Hidden Order in Trades Predicts the Size of Price Moves

Battle Count: 78

Highly relevant for intraday volatility timing and regime detection. The finding that entropy predicts magnitude but not direction is theoretically grounded and practically useful for asymmetric-payoff strategies (e.g., straddles, gamma scalping, stop-loss-based entries). However, the extremely limited sample (36 days, single instrument), concentration risk, and lack of live execution validation temper practical applicability. The theoretical framework (permutation invariance of entropy) is elegant and generalizable. The 2.89x magnitude amplification at low entropy is a strong signal if it persists out-of-sample over longer periods.

Moderate complexity. Requires tick-level data processing, rolling Markov transition matrix estimation (15×15), eigendecomposition for stationary distribution, and entropy computation at second resolution. The trading rule itself is simple (threshold-based entry with stop-loss/timeout). Main challenges: real-time second-resolution data pipeline, maintaining rolling windows, and ensuring low-latency computation for intraday use. No ML training required—purely statistical/information-theoretic.

Analysis code available upon request but not publicly hosted. Data from commercial tick-data vendors requiring licensing. Methodology is well-described with explicit formulas, but the 36-day single-instrument sample and proprietary data source limit independent replication. No GitHub repository provided.