Automating Crypto Factor Mining with IC Validation and Adaptive Weights

Summary

This document outlines an automated research and execution workflow for mining factors across liquid crypto perpetual contracts. At scheduled intervals, it selects markets by trading volume, classifies volatility conditions, reviews the health and coverage of existing factors, asks an AI system for candidates, and evaluates them using cross-sectional Information Coefficients. It then filters overlapping signals, combines the survivors into weighted scores, and adjusts weights as recent IC weakens. Separate slow and fast triggers handle factor research and position protection.

The described validation uses historical walk-forward alignment to avoid using future candles when computing signals. Live observations from a two-day run show that candidate factors’ recent ICs declined and that the system sought replacements, but the document explicitly treats this as evidence that the workflow ran as designed, not that it predicts profit. The short observation period cannot establish durable effectiveness, and the method’s claims remain limited by factor decay and the need for longer testing.

Key ideas

• The workflow automates factor generation, historical validation, redundancy filtering, signal synthesis, and execution.
• It screens perpetual contracts for liquidity and uses Bitcoin volatility to characterize market conditions.
• Walk-forward IC validation is used to assess factor rankings against subsequent returns while avoiding look-ahead bias.
• Recent IC weakness reduces a factor’s influence and guides exploration toward less-covered dimensions.
• The reported two-day operation demonstrates process execution but is too short to establish sustained predictive performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.