Guides · Guide 07
Trend following in crypto: what it is, when it works, when it doesn't
Trend following does not predict. It sizes into moves that have started and out of moves that have ended, and accepts many small losses for a few large gains.
- Backtest CAGR
- 41.8%
- Weakest year (2025)
- +6.1%
- 2022 bear market
- +32.5%
The idea
Trend following buys what has been rising and sells short what has been falling, holds while the move continues, and exits when it stops. It makes no forecast. It relies on one empirical regularity: large moves in most markets, and especially in crypto, have tended to unfold over weeks and months rather than in a single jump, so a strategy that enters after the move has started and exits after it has ended captures a large share of it.
The cost of that regularity is false starts. Most trends fail early, and each failure is a small loss. The strategy's economics are a long series of small losses paid for by a few large gains: in our backtest, 40% of months lost money, and the return came from years like 2021 and 2023 rather than from a steady drip.
Why it suits crypto
- Crypto trends are long and large. Bitcoin's moves of 2021, 2022 and 2024 each lasted the better part of a year. A rule that needs weeks to confirm a trend still catches most of it.
- Both directions are tradeable. Perpetual futures make shorting as easy as buying. In 2022, with the ten largest coins down 52–94%, a long/short trend book gained 32.5% net (Note 05).
- Volatility is high enough to matter and measurable enough to size. Sizing positions inversely to volatility keeps risk stable while the coins swing 50–100% a year.
When it fails
Trend following loses money in markets that move sideways with sharp reversals: the strategy enters on a move, the move reverses, it exits at a loss, repeatedly. It also lags at turning points by design, giving back part of every trend's end. The backtest shows both:
Net return by year, backtest, with and without Binance's historical funding. 2026 to 27 September.
2025 made 6.1% net: a year of range-bound chop with no sustained move to ride. 2021 made 101% net: one long trend. The difference between those years is not skill; it is the market handing the strategy what it needs or not. Anyone allocating to trend following must accept years like 2025 as the price of years like 2021, and must not be able to tell in advance which one is coming.
What it looks like against holding Bitcoin
| Jan 2021 – Sep 2026 | Trend book (backtest) | Bitcoin, held |
|---|---|---|
| CAGR | 41.8% | 20.2% |
| Volatility | 24.3% | 57.7% |
| Sharpe | 1.56 | 0.61 |
| Max drawdown | −18.8% | −77.1% |
| Correlation to BTC, monthly | 0.26 | 1.00 |
The correlation figure is the one to notice. A monthly correlation of 0.26 to Bitcoin means the book is not a leveraged bet on crypto going up; it made money in the year crypto went down. The drawdown figure is the other: the strategy's worst fall was a quarter of Bitcoin's.
What separates implementations
The concept is public and old; the results differ by execution. What decides them: how trend strength is measured and across which horizons, how positions are sized, how exits are placed, how costs are controlled, and whether the operator sticks to the rules through a year like 2025. Our own rules are not published, but every result of them is: the live record, the research notes and the approach.
All strategy figures are hypothetical backtest results, net of 0.07% per trade and Binance's historical funding, computed with hindsight. They are an upper bound, not a forecast.