Backtesting is supposed to talk you out of bad ideas. In crypto it often does the opposite — a flawed backtest takes a fragile strategy and dresses it up as precise, profitable, and ready to scale.

Backtesting itself isn’t the issue. Unrealistic backtesting is. Crypto markets come with violent volatility, liquidity that moves around, quirks specific to each exchange, funding payments, outages, delistings, and regime shifts that arrive without warning. Build a test that ignores all of that and the pretty equity curve you get back isn’t evidence of anything.

Mistake 1: Ignoring Fees and Spreads

A lot of strategies live or die on a small edge, and fees plus spreads are exactly what eats that edge alive.

A strategy that trades frequently must include:

  • exchange trading fees
  • maker/taker differences
  • bid-ask spread
  • slippage
  • funding costs for perpetual futures
  • withdrawal or transfer costs if relevant

A backtest can show 12% a year before costs, and if the strategy trades aggressively the number you actually live with might be negative.

Mistake 2: Survivorship Bias

Crypto is a graveyard of dead tokens. Test only the coins that made it to today and you’ve quietly stacked the deck — the past looks far kinder than it was.

A realistic universe should account for:

  • delisted tokens
  • low-liquidity periods
  • tokens that collapsed
  • symbols that changed
  • exchange listing dates

Backtesting only on today’s big winners hands you a cleaner history than any trader actually faced at the time.

Mistake 3: Overfitting Indicators

Overfitting is what happens when you keep tweaking a strategy until it hugs the historical data a little too lovingly.

Warning signs you’ve done it:

  • too many parameters
  • rules added after each losing period
  • perfect-looking equity curve
  • performance depends on one exact setting
  • no out-of-sample validation
  • strategy logic is hard to explain

The tighter you tune a strategy to the past, the less chance it has of surviving the future.

Mistake 4: Treating Candle Prices as Tradable

Backtests often assume trades happen at open, close, high, or low prices. Real orders do not work that cleanly.

If your strategy somehow buys the exact low of a candle or sells the exact high, be suspicious — real fills don’t land like that. Even close prices flatter you when the signal itself only becomes known after the candle has closed.

The test should model when the signal becomes available and what price could realistically be filled afterward.

Mistake 5: Ignoring Liquidity

Liquidity in crypto can vanish in an afternoon. Something that hums along on Bitcoin can choke on a smaller token, and something that works fine with $500 can fall apart at $50,000.

Check:

  • average volume
  • order book depth
  • spread
  • time of day effects
  • exchange concentration
  • size as a percentage of volume

Capacity matters as much as return. A high percentage gain on tiny size doesn’t make a strategy scalable.

Mistake 6: Forgetting Funding Rates

For perpetual futures, funding is part of the return. Long or short positions can pay or receive funding depending on market conditions.

Ignoring funding makes some strategies look better than they are. In crowded trades, funding can become a major cost.

See funding rates, open interest, and market regimes for a practical explanation of how those variables affect interpretation.

Mistake 7: Testing One Market Regime

Crypto changes regime sharply:

  • bull markets
  • bear markets
  • sideways chop
  • high-volatility crashes
  • low-volatility accumulation
  • exchange stress periods
  • regulatory shocks

A strategy tested only during one regime is incomplete. Mean reversion, trend following, breakout, and carry-style trades all behave differently across regimes.

Mistake 8: No Walk-Forward Testing

An in-sample backtest tells you how the strategy performed on the data used to design it. That is not enough.

Use:

  • in-sample period for development
  • out-of-sample period for validation
  • walk-forward testing for robustness
  • paper trading for operational checks
  • small live trading for final reality test

Each stage should make it harder for the strategy to fool you.

Mistake 9: Optimizing for Return Only

A high return is worthless if you can’t actually stomach the path that gets you there.

Evaluate:

  • maximum drawdown
  • longest losing streak
  • volatility
  • recovery time
  • tail losses
  • trade frequency
  • dependency on one asset
  • dependency on one year

A lower-return strategy with smaller drawdowns and cleaner execution may be more practical than a high-return strategy that requires extreme tolerance.

FAQ

Is a profitable backtest enough to trade?

No. It is only one stage. You still need realistic costs, out-of-sample validation, operational testing, and risk limits.

How do I know if a strategy is overfit?

If small parameter changes destroy performance, the logic is hard to explain, or results depend on one historical period, be skeptical.

Should I include black swan events?

You should at least test stress scenarios. Crypto has repeated exchange, liquidity, and regulatory shocks.

Bottom Line

Treat a backtest as an attempt to kill the strategy, not sell it to yourself. Throw in the friction, the dead assets, the liquidity limits, the funding, the regime changes, and execution that behaves like the real thing. If it’s still standing after all that, maybe it’s earned a small live test. If it only shines in a clean spreadsheet, that’s exactly where it should stay.