What is backtesting? How to backtest a crypto strategy honestly

What backtesting means

Backtesting means running a set of trading rules on past prices to see what they would have done. The software steps through history one price at a time, applies your rules exactly as written, and records every trade the rules would have made, what each one cost and how the account would have grown or shrunk.

A backtest answers one question: would these rules have worked in the past, after costs? It is the cheapest way to throw out bad ideas. A rule that lost money over the last two years is unlikely to be a good place to start.

It cannot answer whether the rules will work next month. Markets change, and a rule that fitted one period may fail in the next. That is why the step after a backtest is paper trading, not real money.

How a backtest works

1. Choose the market and period, for example BTC/USDT on Binance over the last 2 years.
2. Write the rules precisely. "Buy when BTC falls 5% within 24 hours, sell at 4% above the buy, stop at 6% below, at most 2 buys open."
3. Replay the prices. At each step the test checks whether a rule triggers and simulates the order at that price.
4. Apply costs. Every simulated order pays the exchange fee and an allowance for slippage, the gap between the expected and the actual fill price.
5. Summarise. The test reports the result over the period, the number of trades, the share that gained, and the worst drop from a high point along the way.

How to read a backtest honestly

Four numbers matter more than the headline result:

• The period. A result without a period means nothing. Twelve months that include a crash are very different from twelve months of rising prices.
• The worst drop. Also called max drawdown: the largest fall from a high point before recovery. It tells you how painful the strategy would have been to live with. TradeCorgi shows it at the same size as the result, never in small print.
• The comparison with holding. Put the strategy next to simply buying the coin at the start and holding it. Many rules lag holding in rising markets and only help in falling ones. That can be fine, if you know it.
• The number of trades. Five trades is an anecdote. Fifty trades starts to say something. A strategy with very few trades in its backtest has barely been tested.

The mistakes that make backtests mislead

• Leaving out fees and slippage. The most common flattering mistake. A rule that trades daily at a 0.1% fee per order pays about 0.2% for every round trip, and over a year that can erase a small edge entirely.
• Testing one period only. Run the same rules over different stretches, such as a rising year, a falling year and a flat one. A rule that only works in one is a bet on that kind of market returning.
• Overfitting. If you tweak the numbers until the result looks perfect (4.7% instead of 5%, 3.9% instead of 4%), you have probably fitted the past rather than found a pattern. Prefer round, simple rules that work across a range of values.
• Look-ahead bias. A test that uses information the rule could not have known at that moment, such as the day's closing price at noon, will always look better than reality.
• Ignoring the worst drop. A strategy that ends higher but falls 40% on the way is one most people abandon at the bottom.

A worked example

Take the BTC Dip Buyer and backtest it over 90 days on Binance, with a 0.1% fee on every buy and sell and 0.05% slippage per order:

• The rules trade 9 times: 6 trades gain, 3 hit the stop-loss.
• Over those 90 days the account ends 6.8% higher than it started, after fees.
• Simply holding BTC over the same 90 days ends 3.1% higher.
• The worst drop along the way is 4.9% from a high point, against 10.7% for holding.

Read honestly: over this one 90-day period the rules did better than holding with a smaller worst drop, but 9 trades in one period is thin evidence. The next step would be a longer backtest, then paper trading. This is an illustrative example, not a real result.

Backtesting on TradeCorgi

When you describe a strategy, TradeCorgi backtests it automatically once you confirm the rules:

• Fees always included, using the exchange's fee for your tier, plus slippage and the exchange's minimum order size.
• Holding the coin on the same chart, so you can see whether the rules add anything.
• The worst drop as large as the result.
• History by plan: 90 days on Sandbox, 2 years on Copilot, 4 years on Autopilot and the full exchange history on Pro.

Every result is labelled with its period and fees. A good backtest unlocks paper trading, never live trading directly.

A backtest checklist

• The period is stated, and covers more than one kind of market.
• Fees and slippage are included on every trade.
• The result is compared with simply holding the coin.
• The worst drop is shown, and you could live with it.
• There are enough trades to mean something.
• The rules are simple and round, not tuned to the decimal.
• The next step is paper trading, not your full live amount.