Backtest
Testing trading rules on historical data: how a portfolio would have behaved had decisions in the past been taken by those rules.
How to read the number
The worth of the result is decided by the honesty of the conditions: whether commissions and spread were counted, whether the security traded at all that day, whether the algorithm knew only what was known at the moment of the decision.
When the metric lies
The past can always be described by a rule fitted to it. Two quiet distortions come on top: issuers that left the exchange are missing from the sample, and data revised after publication is used as if it had been available.
Also known as: historical simulation of rules