Research tooling for prediction markets.
Poly-Alpha is an offline research and paper-trading platform for prediction-market strategies: provenance-tagged data contracts, uncertainty-aware research, strategy comparison, and risk analysis. It does not execute live trades.
Research Tooling for Quantitative Prediction Markets.
Poly-Alpha is a research platform for prediction-market strategies. It provides shared market and asset data contracts, provenance-tagged research with uncertainty intervals, in-sample strategy comparison, and portfolio risk analysis — all offline and deterministic.
The platform ships deterministic fixture and synthetic datasets alongside an optional real Polymarket market reader. Notes carry sources and caveats; fixture, simulated, and synthetic values are always labeled and are never presented as observed data.
It reads Polymarket market data only. There is no Kalshi or Manifold integration, no order placement, no fund movement, and no validated performance or Sharpe figure. Backtest metrics are in-sample and are not forecasts.
Reason
A deterministic heuristic derives a probability estimate from the de-vigged market price, order-book imbalance, and liquidity, and reports it with an uncertainty interval.
Optimize
A dedicated risk agent applies Kelly Criterion sizing and Shin debiasing to determine optimal position sizes that protect against systemic ruin.
Report
Results are exported as a provenance-labeled Markdown dossier and JSON. The platform reads Polymarket market data only; it does not submit orders.
Research Platform Capabilities
Offline, deterministic, and provenance-labeled from end to end.
Deterministic Research Engine
The platform is a deterministic, offline heuristic over labeled data — not an LLM. It produces provenance-tagged notes with cited sources and explicit caveats, and it makes no claim of live news ingestion or autonomous trading.
Deterministic Tooling
Market data is read from Polymarket; there is no live order execution or low-latency trading claim.
Self-Correcting Risk
Dedicated risk agents continuously monitor portfolio health, employing self-correction heuristics to halt trades and prevent ruin.
Goal-Oriented Optimization
Run deterministic research over labeled fixture, synthetic, and optional real markets; screen candidates by a conservative edge bound, size them with capped fractional Kelly, and analyze portfolio risk. Every estimate carries provenance and an uncertainty interval.
How It Runs
Plain CLI and HTTP commands over deterministic, labeled data. No live orders, no keys, no network by default.
# List labeled markets and rank them by conservative edge
poly-alpha research markets --all
poly-alpha research overview
# In-sample comparison, paper simulation, and risk
poly-alpha research compare
poly-alpha research simulate
poly-alpha research risk
# Provenance-labeled dossier + read-only JSON API
poly-alpha research report --output dossier.md
poly-alpha research serve --port 8000Built by Researchers
Founded by a specialized team of machine learning and quantitative systems engineers.
Transparent SaaS Pricing
Usage-based research tooling with predictable infrastructure costs.
Illustrative concept copy only — no product is for sale and no pricing is offered.
Developer
For independent researchers.
- 1 Research Workspace
- 1M API Requests / month
- Read-only REST API
- End-of-Day Data Sync
Professional
For prop desks and small funds.
- 5 Research Workspaces
- Offline research pipeline
- Unlimited API Requests
- Research journal export
- Priority Slack Support
Enterprise
For institutional capital.
- Unlimited Workspaces
- Petabyte Historical Data Lake
- Custom Debiasing Parameters
- On-Premise Deployment
- SOC-2 Audit Reports
Explore the Research Platform
Run the offline research, comparison, and risk tooling described in the repository README.