How to backtest a crypto strategy

Step 0: write the idea as rules

A backtest can only test what is written down precisely. Before opening any tool, answer four questions about your idea:

1. Market. Spot, or perpetual futures with leverage and funding? Which coins, which timeframe?
2. Entry. The exact condition, with numbers: "the 50-period SMA crosses above the 200-period SMA", not "the trend turns up".
3. Exit. A take-profit, a stop-loss, an opposite signal, a time limit — or several of them.
4. Size. How much of the account goes into one trade.

If you are not sure the idea is precise enough, read what backtesting is first, or describe it in plain words to ChatGPT or Claude and let the assistant turn it into rules — this guide shows how.

Run the backtest in six steps

1. Create a free account. Sign up — no card, no download. Everything runs on the server.
2. Create the strategy. Open Strategies → New strategy and pick a template close to your idea, start from a blank strategy, or choose "Write it with AI". The market — spot or futures — belongs to the strategy.
3. Build the rules. In the visual editor combine indicators (100+ of them, from RSI and moving averages to market structure, funding and open interest) with AND / OR groups, crosses and look-back windows, then add the exits. Under the hood the strategy is a JSON config, so nothing you write is code.
4. Set the costs. Fill in Fees % and Slippage % in the editor. Both start at 0, and a backtest without costs flatters every strategy — see the next section for realistic values.
5. Choose what to test on. Open New backtest, pick the strategy, one or more symbols and timeframes, the period (from 2020-01-01) and the initial capital (10,000 USDT by default). Each symbol × timeframe pair becomes its own backtest. For futures also set the leverage, the margin mode and whether to charge funding.
6. Run and read the report. Progress streams live; when a run finishes, its report shows the equity curve, every trade on the price chart, drawdowns, monthly returns and 60+ metrics.

Set realistic fees and slippage

Costs are charged on both sides of every trade, so a strategy that trades often pays them many times. As a starting point, use the exchange’s standard rates and add a little slippage:

Binance spot

Fee to enter (taker)0.1% per side at the standard tier, at the time of writing

Slippage to start with0.02–0.05%; more on small coins

Also charged—

Binance USDT-M perpetuals

Fee to enter (taker)0.05% per side at the standard tier (0.02% for maker orders), at the time of writing

Slippage to start with0.02–0.05%; more on small coins

Also chargedFunding every 8 hours; liquidation if margin runs out

Check the current rates for your account tier on the exchange’s fee page. When in doubt, round costs up: a strategy whose edge disappears at slightly higher fees had no edge. Janorium’s leaderboard ignores runs without fees for the same reason.

A real example: the golden cross on BTC

Here is one backtest from the golden-cross study: hold BTC on Binance spot while the 50-day SMA is above the 200-day SMA and sell when it drops below, 2020-01-01 → 2026-01-01, 10,000 USDT, 0.05% fee and 0.05% slippage on every fill. The full report is public.

Total return

Value+529.12%

What it saysThe strategy made money…

Buy and hold

Value+859.56%

What it says…but simply holding BTC made more: the cross lost 330 percentage points to it.

Max drawdown

Value57.11%

What it saysAt the worst point the account was more than half below its peak.

Trades

Value6

What it saysSix trades in six years — too few to call the result a statistic.

Win rate · profit factor

Value66.67% · 4.95

What it saysExcellent-looking, and almost meaningless on six trades.

This is the typical first backtest: a positive return that looks good until you put it next to buy and hold and count the trades. The next step is to find out whether anything survives a harder test.

Check that the result survives

• Compare with buy and hold. The report shows it next to the return. A strategy that loses to holding has to earn its place some other way — a much smaller drawdown, for example.
• Test more coins. Run the same strategy on 10–20 symbols in one go. An edge that shows up on one coin only is usually luck.
• Include dead coins. "Take the symbols as of a past date" builds the list from the coins that were large back then, delisted ones included — today’s survivors make every long strategy look better.
• Shift the start date. "Test sensitivity to the start date" reruns the strategy from several starts; if the result swings, it was carried by one lucky stretch.
• Raise the costs. Double the fees and slippage. A real edge shrinks; a fragile one turns into a loss.
• Compare runs side by side. Put up to four runs next to each other with a significance test, and use the Monte Carlo tab to see the range of drawdowns luck alone could produce.

Which numbers to look at, and what counts as good, is the subject of the guide to backtest metrics.

Common mistakes and how to avoid them

Fees left at 0

Why it misleadsEvery trade looks free; frequent strategies look brilliant.

FixSet Fees % and Slippage % before the first run.

One coin, one period

Why it misleadsOne bull market can carry any long strategy.

FixTest many coins and several start dates.

Only today’s coins

Why it misleadsSurvivorship bias: the coins that died are missing.

FixUse a universe taken as of a past date.

A dozen trades

Why it misleadsWin rate and profit factor are noise on small samples.

FixPrefer rules and periods that produce dozens or hundreds of trades.

Tuning until it looks perfect

Why it misleadsOverfitting: the parameters fit the past, not the market.

FixKeep a later period untouched and judge the strategy there.

Futures without funding

Why it misleadsA long position in a hot market can pay more in funding than in fees.

FixLeave "Apply funding" on.

Frequently asked questions