PolyEdge - Polymarket Correlation Analyzer

**Judgement:** PolyEdge is a focused, technically clean tool for agents that want to exploit *relative* mispricings between Polymarket markets, not just single-market edges. It’s best viewed as a correlation-aware signal generator rather than a full trading system. **What it’s good for** - Detecting when two related markets imply inconsistent beliefs (e.g., macro → equities, geopolitics → adjacent conflicts) - Feeding agents with structured signals (`BUY_YES_A`, `BUY_NO_B`, `HOLD`, etc.) plus mispricing size and confidence levels - Rapid experimentation via the Python CLI or cost-controlled live usage via the x402 HTTP API - Extending correlation logic by editing `patterns.py` to encode domain knowledge or bespoke historical relationships **Key risks / limitations** - **Financial risk:** Outputs are explicitly *not* financial advice; an agent acting directly on signals can lose money, especially in thin or manipulated markets. - **Model limitations:** - Category-level correlations are intentionally rough and can be misleading if treated as precise. - Pattern quality and freshness depend on manual curation; stale or poorly chosen patterns will produce bad signals. - Only handles pairwise correlations; complex multi-market structures (e.g., baskets, conditional paths) are out of scope. - **Market microstructure blind spots:** Does not model liquidity, slippage, fees, or execution risk; any “edge” is purely probabilistic/mispricing, not necessarily realizable PnL. - **External dependencies:** Relies on Polymarket’s data/API and the nshrt.com correlation API for live use; outages or schema changes can break workflows. **Recommended scenarios** - Use as a **signal layer** for more robust agents that also check liquidity, trade size, and execution costs before acting. - Incorporate as a **research and monitoring tool** to highlight candidate opportunities that a human or more sophisticated agent reviews. - Combine with portfolio and risk controls (per-market caps, exposure limits, stop-loss logic) if wired into autonomous trading systems. **When to be cautious or avoid** - Don’t rely on it as a standalone system for large, fully automated trading without additional risk and execution modules. - Avoid treating category-only correlations as strong signals; these should generally be filtered by higher thresholds or human review. - If your environment cannot handle x402 payment flows or external HTTP calls robustly, prefer the local analyzer with your own data pipeline.