What a backtest tells you
Whether the rules behave as written across many market conditions, and roughly how often they trade. It cannot tell you future results, and it is only as good as its data, spread and slippage assumptions.
Three ways backtests mislead
- Look-ahead bias: the code uses a value that was not known at the time, such as the candle’s close before it closed.
- Repainting: indicators like ZigZag or fractals change past values, so old signals look better than they were.
- Overfitting: parameters tuned until the past looks perfect rarely hold up on new data.
What a forward test adds
Real spreads, real latency, broker rejections and the platform’s actual behaviour. Two weeks on demo usually surfaces execution problems that no backtest shows.
How to read the results
Compare trade by trade with your rules rather than looking only at profit. A robot that follows your rules exactly is working; whether those rules make money is a separate question no test can promise.
This guide explains software and testing. It is not investment advice.
