BTC trend-following strategy has three parts: a rule that detects a trend (a moving-average cross, a Donchian breakout or a SuperTrend flip), an optional filter that confirms the trend is strong enough, and an exit that lets winners run while cutting losers, usually an ATR trailing stop. Build it on 4H or 1D bars, backtest it net of slippage and funding, and judge it on drawdown and trade count as well as return. This guide walks through each step in Trigr's no-code Studio.
What is trend following, and why try it on BTC?
Trend following buys markets that have been rising and sells or avoids markets that have been falling, then holds until the trend ends. It does not try to predict tops or bottoms. It accepts many small losses in exchange for occasional large gains when a trend persists.
The idea has serious academic backing. Moskowitz, Ooi and Pedersen documented time series momentum across dozens of futures markets (Journal of Financial Economics, 2012): past 12-month returns tended to predict future returns in the same market. Bitcoin has had long directional runs, which makes it a natural candidate.
It is worth being clear-eyed about the trade-offs before building:
- Low win rates are normal. Many trend systems win on fewer than half their trades. The edge, if any, is in the size of winners.
- Chop is expensive. In sideways markets, every breakout fails and each failure costs a fee and a stop.
- Drawdowns are long. A strategy can be flat or underwater for months between trends.
- Much of the return can be market beta. A long-only BTC trend system in a bull market may mostly be holding bitcoin with extra steps. Compare it with simply holding.
Step 1: How do you choose a trend trigger?
The TRIGGER node fires the entry. Every Trigr strategy has exactly one. Three standard choices, all available as indicator nodes in Studio:
| Trigger | Signal | Studio parameters | Character |
|---|---|---|---|
| Moving-average cross | Fast MA crosses above the slow MA | Fast 2–200, slow 5–400 (default 50 × 200) | Slow, few trades, late entries |
| Donchian breakout | Price makes a new N-bar high | Length 2–400 (default 20) | Enters early on breakouts, more false starts |
| SuperTrend flip | ATR-based band flips to long | ATR period (default 10), multiplier (default 3) | Adapts to volatility; built-in trailing logic |
A few practical notes:
- Pick one and keep it. Testing all three and keeping the best is fine as research, but each variant is a trial. Record them; see the section on experiments below.
- Longer lookbacks trade less. A 50 × 200 cross on daily bars might fire only a few times a year. That keeps costs low but gives you a small sample of trades.
- Match the trigger to the timeframe. A 20-bar Donchian breakout means 20 days on 1D bars and about 3 days on 4H bars. Those are different strategies.
For BTC, a reasonable first build is a 4H Donchian breakout (new 20-bar high) or a 1D 50 × 200 moving-average cross. The step-by-step logic of the node graph is covered in our pillar on building a no-code trading strategy.
Step 2: Should you add a trend-strength filter?
FILTER nodes are extra conditions that must also hold for the entry to fire. For trend following, filters try to skip the choppy periods where breakouts fail.
Common options:
- ADX above a threshold. The Average Directional Index measures trend strength regardless of direction. Studio's ADX node defaults to length 14 and a threshold of 25. Requiring ADX > 25 skips entries in directionless markets.
- Price above a long moving average. For a long-only version, require the close to be above a 200-bar EMA or SMA so you only buy breakouts in a broader uptrend.
- A funding-rate ceiling. On perps, extremely positive funding means crowded longs who pay to hold. A filter that skips longs when funding is unusually high can avoid late, crowded entries. See funding-rate and open-interest strategies for how to use the data.
Every filter removes trades. A strategy with five filters might look excellent on 12 trades and mean nothing. Add filters one at a time and check that each one improves the result for a reason you can state, not just the number.
If you want a higher-timeframe confirmation, such as trading 4H signals only when the daily trend agrees, make sure the daily value comes from a completed daily bar. Our guide to multi-timeframe strategies without look-ahead bias explains why this matters.
Step 3: How should a trend strategy exit?
