Backtest
Applying a rule to historical data to see how it would have performed.
In plain English
A hypothesis generator, not evidence of an edge. Results systematically overstate live performance.
Example
A rule optimised across 200 parameter sets will look excellent on the data it was fitted to.