Backtesting:testing rules on the past.

Write the rules down so precisely that someone else would take the same trades: entry, stop, target or exit, position size, and when not to trade. Then step through historical data - manually, bar by bar, or with software - recording every trade the rules produce, including the losing ones.

Record the results in the account's own terms: net PnL, the largest drawdown, the longest losing streak, and the number of trades. A handful of trades is anecdote; most traders look for at least a hundred before drawing conclusions.

Most backtests look better than live trading. The usual reasons:

• Overfitting - tuning rules until they fit the past, so they describe noise rather than a repeatable effect.
• Look-ahead bias - using information that was not available at the time of the trade, such as a bar's close before it closed.
• Ignoring costs - leaving out commission, spread, swap and slippage.
• Selective samples - testing only on the period where the idea obviously worked.

Keep a block of data aside and test on it only once, after the rules are fixed. If results collapse there, the rules were likely fitted to the first sample. Forward testing - trading the fixed rules on a demo or simulated account in real time - is the next step, because it adds real execution and real decisions. Robustness testing asks whether the result survives small changes: nudge each parameter, try nearby markets and other periods, and see whether the edge holds or collapses.