Overfitting hides easily in one-minute parameters

[Amanhasnoname]

A strategy with ten adjustable parameters and a year of one-minute data can be tuned to look brilliant. The settings fit the quirks of that specific history, not any lasting pattern.

Warning signs include sharp peaks in the parameter space, where moving one value slightly destroys the result. A robust strategy shows broad plateaus, where nearby settings perform similarly. If only one exact combination works, it is almost certainly noise.

Too many rules is another warning. Every added filter improves the backtest and reduces the chance of working forward. Strategies that need a long list of conditions to be profitable often trace to data mining.

Simple tests help. Running the strategy on a different pair or period without re-optimising shows if the logic carries over. A sharp drop is a red flag.

Keeping the rule count low and the parameters few makes overfitting harder.

How many parameters does your strategy use, and what happens if the values are nudged slightly?