Empirical Evaluation of Deadline-Resolved Information Leakage on Documented Polymarket Insider Cases

By Maksym Nechepurenko

Rating

1651
Battle Count: 79

Relevance

5/10
The paper addresses informed trading detection in prediction markets (Polymarket), which is adjacent to quantitative trading but not directly about traditional asset pricing, portfolio construction, or algorithmic execution. The ILSdl framework could inform prediction-market trading strategies and regulatory surveillance. The hazard-rate estimation and survival analysis components have transferable methodology to quantitative finance. However, the paper is primarily a methodological proof-of-concept with n=1 empirical case, limiting immediate practical applicability for trading strategy development. The cross-market wallet analysis and blockchain forensics aspects are more relevant to compliance and surveillance than to alpha generation.

Implementation Complexity

8/10
The full pipeline requires: (1) LLM-assisted multi-source event timestamp recovery with web-search tool access and cross-verification across 3+ independent sources; (2) resolution-typology classification of a 911,237-market corpus; (3) per-category exponential hazard-rate fitting via MLE with KS goodness-of-fit testing; (4) CLOB price data collection from market opening (infrastructure-intensive); (5) on-chain trade history retrieval via Polymarket subgraph indexer; (6) cross-market wallet overlap analysis with Herfindahl-Hirschman concentration metrics; (7) scope-condition compliance checks (edge-effect, anchor-sensitivity, positive-tau). The infrastructure requirements (continuous CLOB collection, per-trade collection from T_open) are the primary complexity drivers and are identified as the binding constraints on scaling.

Reproducibility

3/5
The paper releases three accompanying datasets on GitHub (T_news/T_event recovery corpus, per-category hazard fits, population-scale ILS corpus). The methodology is specified end-to-end with formulas and pipeline steps. However, the empirical sample is extremely small (n=1 for the main ILSdl computation, n=18 for hazard estimation), limiting statistical reproducibility. The LLM-assisted T_event recovery pipeline (Claude Haiku 4.5 with web search) introduces non-determinism. The Polymarket subgraph indexer dependency and CLOB price coverage gaps are infrastructure-specific constraints that may not be reproducible by other researchers without equivalent data access.

About this paper

Methodology: Deadline-Resolved Information Leakage Score (ILSdl) with Exponential Hazard-Rate Estimation. Problem types: Survival Analysis, Anomaly Detection, Density Estimation, Classification, Risk Management.

The interactive Everscope explorer (charts, battles, favorites) loads below.