Crypto backtesting: the data traps nobody warns you about

Why crypto is harder to backtest than forex

Forex has one dominant venue and a few liquid pairs; crypto has hundreds of exchanges with different bars, different volume and different fees. There is no single 'truth' price — your backtest inherits whatever feed you chose, and the choice matters more than in any other market.

On top of that, crypto's history is short and regime-driven: a strategy tuned on 2021's bull run is a strategy built on one weather report. Backtesting crypto honestly means confronting these problems explicitly, not assuming them away.

Trap 1 — Survivorship bias

Today's top coins are the ones that survived. If you backtest only the current top-10 list, you include the winners and exclude every coin that died along the way — an automatic, invisible boost to your results.

The fix: define your universe as of the test date, include delisted coins where data exists, and be suspicious of any backtest that cherry-picks the assets that happened to thrive.

Trap 2 — Exchange data is not interchangeable

Binance, Coinbase and Bitstamp produce different bars for the same pair: different volumes, different session behavior, occasionally different closes. A strategy that works on Binance H1 data may not survive Coinbase data — and live trading happens on your broker's feed.

The fix: state the exchange and data source on every backtest, and cross-check the strategy on a second exchange before believing it.

Trap 3 — Fees and funding

Crypto costs are brutal when modeled properly: spot taker fees, perpetual funding rates that charge you every 8 hours, and withdrawal costs. A scalping strategy can look great gross and be dead net — the model must subtract these on every trade and every holding period.

The fix: model taker fees plus funding for any position held past a funding timestamp. If the platform has no funding line, the backtest is incomplete.

Trap 4 — The 24/7 liquidity illusion

Crypto never closes, but liquidity collapses on weekends and in thin altcoins. A strategy that assumes instant fills at the close price all week will bleed on Sunday fills. High-frequency strategies are the first victims.

The fix: model slippage generously, especially for weekend and altcoin entries, and paper trade through a full weekend cycle before live.

A crypto backtesting checklist

• Use at least one full market cycle (3-4 years) covering bull, bear and range regimes.
• State the exchange, data source and date range for every backtest.
• Include taker fees, funding and withdrawal costs in the model.
• Define your coin universe as of each test date; beware of survivorship.
• Validate out-of-sample or walk-forward; never trust a single full-sample curve.
• Paper trade on a demo account across a weekend before scaling live.