We tested one unchanged rule set on daily BTCUSDT candles from 20 July 2018 through 19 September 2026. Entries use only information available at the preceding close, execute at the next open and include 0.15% cost on each side. ETHUSDT is tested separately as a robustness check.
The hypothesis: pay for confirmation, then stay with the trend
The 55-day high asks whether price has exceeded the highs of the previous 55 completed daily candles. Bitcoin trades seven days a week, so this is about eight calendar weeks, not eleven five-day trading weeks. SMA 200 blocks entries in a weak long-term regime. ATR turns current volatility into a stop distance instead of one fixed percentage for every cycle.
These components solve different problems, but they do not predict the future. Breakouts can fail repeatedly in sideways markets, while the trend filter can re-enter late after a rapid recovery. The edge, if it exists, appears only across a sequence of trades.
Exact rules used in the test
Market and direction
BTCUSDT spot, daily candles, long only and no leverage. Only one position may be open.
Regime filter
The signal-day close must be above the 200-day simple moving average.
Breakout signal
The close must exceed the highest high of the previous 55 completed candles; the signal candle is excluded from that channel.
Execution
Enter at the next daily open. A 0.15% fee-and-slippage assumption is charged both on entry and exit.
Protective exit
The initial stop sits two ATR(20) below entry. It then trails at three ATR below the highest close reached during the trade.
Channel exit
A close below the prior 20-day low schedules an exit at the next open. If a stop is crossed first, the stop receives priority.
The full BTC path—not a selected winning window
The first chart displays every daily candle, SMA 200, executed entries and exits, plus the active ATR trail. The equity chart compares the strategy with buy-and-hold from the same starting point; the third panel shows peak-to-trough decline.
BTCUSDT · 1D · SMA 200 · ATR(20)
2018-07-20 — 2026-09-19 · UTC
BTC · Net return / Buy & hold
$10,000 · Test period
BTC · Max drawdown
Test period · %
What the numbers actually say
The BTC model completed 31 trades. 13 were profitable, a 41.9% hit rate. Starting with $10,000, the fee-adjusted final value was $24,325, representing a return of 143.2%, with a maximum drawdown of 39.1%.
Buy-and-hold returned 984.6% over the same measured window. The breakout model therefore made money but sacrificed most of a powerful secular rise without delivering a dramatically smaller drawdown. That is a material weakness, not a footnote.
The distribution is typical of trend following: the majority of attempts do not need to win. What matters is whether the few extended trades are large enough to offset stops, delayed entries and costs.
Ethereum check: useful, but not proof
With identical parameters, ETH produced 32 completed trades, a 50.0% win rate, 204.3% net return and 38.0% maximum drawdown. The positive result on a second liquid asset is encouraging, yet two assets are still one crypto regime family—not independent proof of durability.
Where the test can mislead
Execution gap
A daily open can jump through the intended stop. The model uses the worse of the open and stop when that occurs, but real liquidity can still differ.
Parameter dependence
Fifty-five, 200, two ATR and three ATR are transparent choices, not natural laws. Re-optimising them after seeing the result invites overfitting.
Spot universe bias
BTC and ETH survived the full period. Testing only survivors says little about tokens that disappeared.
Tax and custody
The calculation excludes tax, withdrawal fees, custody failure and venue-specific trading limits.
A responsible next test
- Freeze the rules before opening a new data period.
- Repeat the calculation on several liquid assets and include delisted instruments where data permits.
- Test higher costs and one-day adverse gaps.
- Separate development years from an untouched out-of-sample window.
- Paper-trade the exact next-open and stop logic before risking capital.
A coherent trend model, not a shortcut to outperforming Bitcoin
This breakout framework is reproducible and profitable in the sample, but it did not beat passive BTC ownership and its drawdown remained substantial. Its real value is as a disciplined research baseline: every assumption is visible, every trade can be challenged and the disappointing comparison is left intact.
Data and method
Binance Spot REST API · BTCUSDT & ETHUSDT · daily OHLCV · UTC
Rules were fixed before the calculation. Results include the stated cost model and are not a forecast.