oracle3

Name: oracle3
Author: YichengYang-Ethan

Verified

Autonomous prediction market trading agent | Wang Transform pricing engine calibrated on 291K+ contracts | Kalshi · Polymarket · Solana DFlow · Jito bundles · 633 tests

264stars

32forks

Python

Added 3/10/2026

View on GitHub Download ZIP Scan for vulnerabilities

30 days in the Featured rail · terms & refunds

AI Agents Sports Betting & Prediction Markets Hubai-agentalgorithmic-tradingarbitrageautonomous-agentdefi-tradingdflowhierarchical-mlejito-bundleskalshikelly-criterionllmon-chainpolymarketprediction-marketprediction-marketspythonquantitative-financesolanatrading-botwang-transform

Installation

# Add to your Claude Code skills
git clone https://github.com/YichengYang-Ethan/oracle3

Getting Started

Guides for using ai agents skills like oracle3.

Security ReportVerified

Last scanned: 5/30/2026

{
  "issues": [],
  "status": "PASSED",
  "scannedAt": "2026-05-30T15:24:46.772Z",
  "npmAuditRan": true,
  "pipAuditRan": true
}

README.md

Oracle3

Autonomous prediction market trading agent across Kalshi, Polymarket, and Solana.

Why this exists

Prediction markets price binary contracts at systematically biased levels — a true 50/50 contract typically trades around 0.57 (favorite-longshot bias, $\hat{\lambda} \approx 0.183$). Most trading bots ignore this distortion entirely. Oracle3 operationalizes a peer-reviewed pricing model, calibrated on 291,309 resolved contracts across six venues, to systematically harvest the bias through arbitrage detection and Kelly-sized model trades.

This system deploys the exact $\lambda$ estimates and covariate model from prediction-market-pricing (Yang, 2026) as its real-time pricing engine.

How oracle3 differs from existing prediction-market tools

	Oracle3	polymarket-whales	prediction-market-maker	py-clob-client
Pricing model	Wang Transform (calibrated MLE)	None	Bid-ask MM	None
Constraint-based arbitrage	8 strategies	None	None	N/A
Multi-venue	Kalshi + Polymarket + Solana	Polymarket only	Polymarket only	Polymarket only
On-chain execution	Solana via DFlow + Jito	No	No	N/A (SDK)
Working paper	Yang (2026), SSRN	No	No	No
Tests	633	0	0	50+
License	Apache 2.0	MIT	MIT	MIT

Architecture

graph TD
    A[Wang Transform Pricing Engine<br/>MLE coefficients from paper] --> B[Fair Value Estimator<br/>Model Greeks · Kelly Sizing]
    B --> C[Strategy Layer]
    C --> D[8 Constraint-Based Arbitrage]
    C --> E[2 Model-Driven Strategies]
    C --> F[LLM Agent Strategies]
    D --> G[Trading Engine<br/>SpreadExecutor · Risk Manager · Position Tracker]
    E --> G
    F --> G
    G --> H[Kalshi]
    G --> I[Polymarket]
    G --> J[Solana / DFlow]

Strategies

Constraint-based arbitrage — each exploits a violated probability axiom:

Strategy	Invariant
Cross-Market	Same event, same price across exchanges
Exclusivity	$P(A) + P(B) \leq 1$ for mutually exclusive events
Implication	$P(A) \leq P(B)$ when A implies B
Conditional	$P(A \mid B) \in [L, U]$ within derived bounds
Event Sum	$\sum P(\text{outcome}_i) = 1$ within an event
Structural	$P(A) = \beta \cdot P(B) + \alpha$ from calibrated model

Statistical arbitrage: cointegration spread (self-calibrating z-score), lead-lag (cross-correlation).

Model-driven: fair value divergence (Wang-model edge), premium decay (rides predictable premium lifecycle).

Pricing Engine

Deploys the Wang Transform from Yang (2026), calibrated on 291,309 contracts across 6 platforms:

$$p^{\text{mkt}} = \Phi\bigl(\Phi^{-1}(p^*) + \lambda\bigr), \quad \hat{\lambda} = 0.183 ; (p < 10^{-15})$$

Hierarchical model: $\lambda_i = 0.259 - 0.072 \ln(1+V) + 0.143 \ln(1+D) - 0.477 |p-0.5|$
Model Greeks: $\partial p / \partial \lambda$, Kelly fraction, edge decay rate
Online calibrator: hybrid batch MLE + streaming EWMA with category shrinkage
Correlation-aware risk: EWMA correlation matrix, effective exposure limits

Yang, Y. (2026). Pricing Prediction Markets: Risk Premiums, Incomplete Markets, and a Decomposition Framework. Working Paper, UIUC. [Replication package]

Quick Start

git clone https://github.com/YichengYang-Ethan/oracle3.git && cd oracle3
poetry install

oracle3 market list --exchange polymarket --limit 10
oracle3 dashboard --exchange solana --initial-capital 10000

See docs for full CLI reference.

Key Technical Choices

Event-driven async engine with snapshot persistence and Unix socket control (pause/resume/killswitch)
SpreadExecutor with automatic LIFO unwind on partial fills — no naked multi-leg positions
Dual-layer risk: local position/drawdown/exposure limits + Solana simulateTransaction pre-flight
On-chain audit trail via Solana Memo program; Jito bundle submission for MEV protection
633 tests, ruff, mypy, codespell CI on every push

Star History

Contributors

If oracle3 helps your research or trading, please ⭐ star the repo — it helps others find it.

License

Apache 2.0 — see LICENSE for details.

This software is for research and educational purposes. Trading involves financial risk.

