Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | English |
| Published: |
Zenodo
2026
|
| Subjects: | |
| Online Access: | https://doi.org/10.5281/zenodo.18700972 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866901492495024128 |
|---|---|
| author | Li, Haojun |
| author_facet | Li, Haojun |
| contents | <p>Prediction markets aggregate probabilistic beliefs about future events through continuous trading, yet coherent, rapid probability shifts—particularly those preceding public announcements—may signal early information access by informed participants. We present Polyoracle, a system for detecting statistically significant shifts in binary prediction market probability streams using a four-factor composite signal score: KL divergence, log-volume weighting, historical signal-to-noise ratio (SNR), and trajectory consistency (TC). The system incorporates a two-stage pre-filter tuned to the structural characteristics of Polymarket—the dominant decentralized prediction market venue—and employs cooldown deduplication with a deterministic zone override. Empirical evaluation on 500 live events (8,219 markets, 105,653 snapshots over a 6-hour observation window) demonstrates that the composite score achieves a 130× dynamic range between genuine signals and the noise floor. A factor ablation study confirms that all four components are necessary: removing volume weighting introduces 88% more false positives; removing SNR eliminates 61% of genuine signals.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18700972 |
| institution | Zenodo |
| language | eng |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Polyoracle V1: A Scoring System for High-SNR Anomaly Detection in Prediction Market Data Streams Li, Haojun Decision Support Techniques Forecasting Probability Data processing <p>Prediction markets aggregate probabilistic beliefs about future events through continuous trading, yet coherent, rapid probability shifts—particularly those preceding public announcements—may signal early information access by informed participants. We present Polyoracle, a system for detecting statistically significant shifts in binary prediction market probability streams using a four-factor composite signal score: KL divergence, log-volume weighting, historical signal-to-noise ratio (SNR), and trajectory consistency (TC). The system incorporates a two-stage pre-filter tuned to the structural characteristics of Polymarket—the dominant decentralized prediction market venue—and employs cooldown deduplication with a deterministic zone override. Empirical evaluation on 500 live events (8,219 markets, 105,653 snapshots over a 6-hour observation window) demonstrates that the composite score achieves a 130× dynamic range between genuine signals and the noise floor. A factor ablation study confirms that all four components are necessary: removing volume weighting introduces 88% more false positives; removing SNR eliminates 61% of genuine signals.</p> |
| title | Polyoracle V1: A Scoring System for High-SNR Anomaly Detection in Prediction Market Data Streams |
| topic | Decision Support Techniques Forecasting Probability Data processing |
| url | https://doi.org/10.5281/zenodo.18700972 |