Your Crypto Backtest Is Lying to You: 3 Reasons Why

Key Takeaways

• A backtest that looks great but loses money live usually fails because of how it was measured, not the strategy itself.
• Three common flaws: look-ahead bias (using future information), overfitting (fitting the past too closely), and ignored fees and slippage.
• How to catch them: evaluate signals only on closed bars, cross-check several metrics and trade count instead of one number, and apply real trading costs.

A good backtest result does not mean a good strategy. Many impressive equity curves come not from the strategy but from a flaw in how it was measured. The same data and the same rules can produce completely different results depending on how you test them. Almost every backtest that collapses in live trading does so for one of the three reasons below. Here is how to spot each one and how to prevent it.

1. Look-ahead bias

Look-ahead bias means using information in a signal that was not actually available at that moment. The most common form is deciding an entry from a candle's close before that candle has closed. Quant analyst Michael Harris calls it "the worst type of bias," because the results are simply wrong and, unlike other biases, it shows up immediately in live execution.¹

The same thing happens in indicator math. If an unclosed final bar is folded into a moving average, that indicator carries a value it could never have had at that point in real time.

How to prevent it: evaluate every signal only on closed bars, and for each signal ask, "could I actually have known this at that moment?"

2. Overfitting

Overfitting means fitting a strategy so tightly to past data that it shines only on that stretch and dies on new data. A multiple-testing trap rides along with it: at a 5% significance level, testing 100 variations leaves about 5 that look like winners by pure chance. The 2014 study by Bailey, Borwein, López de Prado, and Zhu in the Notices of the American Mathematical Society showed mathematically that the more configurations you try, the higher the probability that a backtest is overfit, and that when a backtest does not report how many configurations were tried, its overfitting risk cannot be assessed at all.² Running more backtests, by itself, is not the answer.

Three habits reduce the trap:

• Look at several metrics, not one. Check win rate, maximum drawdown, and trade count alongside the Sharpe ratio so a single flattering number cannot fool you.
• Check the trade count. With a small sample, a great result may just be luck.
• Re-run on a different period. Confirm the strategy is not living off one lucky stretch.

3. Fees and slippage

Leave out trading costs and your paper profit drifts away from your real profit. Setting fees and slippage to zero makes a backtest assume a flawless fill that does not exist in real markets. Including only realistic slippage can trim simulated returns by 0.5 to 3 percent per year, by one analysis.³ The more often a strategy trades, the more this gap compounds, turning a paper winner into a real-money loser.

How to prevent it: apply your exchange's actual fees and a conservative slippage estimate. The result that survives after costs is the real result.

The common thread

All three flaws share one root: when the measurement is kinder than reality, the result is inevitably inflated. Whether you peek at the future, fit the past too closely, or skip the costs, you have sat an easier exam than real trading. The point of validation is not to make a strategy look good. It is to leave only the strategies that will survive. The harder you make the test, the more you can trust what passes.

Frequently asked questions

How long should I backtest a crypto strategy? It depends on how often the strategy trades, but to be statistically meaningful you want a large enough sample of trades (often 100 or more) and a span that covers rising, falling, and sideways markets.

Does paper trading replace backtesting? No. They are different stages. Filter candidates with a backtest first, validate in real time with paper trading, then move to live in small size.

Is a high Sharpe ratio always good? Not necessarily. With few trades, even a high Sharpe can be luck. Read it alongside maximum drawdown, trade count, and consistency across different periods.

---

Auto-trading and trading carry a risk of losing your principal. This article is educational, does not guarantee profit, and past backtest results do not predict future returns.

Footnotes

1. Michael Harris, "Look-Ahead Bias in Backtests and How to Detect It". ↩

2. Bailey, Borwein, López de Prado, Zhu (2014), "Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance", Notices of the American Mathematical Society. ↩

3. LuxAlgo, "Backtesting Limitations: Slippage and Liquidity Explained". ↩