Expand StrategyMarsi HK vs origin trade analysis

[yafeng2003]

Motivation

• Improve the existing StrategyMarsi origin vs 港股通 comparison by drilling into trade-level differences to determine whether missed large wins are caused by the universe (stock pool) or by the strategy signals.
• Surface which high-return trades from the origin run are absent in the 港股通 run and quantify their contribution to overall performance.

Description

• Add analysis/strategy_marsi_hkt_comparison.md that extends the previous report with trade-level diffing, per-trade listings, and a summary table of portfolio and trade metrics.
• Identify origin-only trades by key (stockno, entry_date) and classify them by whether the ticker appears in the 港股通股票代码.csv list, and produce top-missing-winners tables.
• Compute and display portfolio-level metrics (start/end assets, total return, CAGR, annualized return/volatility, Sharpe, max drawdown) and trade-level metrics (trade count, win rate, avg/median profit, avg win/loss, avg hold days) using ad-hoc Python analysis that reads backtest_results/StrategyMarsi_origin/* and backtest_results_hkt/StrategyMarsi/* CSVs.
• Add a short conclusions section attributing most missed large wins to stock-pool coverage differences rather than signal failure on commonly-covered tickers.

Testing

• Ran the analysis scripts that load StrategyMarsi_daily_returns.csv and StrategyMarsi_trade_analysis.csv from both universes and 港股通股票代码.csv, and the scripts completed successfully and wrote analysis/strategy_marsi_hkt_comparison.md.
• Executed ad-hoc checks that printed trade counts, top missing profits, and summed profit contributions for origin-only trades, and those commands returned expected numeric outputs.
• No unit or integration tests were added because this is an analysis/report-only change.

---

Codex Task