liu1527-bot/a-share-quant-agent

A multi-factor quant stock-selection project for China A‑shares (CSI 300). Intended for quant researchers and systematic traders to build, backtest and automate a monthly rebalancing long-only portfolio (Top 30 holdings). What it does - Combines value (1/PB, 1/PE), short-term reversal, low-volatility and mid-term momentum into a weighted multi-factor score. - Monthly rebalancing with a one-way trading cost assumption of 0.15%. - Provides vectorized backtesting, single-factor evaluation (IC/IR/group tests), and out-of-sample (OOS) validation. Report generation (charts + Markdown) and GitHub Actions automation produce monthly outputs. Performance (reported) - In-sample (2021–2023) annual return 6.88%; OOS (2024–2025) 7.52%; reported annualized excess vs CSI300 +10.17%; improved max drawdown to -17.24%. - Example factor weights: value_pb 0.30 (IC=0.094, IR=0.314), value_pe 0.25, reversal_5 0.20, low_vol_60 0.15, momentum_120_5 0.10. Repo highlights - Modular code: data loaders, fundamentals, factor construction, factor panel, strategy, backtest, factor_eval, OOS validation, report. - Scripts for data warmup and a monthly_run entry used by GitHub Actions. Notes - For research use only; not investment advice. Real-world frictions (liquidity, market impact, policy) may alter results.