Cherry-picking can be deliberate or accidental. A course seller who shows three winning screenshots is choosing. A trader who remembers the times a pattern worked is being chosen for, by memory.
Backtests are the most common venue. Start date shifted, losing instruments excluded, a rule added after seeing the failures. Each individual choice looks reasonable, and together they manufacture an edge that does not exist outside the sample.
The check is completeness. Report every trade the rule would have taken over the whole period, including the ugly stretch, or accept that the result is decoration rather than evidence.
Related: motivated-reasoning, survivorship-bias-in-advice, base-rate-neglect, gain-porn