Curve fitting to a single news spike

[Amanhasnoname]

One huge day can dominate a backtest. A big move during a central bank surprise might produce a trade worth fifty normal ones, making the whole test look profitable.

Removing the top five or ten trades and re-running the numbers reveals the dependence. If the result flips from profit to loss, the strategy relies on rare events and cannot be judged as a regular edge.

Even worse is tuning parameters so that the strategy catches that specific spike. The settings fit one day of history, and nothing suggests the pattern will repeat.

Better practice is checking the result in segments, such as month by month or week by week, and making sure profit is spread out. A strategy with consistent small gains inspires more confidence than one with a single large outlier.

If the aim is to capture rare events, the strategy should be designed and sized for that, and not disguised as a scalping edge.

Does the strategy still work after the best trades are removed?