Crypto Trend Following: does it work?
Trend following means holding a coin only while its price is rising and stepping aside when it isn't. On crypto it is often pitched as getting most of the upside with far smaller drawdowns.
The rule we tested. Four price-trend rules on the same 18 coins: a graded exposure over three moving averages, a vote across three lookback horizons, a plain four-week momentum filter and a Donchian breakout on Bitcoin, each compared against holding the same coins.
- Crypto trend exposure scaled in over 20/50/100-day averages (18 crypto pairs (BTCUSDT, BNBUSDT, XRPUSDT, ADAUSDT, …)), 2017-08-18 to 2025-03-14, costs 0.10%–0.17% per side: +26.1 pts/yr vs buy-and-hold across 2,069 trades, p=0.177, q=1.000 after correction: beat buy-and-hold on the raw number, but not significantly.
- Crypto 2-of-3 horizon time-series momentum vote (18 crypto pairs (BTCUSDT, BNBUSDT, XRPUSDT, ADAUSDT, …)), 2017-08-18 to 2025-03-14, costs 0.10%–0.17% per side: +25.8 pts/yr vs buy-and-hold across 1,695 trades, p=0.213, q=1.000 after correction: beat buy-and-hold on the raw number, but not significantly.
- Crypto 4-week time-series momentum filter (18 crypto pairs (BTCUSDT, BNBUSDT, XRPUSDT, ADAUSDT, …)), 2017-08-18 to 2025-03-14, costs 0.10%–0.17% per side: +22.8 pts/yr vs buy-and-hold across 1,557 trades, p=0.221, q=1.000 after correction: beat buy-and-hold on the raw number, but not significantly.
- Donchian 20-day breakout on BTC (crypto trend variant) (1 crypto pair (BTCUSDT)), 2017-08-18 to 2025-03-14, costs 0.10% per side: +12.6 pts/yr vs buy-and-hold across 34 trades, p=0.338, q=1.000 after correction: beat buy-and-hold on the raw number, but not significantly.
Not reliably. crypto pairs +26.1 pts/yr vs buy-and-hold (p=0.177, q=1.000); crypto pairs +25.8 pts/yr vs buy-and-hold (p=0.213, q=1.000); crypto pairs +22.8 pts/yr vs buy-and-hold (p=0.221, q=1.000); crypto pair +12.6 pts/yr vs buy-and-hold (p=0.338, q=1.000). None cleared Tickfloor's Benjamini-Hochberg q<0.10 bar once weighed against every other rule tested alongside it.
Stress tests on the discovery data
Same data, harder conditions. Excess is CAGR minus buy-and-hold CAGR.
- Crypto trend exposure scaled in over 20/50/100-day averages: +26.1 pts/yr vs buy-and-hold as tested; +20.8 pts/yr with costs doubled; +22.2 pts/yr with ADA, its biggest contributor, removed; +26.3 pts/yr and +17.6 pts/yr with the main lookback moved down and up by a quarter.
- Crypto 2-of-3 horizon time-series momentum vote: +25.8 pts/yr vs buy-and-hold as tested; +19.1 pts/yr with costs doubled; +22.2 pts/yr with ETC, its biggest contributor, removed; +15.0 pts/yr and +22.8 pts/yr with the main lookback moved down and up by a quarter.
The holdout exam
Data after 2025-03-15 had been kept back. Each rule below was run on it once, with the rule and the pass bar written down first.
- Crypto trend exposure scaled in over 20/50/100-day averages: −6.9 pts/yr vs buy-and-hold from 2025-03-15 to 2026-09-27 (562 days), one-sided p=1.00, Sharpe 0.67 vs 0.66, deepest drawdown 26.3% vs 52.7%. Result: not confirmed.
This is one line in a wider check: Tickfloor's research desk has run 517 strategies against buy-and-hold, and after correcting for how many were tested (Benjamini-Hochberg, 598 tests), 0 survived. A single win-rate claim like this one is exactly the kind of result that check exists to catch before anyone trades on it.
General information only. It doesn't consider your objectives, finances or needs. Tickfloor holds no financial services licence and never places trades.
Does crypto trend following beat holding?
Not reliably. crypto pairs +26.1 pts/yr vs buy-and-hold (p=0.177, q=1.000); crypto pairs +25.8 pts/yr vs buy-and-hold (p=0.213, q=1.000); crypto pairs +22.8 pts/yr vs buy-and-hold (p=0.221, q=1.000); crypto pair +12.6 pts/yr vs buy-and-hold (p=0.338, q=1.000). None cleared Tickfloor's Benjamini-Hochberg q<0.10 bar once weighed against every other rule tested alongside it.
What exact rule did Tickfloor test?
Four price-trend rules on the same 18 coins: a graded exposure over three moving averages, a vote across three lookback horizons, a plain four-week momentum filter and a Donchian breakout on Bitcoin, each compared against holding the same coins.
Did it hold up in stress tests and on later data?
Crypto trend exposure scaled in over 20/50/100-day averages showed +20.8 pts/yr with costs doubled; Crypto 2-of-3 horizon time-series momentum vote showed +19.1 pts/yr with costs doubled. Holdout exam: Crypto trend exposure scaled in over 20/50/100-day averages: −6.9 pts/yr vs buy-and-hold on data after 2025-03-15 (p=1.00), not confirmed. None of this is an edge: Tickfloor's correction for multiple testing left 0 survivors out of 598 tests.
Is this financial advice?
No. This measures a publicly claimed strategy rule, not a recommendation. General information only, not personal advice.
See the full numbers on the Trend following family page, or the full method and every result.
Other rules we've tested: Fear & Greed Timing, VIX Term Structure Timing, Golden Cross / Death Cross.
See a full worked example, free, no account. Or start lesson 1 of the course, also free.