DeepQuant Strategy Backtesting Assistant

Problem

Quant strategy workflows often split strategy design, DeepQuant code generation, data preparation, and result interpretation across separate steps. The DeepQuant Strategy Backtesting Assistant focuses on the DeepQuant quest framework and connects describing a strategy, generating code, running a backtest, and producing reports into one executable workflow.

How It Works

The skill expects initialization at the start of each session: login-state detection, SDK checks, shared module merging, and data-pack validation. It then generates strategies/xxx_strategy.py from the user's description, presents it for confirmation, and runs scripts/run_backtest.py. Key outputs include:
- Return metrics: total return, annualized return, benchmark return, Alpha, Beta
- Risk metrics: max drawdown, annualized volatility, Sharpe ratio
- Files: orders.csv, trades.csv, summary files, and backtest_analysis.png

It also follows DeepQuant conventions: init and handle_bar are required for standard bar backtests, orders are not allowed in before_trading or after_trading, and gid calls should be treated as (data, code, msg) tuples.

Boundaries

It supports A-share stocks, index-constituent strategies, and ETF/fund backtests, with benchmark codes available from indexes.h5. Futures, options, and convertible bonds are not supported and should be rejected clearly. First use requires base data download; minute-bar backtests need --minbar; generated strategy code should be reviewed before execution.