DeepQuant Strategy Builder

From Idea to Backtestable Strategy

When engineers turn an idea such as "buy on MACD crossover, take profit on RSI overbought, and trade CSI 300 constituents" into a strategy, they usually face three tasks: clarifying the target, signal, backtest window, and frequency; generating code that fits the deepquant quest framework with init(context) and handle_bar(context, bar_dict); and running a sample-in validation with deepquant-backtest to interpret the metrics. The DeepQuant Strategy Builder connects this workflow: it first confirms missing inputs and offers sensible defaults instead of assuming; then it generates a .py strategy file with Chinese comments, parameter notes, and risk warnings, and only runs the backtest after user confirmation; finally, it returns indicators such as returns and drawdown, plus commentary on strengths, limitations, improvement directions, and out-of-sample caveats.

Boundaries and Notes

The skill assumes Python 3.8 through 3.12 and the deepquant SDK, and it requires the initialization and authentication flow to complete. Instrument coverage follows the local bundle data: A-shares, ETFs/funds, and index-constituent stock selection are supported, while direct index trading and futures, options, or convertible bonds are not. Before calling gid APIs, version and cache requirements must be satisfied; when data is missing, authentication fails, or an instrument is unavailable, the workflow should stop, explain the issue, and ask for user confirmation rather than substituting around it. Deliverables should be sent as attachments, including strategy code, order logs, trade records, and backtest summaries, instead of pointing only to file paths.