A mathematical transform captures the geometry of stock market data to predict prices 56% better than expert traders.
April 25, 2026
Original Paper
Path Signatures as Universal Feature Extractors for Limit Order Book Mid-Price Prediction
SSRN · 6635378
The Takeaway
Path signatures act as universal feature extractors for high-frequency trading. This method removes the need for humans to manually identify important market indicators. The algorithm identifies the essential shape of price movements to predict where they are going next. It consistently outperforms standard hand-crafted features used by quant funds for years. Wall Street can now replace tedious manual engineering with pure geometric math.
From the abstract
We apply the path signature transform — a canonical lossless encoding of multidimensional time series rooted in Chen's theorem (1954) and Rough Path Theory (Lyons 1998) — to the joint trajectory of bid price, ask price, bid volume, and ask volume extracted from limit order book (LOB) data. Truncating the signature at order 4 yields a 341-dimensional feature vector that constitutes a sufficient statistic for predicting the next mid-price direction without any manual feature engineering. Evaluated