Event-Driven Backtest Engine

Problem It Solves

Event-driven backtesting is not just about comparing strategy equity curves; it is about checking whether a strategy can execute through a realistic trade pipeline. Daily return-only tests can miss order creation, fill confirmation, cash usage, trading halts, T+1 settlement, and transaction costs, which may overstate tradability. This skill targets A-share daily historical backtesting and separates market data, signals, orders, and fills into an inspectable event flow, useful for studying event-driven architecture, cost modeling, and strategy review.

How It Works

• Event chain: MarketEvent → SignalEvent → OrderEvent → FillEvent, with an event queue advancing backtest state.
• Data and strategies: uses akshare for forward-adjusted daily bars and includes ma_cross, rsi, and bollinger.
• Trading rules: models A-share T+1, 100-share lots, commission, stamp duty, transfer fee, and slippage.
• Analytics and attribution: reports Sharpe, Sortino, Calmar, max drawdown, win rate, payoff ratio, and Brinson allocation, selection, and interaction effects.
• Output: generates an HTML report with an equity curve, drawdown curve, monthly return heatmap, and trade log.

Boundaries

It is suitable for daily-bar historical validation, cost-sensitive A-share backtests, and learning event-driven backtest design. It does not support live trading, minute or tick-level backtests, short selling, price-limit constraints, or order-book depth simulation; fills are based on daily close prices, and results are for research only.