Backtesting and forward testing
Test a strategy or a signal source against history and against live data — and read the result honestly.
Backtesting
Backtesting replays historical data through a strategy with your risk settings applied, so the equity curve reflects the sizing you would actually have used.
- Pick the instrument, the period and the risk settings to apply.
- Results show the trade list, equity curve, max drawdown and expectancy.
- Spread and commission assumptions are part of the run — change them and the curve changes, which tells you how fragile the edge is.
Forward testing
Forward testing runs the same logic on live incoming data without sending orders. Slower than a backtest, and much harder to fool.
| Backtest | Forward test | |
|---|---|---|
| Speed | Years in minutes | Real time |
| Overfitting risk | High — you can tune until it looks perfect | Low — the data has not happened yet |
| Slippage realism | Modelled | Real quotes, still no real fills |
| Good for | Rejecting bad ideas fast | Confirming a surviving idea |
Reading the result
- Look at max drawdown before you look at profit. That is the number you have to live through.
- A curve that only works on one instrument in one year is a coincidence, not a strategy.
- Compare against doing nothing, and against a flat fixed-size version of the same trades.
Past performance says nothing about future results. Backtests exclude the two things that actually hurt: real slippage in fast markets, and your own behaviour during a losing month.
Toujours bloqué ?
Envoyez le signal brut, l'horodatage et le nom du compte. N'envoyez jamais de mots de passe, de clés API ni d'identifiants de courtier.