Frequently Asked Questions

What is oracle3?

oracle3 is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by YichengYang-Ethan. Autonomous prediction market trading agent | Wang Transform pricing engine calibrated on 291K+ contracts | Kalshi · Polymarket · Solana DFlow · Jito bundles · 633 tests. It has 264 GitHub stars.

Is oracle3 safe to use?

Yes. oracle3 passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.

How do I install oracle3?

Clone the repository with "git clone https://github.com/YichengYang-Ethan/oracle3" and add it to your Claude Code skills directory (see the Installation section above).

What programming language is oracle3 written in?

oracle3 is primarily written in Python. It is open-source under YichengYang-Ethan on GitHub, so you can review or fork the full source.

Are there alternatives to oracle3?

Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh oracle3 against similar tools.

Agentic AI for Beginners

Build your first AI agent from scratch - tool use, ReAct pattern, memory, deployment

41 minBeginner

Comments (0)

to leave a comment.

No comments yet. Be the first to share your thoughts!

Related Skills

superpowers

by obra

An agentic skills framework & software development methodology that works.

234,966

20,863

Shell

AI Agentsaibrainstorming

View details

Compare

ECC

by affaan-m

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

230,983

35,243

JavaScript

AI Agentsai-agentsanthropic

The agent that grows with you

216,925

40,767

Python

AI Agentsaiai-agent

View details

Compare

everything-claude-code

by affaan-m

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

185,940

28,768

JavaScript

AI Agentsai-agentsanthropic

View details

Compare

claude-code

by anthropics

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

120,031

19,897

Shell

AI Agents

View details

Compare

cc-switch

by farion1231

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Gemini CLI & Hermes Agent. Only official website: ccswitch.io

118,720

7,959

Rust

AI Agentsai-toolsclaude-code

View details

Compare

Browse all AI Agents skills

handclaw claude-007-agents

Oracle3

Autonomous prediction market trading agent across Kalshi, Polymarket, and Solana.

Why this exists

This system deploys the exact $\lambda$ estimates and covariate model from prediction-market-pricing (Yang, 2026) as its real-time pricing engine.

How oracle3 differs from existing prediction-market tools

	Oracle3	polymarket-whales	prediction-market-maker	py-clob-client
Pricing model	Wang Transform (calibrated MLE)	None	Bid-ask MM	None
Constraint-based arbitrage	8 strategies	None	None	N/A
Multi-venue	Kalshi + Polymarket + Solana	Polymarket only	Polymarket only	Polymarket only
On-chain execution	Solana via DFlow + Jito	No	No	N/A (SDK)
Working paper	Yang (2026), SSRN	No	No	No
Tests	633	0	0	50+
License	Apache 2.0	MIT	MIT	MIT

Architecture

graph TD
    A[Wang Transform Pricing Engine<br/>MLE coefficients from paper] --> B[Fair Value Estimator<br/>Model Greeks · Kelly Sizing]
    B --> C[Strategy Layer]
    C --> D[8 Constraint-Based Arbitrage]
    C --> E[2 Model-Driven Strategies]
    C --> F[LLM Agent Strategies]
    D --> G[Trading Engine<br/>SpreadExecutor · Risk Manager · Position Tracker]
    E --> G
    F --> G
    G --> H[Kalshi]
    G --> I[Polymarket]
    G --> J[Solana / DFlow]

Strategies

Constraint-based arbitrage — each exploits a violated probability axiom:

Strategy	Invariant
Cross-Market	Same event, same price across exchanges
Exclusivity	$P(A) + P(B) \leq 1$ for mutually exclusive events
Implication	$P(A) \leq P(B)$ when A implies B
Conditional	$P(A \mid B) \in [L, U]$ within derived bounds
Event Sum	$\sum P(\text{outcome}_i) = 1$ within an event
Structural	$P(A) = \beta \cdot P(B) + \alpha$ from calibrated model

Statistical arbitrage: cointegration spread (self-calibrating z-score), lead-lag (cross-correlation).

Model-driven: fair value divergence (Wang-model edge), premium decay (rides predictable premium lifecycle).

Pricing Engine

Deploys the Wang Transform from Yang (2026), calibrated on 291,309 contracts across 6 platforms:

$$p^{\text{mkt}} = \Phi\bigl(\Phi^{-1}(p^*) + \lambda\bigr), \quad \hat{\lambda} = 0.183 ; (p < 10^{-15})$$

Hierarchical model: $\lambda_i = 0.259 - 0.072 \ln(1+V) + 0.143 \ln(1+D) - 0.477 |p-0.5|$
Model Greeks: $\partial p / \partial \lambda$, Kelly fraction, edge decay rate
Online calibrator: hybrid batch MLE + streaming EWMA with category shrinkage
Correlation-aware risk: EWMA correlation matrix, effective exposure limits

Yang, Y. (2026). Pricing Prediction Markets: Risk Premiums, Incomplete Markets, and a Decomposition Framework. Working Paper, UIUC. [Replication package]

Quick Start

git clone https://github.com/YichengYang-Ethan/oracle3.git && cd oracle3
poetry install

oracle3 market list --exchange polymarket --limit 10
oracle3 dashboard --exchange solana --initial-capital 10000

See docs for full CLI reference.

Key Technical Choices

Event-driven async engine with snapshot persistence and Unix socket control (pause/resume/killswitch)
SpreadExecutor with automatic LIFO unwind on partial fills — no naked multi-leg positions
Dual-layer risk: local position/drawdown/exposure limits + Solana simulateTransaction pre-flight
On-chain audit trail via Solana Memo program; Jito bundle submission for MEV protection
633 tests, ruff, mypy, codespell CI on every push

Star History

Contributors

If oracle3 helps your research or trading, please ⭐ star the repo — it helps others find it.

License

Apache 2.0 — see LICENSE for details.

This software is for research and educational purposes. Trading involves financial risk.