# What leverage does to one strategy: 1x, 2x, 3x, 5x

> The same golden cross on BTC and ETH perpetuals, 2021–2026: BTC +9.78% at 1x, -7.76% at 2x, -99.77% at 5x. Funding times leverage moved the liquidation price.

Research · Janorium · Published 2026-10-20  
Web page: https://janorium.com/articles/crypto-leverage-backtest-1x-2x-3x-5x

It turned a small profit into a wipe-out. The same golden-cross strategy on the BTC perpetual made +9.78% at 1x, -7.76% at 2x, -94.28% at 3x and -99.77% at 5x; on ETH it went from +63.56% to -99.63%. Price was not the main culprit: the 3x BTC long was liquidated above its entry price, because the funding it paid ate its margin.

Leverage multiplies a position, so the intuition is that it multiplies the result: twice the return, twice the
drawdown. On a perpetual future, over a few years, that is not what happened. One textbook strategy, run unchanged at
four leverage levels, went from a modest profit to almost nothing left — and the run without funding shows why.

## Setup

- **Strategy:** the classic golden cross — long while the daily 50-day simple moving average is above the 200-day one,
  flat after the death cross. The same rule as [the golden cross study](https://janorium.com/articles/golden-cross-top-20-crypto-2020-2025),
  not tuned here.
- **Market:** Binance USDT-M perpetuals BTCUSDT and ETHUSDT, 4h bars, 2021-01-01 → 2026-10-01.
- **Sizing:** every entry puts 95% of the account up as **isolated** margin; the position is that margin times the
  leverage. Leverage 1x, 2x, 3x and 5x — eight runs, nothing else changed.
- **Costs:** 0.05% fee and 0.05% slippage per side, fills at the next bar's open, funding charged and paid every
  8 hours.
- **For comparison:** holding the coin over the same window, as each run's report shows it — +185.73% for BTC and
  +261.34% for ETH.

## The results

| Run                      |  Return | Max drawdown | Funding paid, USDT | Liquidations | Sharpe | Profit factor |
| ------------------------ | ------: | -----------: | -----------------: | -----------: | -----: | ------------: |
| [BTC 1x](https://janorium.com/backtests/766) |  +9.78% |       67.93% |           6,776.99 |            0 |   0.26 |          1.19 |
| [BTC 2x](https://janorium.com/backtests/767) |  -7.76% |       88.33% |          12,184.33 |            0 |   0.40 |          0.92 |
| [BTC 3x](https://janorium.com/backtests/768) | -94.28% |       99.62% |          11,333.91 |            1 |   0.39 |          0.08 |
| [BTC 5x](https://janorium.com/backtests/769) | -99.77% |       99.99% |          18,267.41 |            2 |   0.60 |          0.01 |
| [ETH 1x](https://janorium.com/backtests/770) | +63.56% |       73.90% |          14,425.20 |            0 |   0.46 |          1.40 |
| [ETH 2x](https://janorium.com/backtests/771) | -35.23% |       96.84% |          32,596.06 |            0 |   0.44 |          0.92 |
| [ETH 3x](https://janorium.com/backtests/772) | -95.50% |       99.86% |          40,772.25 |            2 |   0.38 |          0.84 |
| [ETH 5x](https://janorium.com/backtests/773) | -99.63% |      100.00% |          83,556.56 |            3 |   0.51 |          0.89 |

Doubling the leverage did not double anything useful. At 2x neither coin was ever liquidated, and both still lost
money: the drawdown deepened to 88.33% and 96.84%, and the funding bill roughly doubled with the position. From 3x up,
the first liquidation took almost the whole account, and the strategy went on trading the same signals with what was
left.

**Backtest: [Golden cross 50/200 daily — leverage study](https://janorium.com/backtests/766)**  
BTCUSDT · 4h · Futures · 2021-01-01 → 2026-10-01

| Return | Max drawdown | Sharpe | Win rate | Trades |
| --- | --- | --- | --- | --- |
| +9.78% | 67.93% | 0.26 | 57.1% | 7 |

*BTC, golden cross at 1x — +9.78%, against +185.73% for holding*

**Backtest: [Golden cross 50/200 daily — leverage study](https://janorium.com/backtests/769)**  
BTCUSDT · 4h · Futures · 2021-01-01 → 2026-10-01

| Return | Max drawdown | Sharpe | Win rate | Trades |
| --- | --- | --- | --- | --- |
| -99.77% | 99.99% | 0.60 | 44.4% | 9 |

*BTC, golden cross at 5x — -99.77%, two liquidations*

**Backtest: [Golden cross 50/200 daily — leverage study](https://janorium.com/backtests/770)**  
ETHUSDT · 4h · Futures · 2021-01-01 → 2026-10-01

| Return | Max drawdown | Sharpe | Win rate | Trades |
| --- | --- | --- | --- | --- |
| +63.56% | 73.90% | 0.46 | 71.4% | 7 |

*ETH, golden cross at 1x — +63.56%*

**Backtest: [Golden cross 50/200 daily — leverage study](https://janorium.com/backtests/773)**  
ETHUSDT · 4h · Futures · 2021-01-01 → 2026-10-01

| Return | Max drawdown | Sharpe | Win rate | Trades |
| --- | --- | --- | --- | --- |
| -99.63% | 100.00% | 0.51 | 50.0% | 10 |

