Over-optimization is the trading term for overfitting: adjusting a strategy's rules and parameters so precisely to historical data that it captures noise rather than a lasting pattern. It is also called curve fitting. An over-optimized system can show excellent results in a backtest and then lose money once it trades live, because the quirks it was tuned to do not repeat.
The risk grows with the number of parameters tested and the number of variations tried on the same data. Warning signs include results that collapse when a setting is changed slightly, very few trades behind the headline returns and rules that only make sense for one market period. Common safeguards include keeping out-of-sample data aside, walk-forward optimization and favouring simple rules. The issue affects any rule-based trading strategy, including crypto trading bots sold on the strength of past performance.
Ref 1320000460