A trend-following strategy spends a large part of its time out of the market, earning nothing. Meanwhile the exchange’s own market-making vault earned steady returns with small drawdowns. The two earn in opposite regimes: BTC in uptrends, market making through volatility and downtrends.
The rule: hold BTC, unlevered, while a slow moving-average trend filter is up; otherwise park the capital in the exchange-run market-making vault, respecting its deposit lockup.
Once a day, after the closeCheck whether Bitcoin sits above its slow trend line.
Trend up: hold BitcoinUnlevered, never borrowed against.
Trend down: earn yieldThe money moves to the exchange’s market-making vault instead of idling in cash.
| Mid-2023 to late 2026 | Growth of 100 | Per year | Worst drawdown |
|---|---|---|---|
| Trend + carry | 553 | +69% | −25% |
| Trend, parked in cash | 263 | +34% | −29% |
| Market-making vault only | 296 | +40% | −6% |
| BTC buy-and-hold | 283 | +37% | −53% |
Out of sample
−15.9%
−0.8%
−53.0%
Positive every calendar year: +102% in 2023, +76% in 2024, +18% in 2025 and +25% in 2026 to date.
Robustness
- Not tied to one setting: every trend length from 20 to 200 days works, beats parking in cash and stays positive in both periods (+13% to +25% a year since mid-2025).
- Not tied to one coin: the same structure works on ETH.
- Not tied to one cycle: the trend rule alone holds over 2018–2026, including the 2018 and 2022 crashes.
Caveats
- The strongest setting is the best of the grid; plan on the middle of the range: about +15–20% a year with −15% to −25% drawdowns.
- The vault’s yield has fallen as it grew, from about 57% to about 15% a year, and it carries protocol risk that three years of history do not show.
- Unlevered only: even the trend rule alone drew down −57% in 2018–2022.