Research ·
Leverage set to 5×, used at 0.3×: what the bot really holds
The exchange setting says 5×. The positions averaged 0.30× the account, were below 0.63× nine-tenths of the time, and peaked at 1.02×. The leverage setting and the risk taken are different things.
Copy-trading pages show a leverage number next to every trader, and people read it as a risk score. Ours says 5×. That number is real, and it tells you almost nothing about how much risk the bot takes.
Two different numbers
The leverage setting is the maximum the exchange will let a position borrow. It decides how much margin a position locks up, not how big the position is.
Exposure is what the bot actually holds: the total value of all positions divided by the account. A $1,000 account holding $300 of Bitcoin has 0.3× exposure, whether the leverage setting is 1× or 20×. The price moves you gain or lose on come from exposure.
The bot sets 5× so that positions use little margin (useful for copiers, whose margin is tied up by every copied order) and sizes the positions itself.
What six years of backtest held
| Total exposure, Jan 2021 – Sep 2026 | × the account |
|---|---|
| Average | 0.30× |
| Median (half the time below this) | 0.26× |
| 90% of the time below | 0.63× |
| 99% of the time below | 0.91× |
| Highest ever | 1.02× |
At the 5× setting, that means about 6% of the account was used as margin on average, and 20.5% at the busiest moment. The rest sits free.
By year, average exposure stayed in a narrow band: 0.21× in 2021, 0.27× in 2022, 0.37× in 2023, 0.32× in 2024, 0.27× in 2025 and 0.36× so far in 2026. Even in the 2021 bull market it didn't pile in.
Why so little
Three rules keep it small, in this order:
- A volatility target. The portfolio aims to swing about 20% a year. Our ten coins have swung between about 50% (Bitcoin) and 100% (Solana, Dogecoin, Avalanche) a year since 2021, so reaching 20% needs only a fraction of the account in them. When markets calm down, positions grow; when they get wild, positions shrink automatically.
- Four coins at most. Only the strongest trends get a seat (why four). The average was 2.8 positions; 12.6% of the time it held nothing at all.
- Hard caps. No coin above 25% of the account, and never more than 1.5× in total. The 1.5× cap was never reached in the backtest. The 25% cap is checked when positions are resized; between resizes a winning position can drift a little above it. The largest single position in six years was 28.9% of the account, and 95% of the time the largest was at most 25.2%.
Long and short, roughly half each
The bot can bet on prices falling as well as rising. Over the six years it was net long 43.9% of the time, net short 43.4%, and roughly flat the rest. Short positions made up just over half of all position-time (51.2%). That surprises people who assume a crypto bot is a leveraged bet on crypto going up. It isn't: in the 2022 bear market the backtest made 36.3%, mostly from shorts.
What it means if you copy
- Liquidation isn't the risk to worry about. With exposure around 0.3× and a stop-loss 25% from every entry, the stops act long before the exchange would liquidate anything.
- The real risk is ordinary: a run of trends that reverse. The backtest's worst peak-to-trough drop was −18.5%; every drop over 8% is listed here.
- If you copy with a fixed ratio, most of your copy account sits unused as margin most of the time. That's by design, not a mistake.
These figures come from the same backtest as the strategy page: the live rules, Binance 4-hour candles, 0.07% cost per trade, funding not included. The live exposure is shown on the performance page under "Open now".