At first glance a leverage backtest seems to promise the obvious: more leverage, more return. We ran the same strategy – RSI/SMA Cross on BTCUSDT, 2018 to today, 510 trades – once unleveraged and once at 5× with a full position. Unleveraged: +21.1 % CAGR. At 5× with a 100 % position: account bankrupt on December 7, 2018, −100 % CAGR. The backtest ends where the account is wiped out.
That is neither a fluke nor a rounding error. It is the direct consequence of how leverage actually works – and what most people underestimate. This post explains the model behind the new leverage backtester before you use it yourself.
Position size is margin, not stake
The root of nearly every misunderstanding: your position size is the margin – the share of your account that one position risks. Not the amount you "send into battle," but the collateral that burns when the trade goes wrong.
The rest stays in cash and does nothing.
From this follows the only formula you need to remember:
Effective exposure = position size × leverage.
| Setting | Margin | Effective exposure | Cash | Behavior vs. unleveraged |
|---|---|---|---|---|
| 20 % / 2× | 20 % | 40 % | 80 % | clearly more defensive |
| 20 % / 5× | 20 % | 100 % | 80 % | like unleveraged (all-in) |
| 50 % / 5× | 50 % | 250 % | 50 % | aggressive |
| 100 % / 5× | 100 % | 500 % | 0 % | extreme – rekts early |
An unleveraged run puts practically the whole account into each trade (100 % exposure). Choose 2× at a 20 % position and you land at 40 % exposure – less market participation than with no leverage at all. That is exactly why such a run sits below the unleveraged curve in the backtester: not despite the small position, but because of it.
The counterintuitive part
"Leverage" sounds like "more." Below 100 % effective exposure it means the opposite. The same RSI/SMA run, three settings:
| Setting | Exposure | CAGR | Result |
|---|---|---|---|
| unleveraged | 100 % | +21.1 % | reference |
| 2× / 20 % | 40 % | +8.5 % | more defensive, less return |
| 5× / 50 % | 250 % | −23.0 % | 1 liquidation, funding bites |
| 5× / 100 % | 500 % | −100 % | REKT on 2018-12-07 |
Anyone expecting "more leverage = more return" reads this table backwards. Return does not come from the leverage factor alone, but from exposure – and exposure above 100 % amplifies gains and liquidation risk in equal measure.
Liquidation hits a position, not the account
The leverage runs on isolated margin: each trade opens its own isolated position. If price moves roughly 1/leverage − 0.5 percentage points against you, that position is liquidated – its margin is gone. The maximum loss of one position is therefore exactly your position size.
At 2× this threshold sits at −49.5 %, at 5× already at −19.5 %.
Crucially, a liquidation does not stop the backtest. The account keeps trading with the remaining capital and opens the next position on the next signal. The 5× / 50 % run above shows exactly that: "1 liquidated," but no bankruptcy – one position lost its margin, the rest carried on.
Bankruptcy (REKT) only happens when cumulative account equity falls to zero or below. At a 100 % position, a single liquidation is enough – which is why the 5× run was already over in 2018.
What leverage does not touch
The leverage layer is a pure position-sizing overlay. It does not change the strategy engine: entry and exit signals stay identical to the unleveraged run. Leverage only scales how much is risked per signal, then adds liquidation and funding on top. So you compare the same strategy against itself – once bare, once leveraged.
Limits of the test
No backtest is reality, and a leveraged one least of all. Three points that make the result lean optimistic:
- Liquidation is checked on the daily low. The test looks bar-by-bar at each candle's low. Intraday wicks below the daily low – which do liquidate on real exchanges – go undetected. At high leverage the real liquidation rate is likely higher.
- Funding is simplified. You can pick "ignore" or a conservative flat 0.05 %/day; historical funding exists only for BTCUSDT from September 8, 2019. Real perp funding rates swing wider.
- No slippage, no partial fills, no cross-margin. The model is deliberately isolated-margin and long-only in this first stage.
A sober conclusion
A leverage backtest is not a one-click return amplifier. It is a risk calculator: it shows you at which combination of position size and leverage your strategy survives – and at which it stops existing on a specific date. For RSI/SMA on Bitcoin that date, at 5× / 100 %, was December 7, 2018.
FAQ
Is more leverage always worse in a backtest? No. Below 100 % effective exposure a leveraged run is more defensive than the unleveraged one; above it, more aggressive. What matters is position size × leverage, not the leverage factor alone.
What does "1 liquidated" mean in the result? A single position lost its margin. The account is not bankrupt because of it – it keeps trading with the remaining capital. Bankruptcy (REKT) only when total equity hits zero.
Why is my leveraged run below the unleveraged one? Because your effective exposure is under 100 %. Example: 20 % position × 2× = 40 % exposure – less market participation than the unleveraged all-in run.
Does leverage change my entry/exit signals? No. Leverage is a position-sizing overlay. The strategy buys and sells at exactly the same times as unleveraged.
Why is the result "optimistic"? Liquidation is only checked on the daily low, not on intraday wicks below it. Real liquidations can occur earlier and more often, especially at high leverage.
Test it yourself
The leveraged run sits on the same engine as the regular crypto backtest – every strategy, every pair. You set leverage, position size and funding; the rest stays your usual backtest. The leverage backtester itself lives at dashboard/leverage.
Not investment advice. All backtests are historical simulations with no guarantee of future results.