For trend following, the exit decides more than the entry. The goal is asymmetric: let winners run, cut losers quickly. The RISK node in Studio offers the relevant tools:
- ATR trailing stop. The stop trails price at a multiple of the Average True Range. It widens in volatile markets and tightens in quiet ones. This is the classic trend exit, and it is usually the one to start with.
- Exit on signal flip. Close the position when the entry signal reverses (for example, the fast MA crosses back below the slow). This is on by default in Studio.
- Stop-loss. A fixed percentage stop, between 0.2% and 50%, caps the loss on a failed breakout.
- Take-profit. Usually left off or set wide for trend following, since a tight target caps exactly the large winners the strategy depends on.
- Time stop. A maximum number of bars in a trade. It can help a breakout system exit trades that go nowhere.
Sizing and leverage deserve care. Trend systems have long losing streaks, so a size that feels comfortable per trade can compound into a deep drawdown. Start at low leverage and a modest percentage of capital. The risk settings guide goes through sizing, ATR trails and time stops in detail.
Finally, set the direction: long only, short only or both. With both, the opposite-signal setting decides whether a short signal reverses a long position, is ignored, or can only reduce it. Test long-only and both-sides versions separately; on BTC they often behave very differently.
Step 4: How do you backtest it honestly?
A trend strategy that trades rarely is sensitive to every assumption. Three matter most on Trigr:
- Fills at the next bar's open. A signal on the 4H close is filled at the next 4H open, not at the close that triggered it. Breakout strategies often look worse under this rule, because price has already moved. See next-bar-open fills.
- Fees are always applied. Trading and builder fees come off every fill.
- Slippage and funding are opt-in. A first result is labelled gross. Turn on a flat slippage assumption in basis points per fill and turn on funding. Historical funding comes from Binance's 8-hour USDT-M series (Hyperliquid itself pays hourly; see Hyperliquid's funding documentation). For a long-biased trend system, funding can be a real cost in bull markets, when longs usually pay.
Everything runs on point-in-time data, and when a take-profit and stop-loss are both touched in the same bar, the engine replays 5-minute bars to see which came first; if it is still ambiguous, the stop wins. The backtesting docs list what is and is not modeled.
What should you read in the results?
| What to check | Why it matters for trend following |
|---|---|
| Net return vs holding BTC | Separates a trend edge from market exposure |
| Maximum drawdown and its length | Trend systems can sit underwater for months |
| Number of trades | Under a few dozen trades, results are fragile |
| Win rate and average win / average loss | Low win rates are fine only if winners are much larger |
| Monthly returns | Shows whether returns come from one or two trends |
| Trade log | Confirms entries and exits happen where you expect |
A standard backtest does not report a Deflated Sharpe Ratio or PBO; those come from an ML optimization run. The trailing 25% shading in Studio is a recent-period diagnostic, not an out-of-sample holdout.
Step 5: How do you iterate without fooling yourself?
You will want to try other lookbacks, filters and trail widths. Do it as labelled experiments inside one strategy rather than as dozens of saved near-duplicates. In Studio, the Copilot iterates as Experiment 1, 2, 3 in the strategy's version history, and any version can be restored. Over MCP, backtests can carry an experimentLabel on the same strategy.
Keeping every variant in one history keeps the number of trials visible, which is what you need to judge whether the best variant is real or lucky. The approach is explained in iterating on a strategy with AI experiments.
What this means for you
You can build a complete trend system, including ATR trailing exits and funding-aware backtests, without writing code, and see a net result that reflects next-bar fills rather than an idealized one. When you settle on a version, freeze it and run it on a paper agent before considering real capital.
Backtests are not guarantees; perps are leveraged and can lose more than expected.
Next steps
Build the first version in Studio with a single trigger and an ATR trail, then compare it with holding BTC. If you would rather start from an existing strategy, browse verified, versioned strategies on the strategy marketplace.