Velocity- and Regime-Aware Detection of Intraday Options Market Manipulation, with Explainable Attribution

Battle Count: 60

Highly relevant to market surveillance, compliance, and risk management in quantitative trading. The velocity-based features and pump-reversal shape detection are directly applicable to monitoring trading activity for manipulation. However, the paper focuses on regulatory detection rather than alpha generation or trading strategy development. The findings about regime conditioning, label incompleteness, and the distinction between shape and magnitude are valuable for quantitative risk teams. The cross-market transferability of the dynamic signature is relevant for multi-asset surveillance systems.

Moderate complexity. The core pipeline involves: (1) computing Black-Scholes Delta from option quotes, (2) smoothing and computing velocity features, (3) training an autoencoder on 3 features, (4) HMM fitting with BIC selection, (5) exact SHAP computation (tractable with 3 features = 8 coalitions). The time-partitioned protocol and threshold calibration add procedural complexity. The cross-market shape score is simpler. Main challenges: obtaining minute-level options data, correct Delta computation, and handling the label structure (positive-only enforcement labels).

The paper uses a public Kaggle dataset for BANKNIFTY data and references a frozen working copy in a project archive. The SEC v. Patel complaint is public. However, the specific enforcement order (SEBI interim order) is referenced but the exact data pipeline code is not publicly linked. The authors state all results come from a single checkpointed run. SHAP computation is exact (8 coalitions for 3 features). Delta computation validated against vendor Greeks (Pearson r >= 0.9995). No GitHub repository is explicitly provided.