Valuation Tearing and Deep Drawdowns: A Leading Measure and a Three-Layer Decision System for China's A-Share AI Economy (frozen datasets and construction code)

Verification-level reproduction materials for the paper of the same name. Data window 2024-01-01 to 2026-08-26, 642 trading days. data/ — 19 frozen datasets: the ledger of deep drawdown episodes (12 episodes), the daily series of the three decision quantities, the valuation tearing series (gap) between technology and traditional industries, the industry-level return and drawdown series behind the panel test together with its results, the calibration of every threshold, and snapshots of the industry and sector classifications. All are taken from the sealed area of the production system. code/ — 16 construction scripts: episode identification, the causal construction of the valuation temperature T, the construction of gap and the assessment of its forward-looking discriminative power, the quarterly capitalisation series with its coverage gate, and the three decision quantities. reproduce.py reads only the data in this repository, recomputes the paper's main numbers and compares them item by item against the values printed in the text. Python standard library only, no third-party dependencies. Thresholds are not hard-coded. All 14 thresholds are collected in data/gate_params.json, each with its provenance and the method by which it was calibrated. A reader can recompute everything on the paper's own terms, or recalibrate on data of their own. Boundary of reproducibility: this repository supports verification-level reproduction, meaning every number printed in the paper can be recomputed from the data released here. It does not support rebuilding all intermediate quantities from raw market data, which would require roughly 145 MB of intermediate files and gigabytes of raw quotes, and would depend on the continued historical availability of third-party data interfaces. Disclaimer: the outputs of this repository are statistical descriptions of market states, not investment advice. They contain no buy or sell recommendations, no price targets and no position sizing.