This strategy is a fully automated cryptocurrency quantitative trading system powered by a large language model, orchestrated through a workflow engine that coordinates multiple functional modules — spanning market data collection, AI decision generation, and order execution — into a complete end-to-end trading loop.

Strategy Architecture

The system consists of two parallel workflows:

Main Strategy Workflow (Hourly Execution): Periodically triggers a full analysis and decision cycle, sequentially handling account state initialization, multi-dimensional market data collection, position retrieval, historical performance aggregation, and final data consolidation fed into the AI model, which generates structured trading signals that are then executed.

Take-Profit/Stop-Loss Monitoring Workflow (Second-Level Execution): Runs at high frequency independently, continuously monitoring the P&L status of all open positions. Supports both fixed-ratio TP/SL and AI-defined dynamic price targets. Upon trigger, it automatically closes positions and logs results to the historical database.

Core Functional Modules

Market Data Collection: Simultaneously fetches candlestick data across 3-minute and 4-hour timeframes, computing EMA, MACD, RSI, ATR, and other technical indicators alongside funding rate data, forming a multi-timeframe signal fusion view.

Historical Performance Analysis: Maintains an independent trade record database for each asset, tracking win rate, profit/loss ratio, long/short directional performance, average holding duration, and other key metrics to provide the AI with quantitative performance context.

Freeze Protection Mechanism: Dynamically monitors consecutive losses per coin over the past 4 hours. Upon exceeding the threshold, trading rights for that coin are automatically suspended for 4 hours, effectively preventing emotional loss spirals.

AI Decision Engine: A structured prompt consolidates market data, position details, account state, and historical performance into the AI model. The model outputs a complete JSON decision object per coin, including direction, leverage, position size, TP/SL prices, confidence score, and historical directional bias.

Smart Position Sizing: Dynamically calculates order quantities based on risk amount, stop distance, and contract multiplier, while respecting account balance constraints and maximum per-trade risk limits.

Fault-Tolerant Parsing System: A four-layer AI output parsing mechanism (standard JSON → object-by-object extraction → regex matching → keyword scanning) ensures reliable signal extraction even when AI output formatting is imperfect.

AI Decision Logic

At each execution cycle, the AI first checks the freeze status of every coin, mandating a hold signal with zero risk allocation for frozen assets and skipping all further analysis. For tradeable coins, the AI dynamically adjusts base risk exposure according to historical performance: high-performing assets receive increased allocation while consistently losing assets are scaled back. Direction is determined by combining technical signals with historical long/short bias, and the model is explicitly required to evaluate both long and short opportunities at each cycle. During the initial phase when no historical data exists, the system relies entirely on technical analysis to initiate trades and begin building a performance record.

Suitable For

Quantitative traders seeking to combine AI judgment with systematic risk controls for fully autonomous 24/7 operation. The strategy supports both live trading and paper trading modes.

Vedio Link：AlphaArena Clone Trading System – Optimized Version 2.0