Backtesting: did the rules work in the past?
A backtest runs your rules over historical prices. It is fast, cheap and repeatable, which makes it ideal for discarding weak ideas and comparing variations. Its weakness is that you already know what happened, so it is easy, even unintentionally, to shape the rules to fit it.
Forward testing: do they hold up on new data?
A forward test runs the same rules on current market data as it arrives, with simulated orders and notional capital, so no real money is at risk. Because the future has not happened yet, you cannot have tuned the rules to it. It is the closest honest check before real orders.
Forward testing takes time. A week or two of data is rarely enough for a strategy that trades infrequently, so be patient and judge it by trades observed, not days elapsed.
Side by side
- Data: backtesting uses historical data; forward testing uses data arriving now.
- Speed: a backtest takes seconds or minutes; a forward test takes weeks.
- Main risk: backtests invite overfitting; forward tests can be too short to mean much.
- Money at risk: neither one uses real money. Live trading does.
- Best use: backtesting to filter ideas; forward testing to confirm survivors.
Why live results still differ
Even a strategy that passes both can behave differently with real orders. Fills may be worse than simulated, a broker session may drop, and size changes market impact. This is why many people start live trading very small and increase only after the real numbers match the tested ones.
A practical workflow
- Backtest the idea over several periods with realistic costs.
- Fix the rules and write them down.
- Forward test those exact rules without changing them.
- Compare forward results with the backtest and note any gap.
- Only then consider live orders, at a size you can comfortably lose.