Oracle3
Autonomous prediction market trading agent | Wang Transform pricing engine calibrated on 291K+ contracts | Kalshi · Polymarket · Solana DFlow · Jito bundles · 633 tests
Last verified:
What is Oracle3?
Oracle3 is an autonomous on-chain trading agent for prediction markets on Solana, Polymarket, and Kalshi. It combines LLM reasoning, quantitative signals, and atomic on-chain execution into a single autonomous system with no human in the loop. The tool operationalizes a peer-reviewed Wang Transform pricing model calibrated on 291,309 resolved contracts across six venues to systematically harvest the favorite-longshot bias through arbitrage detection and Kelly-sized model trades.
Key features include 8 on-chain agent capabilities (arbitrage, risk management, MEV protection, reputation, flash loans, and more), an AI + Quant hybrid using LLM agent strategies via OpenAI Agents SDK alongside adaptive quantitative strategies, multi-exchange support for Solana/DFlow, Polymarket CLOB API, and Kalshi REST API, cross-platform arbitrage to detect and trade price discrepancies, a live trading dashboard with real-time web UI showing equity curve and execution pipeline, dual-layer risk management with local limits plus Solana simulateTransaction pre-flight validation, and an on-chain audit trail where every trade is logged to Solana via Memo program.
Oracle3 is designed for quantitative traders, prediction market researchers, algorithmic trading developers, and crypto traders who want autonomous trading bots for binary prediction market contracts. It is particularly useful for users who want to exploit systematic pricing biases in prediction markets using academically-grounded pricing models.
The system deploys 8 constraint-based arbitrage strategies exploiting violated probability axioms (cross-market, exclusivity, implication, conditional, event sum, structural), statistical arbitrage via cointegration spread and lead-lag strategies, and model-driven strategies including fair value divergence and premium decay. The pricing engine uses hierarchical MLE with empirical Bayes priors and includes model Greeks for Kelly fraction and edge decay rate.
Oracle3 pricing
Pricing model: Freemium
Oracle3 is free and open source under the Apache 2.0 license. The software is available via PyPI (version 1.1.2) and can be installed via git clone and poetry install. There are no paid plans or subscription tiers. Users only pay their own trading capital, exchange fees, and any third-party service costs like OpenAI API usage. The code includes 633 tests and full CLI reference documentation at no cost.
Oracle3 pros
- Autonomous execution with no human in the loop required
- Wang Transform pricing calibrated on 291,309+ resolved contracts
- Multi-venue support across Kalshi, Polymarket, and Solana
- 8 constraint-based arbitrage strategies exploiting probability axioms
- Live trading dashboard with real-time equity curve visualization
- MEV protection via Jito bundle submission on Solana
- Dual-layer risk management with pre-flight transaction simulation
- On-chain audit trail via Solana Memo program for full transparency
- LLM agent strategies via OpenAI Agents SDK integration
- 633 unit tests with CI validation on every push
- Open source under Apache 2.0 license
- Event-driven async engine with snapshot persistence
- Automatic LIFO unwind on partial fills prevents naked positions
- Kelly criterion position sizing for optimal bet allocation
- Cross-platform arbitrage detects price discrepancies across exchanges
Oracle3 cons
- Requires technical expertise to set up with poetry and git
- Only supports 3 prediction market exchanges
- OpenAI API costs not included or subsidized
- Primarily designed for research and educational purposes
- Solana network congestion may affect transaction execution
- Requires API keys for Kalshi and Polymarket access
- No hosted cloud version available - self-hosted only
- Python-only implementation limits integration options
Frequently asked questions about Oracle3
What is Oracle3?
Oracle3 is an autonomous on-chain trading agent for prediction markets on Solana, Polymarket, and Kalshi. It combines LLM reasoning, quantitative signals, and atomic on-chain execution into a single autonomous system with no human in the loop. It uses a Wang Transform pricing model calibrated on 291,309 resolved contracts to systematically harvest the favorite-longshot bias.
What exchanges does Oracle3 support?
Oracle3 supports three exchanges: Solana/DFlow for on-chain execution, Polymarket via their CLOB API, and Kalshi via their REST API. This multi-exchange support enables cross-platform arbitrage to detect and trade price discrepancies across venues.
How does the Wang Transform pricing work?
The Wang Transform pricing engine uses the formula p_mkt = Φ(Φ^(-1)(p*) + λ) with λ̂ = 0.183. It employs hierarchical MLE with empirical Bayes priors where λi = 0.259 - 0.072 ln(1 + V) + 0.143 ln(1 + D) - 0.477 |p - 0.5|. The model includes model Greeks (∂p/∂λ, Kelly fraction, edge decay rate) and an online calibrator using hybrid batch MLE plus streaming EWMA with category shrinkage.
What arbitrage strategies does Oracle3 use?
Oracle3 uses 8 constraint-based arbitrage strategies: Cross-Market (same event, same price across exchanges), Exclusivity (P(A) + P(B) ≤ 1 for mutually exclusive events), Implication (P(A) ≤ P(B) when A implies B), Conditional (P(A|B) ∈ [L, U] within bounds), Event Sum (sum of outcomes equals 1), and Structural (P(A) = β·P(B) + α from calibrated model). It also uses statistical arbitrage via cointegration spread and lead-lag strategies.
How do I install Oracle3?
Install Oracle3 by running: git clone https://github.com/YichengYang-Ethan/oracle3.git && cd oracle3, then poetry install. After installation, you can browse markets with 'oracle3 market list --exchange solana --limit 10' or start paper trading with 'oracle3 dashboard --exchange solana --strategy-ref oracle3.strategy.contrib.adaptive_onchain_strategy:AdaptiveOnChainStrategy --initial-capital 10000'.
Is Oracle3 free to use?
Yes, Oracle3 is free and open source under the Apache 2.0 license. It is available via PyPI (version 1.1.2) and can be installed via git clone and poetry install. There are no paid plans or subscription tiers. Users only pay their own trading capital, exchange fees, and third-party service costs like OpenAI API usage.
What risk management features does Oracle3 have?
Oracle3 has dual-layer risk management: local position/drawdown/exposure limits plus Solana simulateTransaction pre-flight validation. It also includes correlation-aware risk with EWMA correlation matrix and effective exposure limits. The SpreadExecutor provides automatic LIFO unwind on partial fills to prevent naked multi-leg positions.
How does MEV protection work in Oracle3?
Oracle3 provides MEV protection through Jito bundle submission on Solana. This ensures transactions are bundled together and executed atomically, preventing front-running and other MEV extraction attacks that could negatively impact trading performance.
What is the live trading dashboard?
The live trading dashboard is a real-time web UI accessible at http://localhost:3000/live that shows the equity curve, feature cards, and execution pipeline. It provides real-time visualization of trading performance and allows users to monitor the autonomous agent's activities.
Who created Oracle3?
Oracle3 was created by Yicheng Yang (YichengYang-Ethan), a CS + Statistics + Economics student at UIUC with 4.0 GPA. The pricing model is based on Yang's peer-reviewed working paper 'Pricing Prediction Markets: Risk Premiums, Incomplete Markets, and a Decomposition Framework' (UIUC, 2026), which was calibrated on 291,309 resolved contracts across six venues.