v1.0 — replication package for arXiv v1

First public release of the replication package for The Anatomy of a Decentralized Prediction Market: Microstructure Evidence from the Polymarket Order Book.

What's included

• Compiled paper (paper/build/anatomy.pdf)
• Full LaTeX source, figures, tables, and bibliography (paper/)
• Python package for orderbook replay, on-chain trade joining, microstructure measures, and stylized-fact computations (polydata/)
• End-to-end pipeline scripts (scripts/)
• pytest suite (tests/)
• 600-market pre-registered panel + per-market measure parquets (data/panel.parquet, data/panel_metadata.parquet, data/panel_trade_measures.parquet, data/panel_quote/*)
• Stylized-fact tables and figures (artifacts/sf*.{parquet,png}), STRICT-vs-on-chain comparison panels, Glosten-Harris decomposition, sign-agreement robustness, and sample-step sensitivity grid
• Locked pre-registration document (docs/preregistration_plan3c.md) with the panel SHA-256

What's not included

• The 30-billion-event raw WebSocket archive (623.8 GB across 1,262 hourly Parquet files). Acquisition is described in §3.1 of the paper; the collector configuration is sufficient to reproduce.
• The 14 GB on-chain OrderFilled scrape (242 slice Parquets). Reproducible via scripts/scrape_onchain_fills.py against any Polygon archive RPC provider.
• The 99 MB clob_token_map.parquet token mapping (over GitHub's 100 MB file limit). Regenerate with scripts/pull_token_ids_clob.py (~6 minutes).

Headlines

• 8 cross-sectional stylized facts on a pre-registered 600-market panel.
• Trade-direction inference from Polymarket's public WebSocket feed agrees with on-chain ground truth on only ~59% of comparable buckets.
• This translates to a 67% sign-flip rate on the effective half-spread and a 60% sign-flip rate on Kyle's λ between feed-inferred and on-chain estimates within the comparable subset of the top-100 panel.

A DOI is being minted for this release via Zenodo and will be added to the README + paper in a follow-up release.