Spurious Predictability in Financial Machine Learning

2026/04/16 by Sotirios D. Nikolopoulos · 1 voice · 1 citation
Economics, Econometrics and Finance · Mathematics · #q-fin.ST #stat.ME #stat.ML

paper · pdf

Abstract

Adaptive specification search generates statistically significant backtests even under martingale-difference nulls. We introduce a falsification audit testing complete predictive workflows against synthetic reference classes, including zero-predictability environments and microstructure placebos. Workflows generating significant walk-forward evidence in these environments are falsified. For passing workflows, we quantify selection-induced performance inflation using an absolute magnitude gap linking optimized in-sample evidence to disjoint walk-forward realizations, adjusted for effective multiplicity. Simulations validate extreme-value scaling under correlated searches and demonstrate detection power under genuine structure. Empirical case studies confirm that many apparent findings represent methodological artifacts rather than genuine predictability.

Citations

• Exact post-selection inference, with application to the lasso
• A Test for Superior Predictive Ability
• Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation

Cited by

• When Alpha Disappears: A One-Switch Benchmark for Decision-Time Leakage in Financial Backtests

Discussions

• Sotirios D. Nikolopoulos: Spurious Predictability in Financial Machine Learning https://arxiv.org/abs/2604.15531 https://arxiv.org/pdf/2604.15531 https://arxiv.org/html/2604.15531 [bsky, 0 points, 0 comments]

Related

• High-Dimensional Learning in Finance
• Limits To (Machine) Learning
• When Alpha Disappears: A One-Switch Benchmark for Decision-Time Leakage in Financial Backtests
• Interpretability in Safety-Critical FinancialTrading Systems
• Adversarial Robustness in Financial Machine Learning: Defenses, Economic Impact, and Governance Evidence