Can an investor use relative strength without pretending to forecast the next winner? We calculate each coin’s trailing 90-day return at the previous UTC close. A positive leader is held for the coming month; if both returns are zero or negative, the model holds cash that earns no interest.
The question
The idea in plain words
The aim is to hold whichever of BTC and ETH had the stronger recent trend, and avoid both when neither had a positive 90-day return. The rule does not predict tomorrow’s price.
Walk through the rule
- Look backOnce a month, compare each coin’s last completed daily close with its close 90 days earlier. For example, 120 / 100 − 1 = +20%. No price from the coming month enters this calculation.
- Choose one holdingSelect the higher positive return. If both are zero or below, hold a dollar-denominated cash position that earns no interest in the model. There is no 50/50 split while this rule is active.
- Change only if neededBuy or sell at the next daily open after calculating the signal. A direct BTC-to-ETH switch means selling BTC and buying ETH; the model deducts 0.10% on each side. If the choice is unchanged, it makes no trade.
- Wait until next monthHold the choice without a protective stop, then repeat the calculation. This simplicity also explains why a large temporary loss is possible.
A real observation from the test
On February 1, 2021, the previous completed candles showed BTC +144.24% and ETH +242.68% over 90 days. ETH had the larger positive return, so the model switched from BTC to ETH at that day’s open and paid two trading fees. These are past returns, not a promise that ETH rose afterwards.
What the numbers mean: A return is the change in the account from its starting value. Maximum drawdown measures the largest fall from an earlier account peak. A positive final return can still include a severe drawdown.
Exact trading rules
On the first available daily candle of every calendar month, calculate BTC and ETH returns using only closes already known. Sell the previous holding at that day’s open, then buy the new holding at the same open. Charge 0.10% on each executed side; a direct BTC-to-ETH switch therefore pays twice. No borrowing, leverage or short selling.
90D → 1M · BTC / ETH / USD
0.10% · UTC · BTC / ETH / USD
Real market charts
Model return · Buy and hold BTC · BTC/ETH 50/50
2018-04-02 — 2026-09-19 · USD · $10,000
Historical data: Binance Spot · BTCUSDT / ETHUSDT · 1D · UTC. 2018-04-02 — 2026-09-19.
What the test found
The model gained +2,098.14% with a -71.48% maximum drawdown and 37 changes of holding. Passive BTC returned +1,091.38% over the same dates. Most of the model’s growth came before 2022; the later period returned +45.58%. A good whole-period number does not imply a smooth ride.
Switches: 37 · Maximum drawdown: -71.48% · Final value: $219,814
| Holding | Return | Final value |
|---|---|---|
| Model return | +2,098.14% | $219,814 |
| Buy and hold BTC | +1,091.38% | $119,138 |
| BTC/ETH 50/50 | +843.05% | $94,305 |
Costs and robustness
Doubling fees to 0.20% per side changes the result to +1,997.07%. More troubling, shortening the signal to 60 days produces +61.87%, while 120 days produces +1,136.13%. These variants were checked after the main rule; they are a sensitivity warning, not alternative “winning settings”.
Where the model can fail
BTC and ETH survived this sample, so survivorship matters. Execution at the next daily open can differ from the historical candle, and taxes, exchange outages, spread, market impact and cash interest are excluded. A -71.48% fall may be intolerable even with a positive final return. This backtest is an illustration, not an out-of-sample forecast.
Editorial verdict
The rotation beat passive BTC in this particular sample, but its large losses and dependence on the 90-day window make the result fragile. Freeze the rule, reserve new data for forward evaluation and test execution costs before considering a live allocation.
Data and method
Binance Spot · BTCUSDT / ETHUSDT · 1D · UTC · 2018-04-02 — 2026-09-19.