Do trading strategies work on crypto?
None of the technical indicators and other rules passed the corrected test in these backtests. Of 510 strategies tested on crypto pairs, 94 finished ahead of the equal-weight benchmark on the point estimate, 13 of those with an unadjusted p-value under 0.05, and none met the corrected Benjamini-Hochberg bar (q below 0.10). The median strategy returned −50.6 pts/yr against the benchmark. These are results from our older test setup, which was built before our current testing rules. They have not been redone under those rules.
The universe is crypto pairs only, with data up to 15 March 2025. Technical analysis here means rules built from price and volume, such as moving averages, RSI and MACD; the technical indicators were tested alongside other rules, such as seasonality. Most strategies are published or popular rules coded from a source description (a few are Tickfloor research-desk hypotheses), charged modelled trading costs and scored against an equal-weight benchmark of the same assets.
The numbers
- 510 strategies tested on crypto pairs.
- 94 finished ahead of the benchmark on the point estimate, 13 with an unadjusted p-value under 0.05.
- None met the corrected bar (q below 0.10).
- The median strategy returned −50.6 pts/yr against the benchmark.
Research map: crypto pairs
Each family's size shows how many strategies we tested, and its colours show how each one did. Pick a family to open its page.
- Edge excluded 77
- Behind, not excluded 202
- Behind, one good half 125
- Too few trades 12
- Ahead, not significant 94
Families by number of strategies; bar colours are the outcomes above.
- Trend following83
- Momentum68
- Mean reversion57
- Volatility38
- Sentiment extremes25
- Equity factors24
- Other tested patterns23
- Social media chart patterns22
- Top 100 retail strategies17
- Funding carry15
- Trending techniques14
- Day and month seasonality13
All 51 families
- Popular rules, strongest version12
- Positioning10
- Popular indicator rules9
- Chart structure8
- Macro7
- Machine learning5
- Calendar-event drift4
- Open interest vs spot4
- TJR core model4
- Event clock3
- Liquidity sweep3
- Pairs and stat arb3
- Anchor occupation2
- Breadth2
- Breakout2
- Daily profile2
- Fair value gap2
- Macro and intermarket2
- Managed futures trend2
- Path order2
- Premium discount2
- Session liquidity2
- Tail budget2
- Volume order2
- Failed breakout1
- Holding clock1
- Judas swing1
- Liquidity draw1
- Liquidity run1
- Low-volatility squeeze and breakout1
- Open-interest and price divergence1
- Participation1
- Realized-volatility expansion1
- Regime1
- Session range1
- Session relay1
- Stop reentry1
- Underwater entry1
- Volatility-gated sentiment extremes1
Historical backtests net of modelled costs, not a recommendation. A colour describes one strategy's result over its test window and says nothing about future returns.
The closest results
These five came out furthest ahead of the benchmark. Being furthest ahead among 510 tries is what luck alone produces, which is why the corrected q matters more than the headline figure.
- Alt/BTC ratio above its 50-day SMA (alts that beat Bitcoin) (Momentum): +130.8 pts/yr vs the benchmark, unadjusted p=0.015, corrected q=0.693.
- Alt/BTC ratio above its 50-day SMA, restricted to the 9 alts with the lowest 90-day BTC beta (A x B) (Momentum): +87.9 pts/yr vs the benchmark, unadjusted p=0.057, corrected q=0.904.
- Double sort: bottom-half BTC-beta intersected with top-half 4-week momentum (Equity factors): +59.8 pts/yr vs the benchmark, unadjusted p=0.025, corrected q=0.766.
- Crypto momentum consensus: top half on every horizon of 14/21/28 days (Momentum): +56.9 pts/yr vs the benchmark, unadjusted p=0.008, corrected q=0.635.
- r2-09 leg (a): two-week momentum top 4 in greed, cash in fear, equal weight when neutral or unavailable, weekly (Sentiment extremes): +54.2 pts/yr vs the benchmark, unadjusted p=0.094, corrected q=1.000.
Tickfloor's research desk has backtested 1,346 strategies, most against an equal-weight benchmark of the same assets rebalanced daily that pays no costs (the strategies pay theirs). After correcting for the 1,778 tests run, 0 passed. That does not prove no strategy works. These are results from our older test setup, which was built before our current testing rules. They have not been redone under those rules.
General information only. It doesn't consider your objectives, finances or needs. Tickfloor holds no financial services licence and never places trades.
Strategy families tested on crypto pairs
- Trend following (83 on crypto pairs)
- Momentum (68 on crypto pairs)
- Mean reversion (57 on crypto pairs)
- Volatility (38 on crypto pairs)
- Sentiment extremes (25 on crypto pairs)
- Equity factors (24 on crypto pairs)
- Other tested patterns (23 on crypto pairs)
- Social media chart patterns (22 on crypto pairs)
Do trading strategies work on crypto?
None of the technical indicators and other rules passed the corrected test in these backtests. Of 510 strategies tested on crypto pairs, 94 finished ahead of the equal-weight benchmark on the point estimate, 13 of those with an unadjusted p-value under 0.05, and none met the corrected Benjamini-Hochberg bar (q below 0.10). The median strategy returned −50.6 pts/yr against the benchmark. These are results from our older test setup, which was built before our current testing rules. They have not been redone under those rules.
What is the benchmark?
An equal-weight long position in the same assets, rebalanced daily, with no trading costs taken off it. The strategies pay trading costs and the benchmark does not, which makes the test harder to pass, not easier.
Why correct the p-values?
When 1,778 tests are run, some will look significant by luck. The Benjamini-Hochberg correction accounts for how many were tried before calling any result real.
Is this advice on what to trade?
No. It reports what happened to published and popular rules in the past. General information only, not personal advice.
More
- Strategies tested on US stocks
- Strategies tested on ETFs
- Every popular rule we tested
- Trading glossary
- The full method and every result