Research · all tests in one place
Crypto trading strategies, tested
Most lists of crypto trading strategies explain how each one works. This page shows what happened when we tested the popular ones on real exchange data, with the rules written down before any data was loaded, a control for luck, a check on data the idea was not built on, and trading costs. Six tests so far: no edge in five, one undecided.
In short: fading pumps, buying crashes, trading against extreme funding, Uptober, the rainbow chart and altcoin season all failed to beat chance after costs. Several of them win most of the time and still make nothing, because a few trades lose far more than the usual win. In the two tests that compared a timing rule with simply holding bitcoin, holding won both times. Liquidation-cluster "magnets" are still being tested on new data.
The scoreboard
| Strategy | The popular claim | What the test found | Verdict |
|---|---|---|---|
| Short the pump | "Every pump gets dumped" | 69% of shorts won, average +1.1% (t 1.03); 27 of 1,215 lost more than the whole position | No edge |
| Buy the crash | "Coins bounce after a dump" | 59% of buys lost over 72 h; average +1.3% (t 0.71) from a few big rebounds | No edge |
| Fade extreme funding | "Crowded side gets squeezed" | Short at high funding: 55% won, average -0.33%; long at negative funding: 43% won, +0.07% | No edge |
| Uptober | "Bitcoin rises in October" | Up in 10 of 13 Octobers, but p = 0.125 - chance can produce that with 13 years | No edge |
| Rainbow chart | "Buy blue, sell red" | $1 became $22 with bands known at the time, $193 by holding ($397 with hindsight) | No edge |
| Altcoin season | "Altseason - rotate into alts" | Altcoins beat bitcoin the next month after 3 of 8 readings (p 0.24) | No edge |
| Liquidation magnets | "Price hunts liquidation clusters" | Bigger cluster hit first 55.2% of 105 events (p 0.165); forward test running | Undecided |
Want to guess before reading? Take the Edge or Not? quiz.
"No edge" means the result was not distinguishable from chance with the data available, after costs. It does not prove an idea can never work - only that this version, on this data, did not.
Winning often is not the same as making money
The most common pattern in our tests: a strategy that wins most trades and still averages about zero. Four contrarian strategies, each dot is all the trades of one rule after costs:
Shorting pumps won 69% of the time; buying crashes won 41%. Both averaged about +1% and neither is statistically different from zero. A high win rate usually means small wins paid for by rare large losses: 27 pump shorts lost more than the whole position, and 21 funding shorts lost more than all of it in a single day. With leverage, each of those trades is a liquidation.
Popular strategy types and what the data says
Buy and hold (HODL)
Not a test of its own here, but the benchmark every timing rule has to beat. In the two tests that compared timing with holding, holding bitcoin won: the rainbow rule turned $1 into $22 from 2016 to 2026 while holding made $193, and holding only through Octobers made ×8.5 from 2013 while holding through every month made ×6,311. Holding also had the deepest drawdown in the rainbow test (-75%), and past growth says nothing about the next ten years.
Timing with valuation models
Models like the rainbow chart fit a curve to the whole price history, so they always look right in hindsight. Redrawn each month with only the prices known at that time, the rainbow sold bitcoin in May 2017 at $2,298 and stayed out for five years. Full test.
Seasonality
"Uptober", "sell in May", the September dip. With 13 to 14 years of monthly data, no calendar month passes a test corrected for checking all 12. October's record is good and still within what chance produces. Full test, live table: bitcoin monthly returns.
Contrarian and mean reversion
Fading pumps, buying crashes, shorting when funding shows a crowded long side. All three failed the same way - frequent small wins, rare huge losses, average near zero. Live data for the same signals: pump & dump scanner, funding rates.
Rotation between bitcoin and altcoins
The altcoin season index describes the last 90 days well and the next month poorly. The steadier fact: an equal-weight basket of the top 50 altcoins lagged bitcoin in 62% of months since late 2019, by 1.8% a month on average. Full test.
Liquidity and liquidation hunting
The idea that price is pulled toward big liquidation clusters is everywhere and almost never checked. On the first 30 days it did not beat a coin flip; the forward test on new data should reach a verdict around December 2026. Test, how to read the map.
Day trading and scalping with leverage
We have not tested a day-trading system - there are too many variants to call one "the" strategy. What our data does show is the cost of the usual tools. Costs are charged on the position: at a 0.05% taker fee, opening and closing a 10x position takes 1% of the margin. And leverage gets liquidated often:
| Bitcoin long held 1 day | 10x | 25x | 50x | 100x |
|---|---|---|---|---|
| Liquidated in this share of past days | <1% | 6% | 34% | 69% |
| Move against you that ends it | 9.6% | 3.6% | 1.6% | 0.6% |
Binance hourly data 2026-04-14 to 2026-10-20, isolated margin, 0.4% maintenance, before fees and funding. Any coin and holding time: how fast would you get liquidated? Explained: crypto leverage trading.
Arbitrage
Funding rate arbitrage - short the perpetual where funding is high, long where it is low - is a hedged trade, not a bet on direction, so a "does it predict price" test does not apply. Its risks are different: rates flip, prices on two exchanges drift, one leg can be liquidated. Live spreads and fee payback: funding rate arbitrage scanner.
Not tested yet
Dollar-cost averaging, trend following, breakouts, grid trading and indicator signals (RSI, MACD, moving-average crosses). They will be tested the same way - or listed here as untested until they are. "Not tested" is not the same as "failed".
How a strategy gets a verdict here
You can apply the same checklist to any strategy you are offered, from a course, a signal channel or a bot: were the rules fixed before the results were known? Was it compared with random entries? Does it hold on a period it was not built on? Are fees, funding and the worst trades included? A backtest that skips these is a description of the past, not evidence. More on the method: about.
FAQ
What is the best crypto trading strategy?
None of the six popular strategies we tested had an edge after costs. In the two tests that compared a timing rule with simply holding bitcoin, holding won both times: $193 vs $22 for the rainbow rule, ×6,311 vs ×8.5 for holding only in Octobers. A record of past data, not investment advice.
Is crypto trading profitable?
For some traders, yes - but none of the six popular rules we tested made money beyond what chance produces once costs were included. The two reasons we saw most often: trading costs, and rare, very large losing trades.
Is crypto day trading profitable?
We have not tested a day-trading system. Our data shows its two biggest headwinds: costs (1% of margin to open and close a 10x position at a 0.05% taker fee) and leverage (a 50x bitcoin long held for one day was liquidated in 34% of past days).
Do crypto trading bots work?
A bot only executes a strategy, so it has an edge only if the strategy does. The checklist above applies to any bot's logic.