Backtest Audit Checklist Before Running an EA

A backtest that looks amazing on a screenshot is often the product of overfitting — tuning the strategy so precisely to past data that it has no predictive power going forward. The equity curve looks beautiful because the strategy was built to fit that exact history.

Before you trust an EA with real risk, run it through this audit. The goal is not to find a perfect backtest — it's to find the flaws that would cost you money live.

• Use tick data, not just M1 bars — M1 or higher-timeframe bars hide intra-bar slippage and gaps.
• Check for data gaps — missing periods can make a strategy look better than it is.
• Verify the spread model — a backtest with a fixed, tiny spread overstates profitability on a real, variable-spread market.

• Test out-of-sample — optimize on one period, then validate on a period the optimizer never saw.
• Keep the parameter count low — the more parameters you tune, the easier it is to overfit.
• Be suspicious of "perfect" results — a strategy with a 99% win rate and no drawdown is almost certainly curve-fit.

• Add realistic slippage — a few pips per trade on a liquid instrument, more on illiquid ones.
• Model commission and swap — costs that the backtest ignores will eat into live results.
• Check fill assumptions — does the backtest assume every order fills at the exact price? That rarely happens live.

The profit factor matters less than the drawdown profile — because drawdown is what ends accounts.

• Look at max drawdown — is it within what you (and your prop firm) can survive?
• Check the worst losing streak — how many consecutive losses can the strategy take?
• Compare max drawdown to the daily loss limit of your prop firm — a strategy that can draw down 8% in a day is dangerous on a 5% daily limit.

1. Does the strategy still work out-of-sample?
2. Would it survive a realistic spread, slippage, and commission?
3. Can its worst drawdown fit inside your prop firm's limits?
4. Is the edge based on a real, explainable mechanism — or just a curve fit?
5. How would it behave in a market regime it hasn't seen?

• A backtest is a historical simulation, not a guarantee.
• Most "amazing" backtests are overfit — test out-of-sample and keep parameters low.
• Model realistic slippage, spread, commission, and swap — costs the backtest ignores will eat live results.
• The drawdown profile matters more than the profit factor — because drawdown is what ends accounts.