DRADIS: A High-Frequency Predictive Market Automated Trading Platform Built with Rust

DRADIS (Direct Reaction And Dynamic Intelligence System) is a low-latency multi-strategy automated trading platform developed with Rust, designed specifically for predictive markets like Polymarket. It aims to address the issues of delayed price discovery and low efficiency in predictive markets. The platform integrates architectural design, six core trading strategies, and a machine learning module to capture market inefficiencies and diversify risks with millisecond-level responses.

Project Background

Predictive markets (such as Polymarket) have risen in recent years, but there are issues of delayed price discovery and low efficiency. DRADIS is positioned as a complete automated trading platform, not just a simple bot.

Advantages of Rust Language

Rust has no Global Interpreter Lock (GIL) or Garbage Collection (GC) pauses, supporting concurrent evaluation of multiple strategies at a 50-millisecond frequency, and maintains stable performance in high-frequency scenarios.

The core of DRADIS's architecture is the Orchestrator (CIC) coordinator, which is responsible for maintaining data connections with Polymarket's CLOB (Central Limit Order Book) and Binance Oracle:

• Data Flow: Binance Oracle provides price/funding rate data, Polymarket pushes order book data via WebSocket, which is aggregated and processed by the Orchestrator.
• Architectural Advantages:

1. Parallel Scheduling: All strategies are evaluated in parallel within each 50-millisecond heartbeat cycle, avoiding cumulative serial delays;
2. Isolated Positions: Each strategy has an independent capital budget and position book, so risks do not affect each other;
3. Signal Filtering: OBI (Order Book Imbalance) veto mechanism—when the indicator reaches -0.60, it blocks entry into toxic flows.

DRADIS deploys six strategies to address different market inefficiencies:

1. Momentum (Interceptor): Captures high volatility on Binance; when the target moves $75 within 5 seconds, it seizes the pricing first-mover advantage on Polymarket;
2. Maker (Sentinel): Maintains bilateral orders on the Window venue to profit from spreads; skips quoting when the order book is severely skewed;
3. Arbitrage (Surveyor): Monitors the sum of YES/NO prices, looking for arbitrage opportunities where the sum (after deducting fees) is below $1.00;
4. Time Decay (Ghost): Leverages the price convergence feature; issues GTC orders on the Hourly venue near expiration to profit with 0% fees;
5. Basis/Funding (Analyst): Compares Polymarket sentiment with Binance's funding rate, following the direction of smart money;
6. GBoost (Cylon): Online gradient boosting classifier trained on 13 features; retrained every N ticks, and trades based on predicted probabilities.

Risk Control

• Orphan Position Detection: Automatically unwinds one-sided hedging positions after 60 seconds;
• Fee Threshold: Hard-coded to prevent Taker strategies from entering high-fee environments;
• Circuit Breaker Mechanism: System locks after three consecutive execution failures;
• TOCTOU Safety Entry: Atomic lock to prevent duplicate orders.

Deployment Configuration

• Preset Templates: Conservative (<$100, only Maker/Time Decay), Balanced ($100-$300, all strategies), Aggressive ($200+, maximum positions);
• GHOST_MODE: Simulated trading test to validate strategies with zero risk;
• Trading Records: Automatically generates daily CSV files containing fields like timestamp, strategy, profit, etc.

Technical Highlights

1. Model Version Management: GBoost models are named with versioning to avoid dimension mismatch issues;
2. No Backtesting Framework: Believes traditional backtesters have data/strategy fidelity issues in predictive markets, so GHOST_MODE is more reliable;
3. Performance Optimization: Provides guidelines for kernel tuning, CPU binding, cloud instance selection, etc.

Summary Thoughts

DRADIS is a mature engineering practice for automated trading in predictive markets, covering end-to-end trading infrastructure. It demonstrates the potential of Rust in financial applications for developers, provides an extensible framework for traders, and has reference value for researchers. Note: This project is experimental software; there is no guarantee of profit from the strategies, and there is a risk of loss.