Data Snooping

Data Snooping - A statistical bias that occurs when one exhaustively searches through data for profitable patterns, often without a predefined hypothesis. In trading, data snooping (or data dredging) means testing many combinations of rules until one “works” on historical data, by chance. This inflates false positives: patterns that won’t hold in future data. Proper research practices (a priori rules, separate test sets) are used to mitigate data snooping bias.