| --- |
| title: ArbIntel |
| emoji: 📈 |
| colorFrom: blue |
| colorTo: green |
| sdk: streamlit |
| sdk_version: 1.32.0 |
| app_file: app.py |
| pinned: false |
| --- |
| |
| # ArbIntel: Prediction Markets Alpha Engine |
| ## System Architecture & Documentation |
|
|
| ### 1. Overview |
| ArbIntel is a real-time trading intelligence system designed for Polymarket and Kalshi. It features cross-platform arbitrage scanning, Bayesian fair-value probabilistic modeling, HMM regime detection, and an NLP sentiment pipeline for event-driven trading. |
|
|
| ### 2. Core Modules |
|
|
| #### Data Ingestion (`src/clients/` & `src/ingestion.py`) |
| - Async WebSocket and REST clients for Polymarket and Kalshi. |
| - Real-time normalization of order books to implied probabilities (0-1). |
| - TimescaleDB storage schema for tick-level data. |
|
|
| #### Arbitrage Engine (`src/strategies/arbitrage.py`) |
| - **Cross-Platform Arbitrage**: Identifies risk-free margins between Polymarket and Kalshi considering fee schedules. |
| - **Intra-Market Parity**: Detects structural pricing violations (e.g., YES + NO < 1.0). |
|
|
| #### Bayesian Fair Value (`src/models/bayesian.py`) |
| - Evaluates real-time market action against a Beta-distribution belief system. |
| - Adjusts expected probability (Fair Value) based on trade volume and decays older beliefs. |
|
|
| #### NLP Sentiment Pipeline (`src/models/nlp.py`) |
| - Hugging Face `ProsusAI/finbert` integration. |
| - Scores incoming news and social feeds for bullish/bearish confidence to feed momentum strategies. |
|
|
| #### HMM Regime Detection (`src/models/hmm.py`) |
| - 2-state Gaussian Hidden Markov Model. |
| - Utilizes log-returns, volume spikes, and bid-ask spreads to classify "Stable" vs "Volatile" market conditions for dynamic position sizing. |
|
|
| #### Backtesting & Execution (`src/backtest/engine.py` & `src/execution.py`) |
| - Event-driven Pandas backtester calculating Sharpe, Max DD, and Win Rates accounting for basis-point slippage. |
| - Paper Trading engine enforcing risk limits (maximum capital deployed, daily loss limits). |
|
|
| #### Blockchain Integration (`src/blockchain.py`) |
| - Polygon RPC listener monitoring USDC (`0x2791Bca...`) transfers into the Polymarket Exchange Contract. |
| - Generates "Whale Alerts" for transaction sizes exceeding expected retail limits. |
|
|
| #### Alerts (`src/alerts.py`) |
| - Discord & Slack webhook integration for instant arbitrage notification. |
|
|
| ### 3. Running the Dashboard |
| The Streamlit dashboard allows manual overriding and live viewing of the `CrossPlatformArbitrage` scanner. Deployable directly to Hugging Face Spaces. |
| ```bash |
| streamlit run app.py |
| ``` |
|
|
