Maximum Adverse Excursion (MAE) as an Algorithmic Objective & Diagnostics Tool

Definition recap

Given an open position opened at \(t_0\), MAE measures the worst unrealised loss excursion before flat:

\[ \text{MAE} = \min_{\tau \in [t_0, t_{\mathrm{exit}}]} \big( P(\tau) - P(t_0) \big) \]
(price scaled by contract factor). Positive MAE magnitude means deep underwater travel even if trade ultimately wins.

Why expectancy alone conceals pathology

Two strategies might share identical average R yet differ wildly in MAE distributions — one cleanly mean-reverts, the other survives via occasional bailouts beyond risk tolerance.

Optimisation viewpoint

Including MAE percentile constraints in calibration (e.g. cap 90th percentile MAE relative to median winner) aligns risk of journey with risk of outcome.

Implementation notes

Rolling MAE dashboards per setup provide explainability feeds for allocators auditing robot behaviour historically.

Finvestopia context

Bots page metrics pair recovery style stats with directional bias context — comparing MAE vs MFE (Maximum Favourable Excursion) communicates whether entries are fundamentally early vs late.