*ETH, golden cross at 5x — -99.63%, three liquidations*

## A 3x long liquidated above its entry price

The first BTC trade opened on 2021-01-01 at 29,317 and stayed long through the 2021 rally. At 3x without funding its
liquidation price was 19,662 — a third below entry. With funding it was liquidated on 2021-05-19 at 30,798, **above the
price it was bought at**. Between those two dates BTC funding averaged 0.0554% per 8 hours — more than five times its
0.0098% average over the whole 2021–2026 window — and was above 0.05% at 44.8% of settlements[^1]. On a position whose
value rose with the rally, that came to 10,846 USDT — more than the trade's whole 9,505 USDT margin. At 5x the same trade paid 18,077 USDT and
was liquidated the same day at 34,705.

The mechanism is isolated margin. A long's funding payments are taken out of the margin that holds the position open;
less margin means a smaller loss is enough to hit maintenance, so the liquidation price climbs a little at every
settlement. Over five months of hot funding it climbed past the entry. Even the 1x run felt it: its first trade
gained 21% on price (29,317 → 35,505) and still closed at a loss of 1,683 USDT, because it paid 3,679 USDT of funding.

Switch funding off and the same 3x strategy does something else entirely — never liquidated, +359.94%. That run is
not achievable on a perpetual; it only isolates how much of the damage was funding rather than price:

**Backtest: [Golden cross 50/200 daily — leverage study](https://janorium.com/backtests/774)**  
BTCUSDT · 4h · Futures · 2021-01-01 → 2026-10-01

| Return | Max drawdown | Sharpe | Win rate | Trades |
| --- | --- | --- | --- | --- |
| +359.94% | 85.38% | 0.74 | 71.4% | 7 |

*BTC, golden cross at 3x with funding switched off — +359.94%, for comparison only*

> **How Binance handles it, and how this backtest does**
>
> Binance's funding FAQ, checked 2026-10-20: "Funding fees (if any) will be deducted from the available balance in
> your Futures Account. If your account balance is insufficient, the funding fees (if any) will be deducted from your
> position margin, which may affect your liquidation price."
> ([Introduction to Binance Futures Funding Rates](https://www.binance.com/en/support/faq/introduction-to-binance-futures-funding-rates-360033525031))
>
> The backtest takes funding from the isolated margin straight away. Here that is the case Binance describes once the
> free balance is gone: with 95% of the account committed, only about 500 USDT stayed free, against 10,846 USDT of
> funding on the first 3x trade. A trader holding more cash outside the position — or topping the margin up — would
> have been liquidated later, at the price of more capital at risk.

## Sharpe barely noticed

The [metrics guide](https://janorium.com/guides/backtest-metrics-explained) separates _shape_ metrics — Sharpe, Sortino, profit factor,
which describe how a strategy trades — from _scale_ metrics — return and drawdown, which grow with the bet size. Here
the split shows up in its sharpest form. Sharpe stayed between 0.26 and 0.60 on BTC and between 0.38 and 0.51 on ETH
while the return went from a profit to -99%; the 5x BTC run even has the highest Sharpe of the four. Sharpe is an
average of daily returns over their spread, and leverage scales both by the same factor — it does not see that the
compounded path ran into a liquidation.

Profit factor is a sum of money, so it depends on when the losses came. On BTC the liquidation came first, on a full
account, and the later winners were tiny — profit factor fell to 0.08 and 0.01. On ETH the big winner (2021) came
before the liquidation, so from 2x up profit factor stayed between 0.84 and 0.92 while the account was emptied.

The practical reading: judge the rule by its shape at 1x, then choose the leverage from the drawdown and from where
the liquidation price can drift — not from the Sharpe, which will look the same at every leverage right up to zero.

> **What this test does not show**
>
> - **One rule, one window.** The golden cross holds for months, which is exactly what lets funding pile up; a strategy
>   that holds for days pays far less funding per trade. And early 2021 had unusually hot funding; a window that skips it
>   would probably look less brutal.
> - **No margin top-ups.** Each trade keeps the margin it opened with; nobody adds collateral as the liquidation price
>   rises.
> - **Isolated margin only.** Cross margin, where the whole account backs the position, is not compared here.
> - **Two coins.** Smaller coins swing harder, which makes each step of leverage more dangerous, not less.

## Conclusion

- The same golden cross went from +9.78% (1x) to -99.77% (5x) on BTC and from +63.56% to -99.63% on ETH, 2021–2026.
  Even 2x, never liquidated, lost money on both coins.
- On a perpetual, leverage multiplies funding as well as price. Under isolated margin funding drains the margin, so a
  long held through hot funding gets a rising liquidation price — at 3x on BTC it overtook the entry.
- Sharpe stayed in the same band at every leverage; drawdown and return did not. Use the shape metrics to judge the
  rule and the scale metrics to judge the bet.

More research: [funding as a signal](https://janorium.com/articles/funding-rate-signal-backtest) and
[how to backtest a crypto strategy](https://janorium.com/guides/how-to-backtest-a-crypto-strategy).

[^1]:
    Computed from Binance's public USDT-M funding-rate history for BTCUSDT (415 settlements from 2021-01-01 to
    2021-05-19). The backtest's funding on this trade matches a replay of those rates to the cent.