wbh604/fund-guy-skill

What it solves

It addresses the problem where investors rely solely on net asset value (NAV) curves and platform ratings to judge fund managers. These metrics often ignore whether a manager's gains were due to skill or simply a favorable market style (Beta), and they fail to verify if the timing of specific buys and sells was actually correct.

How it works

The engine takes a fund code as input and uses an AI Agent to audit the manager's trading behavior over several years. It scrapes public data (NAV, holdings, announcements, and K-lines) from sources like EastMoney and Baostock to perform a "post-mortem" on every trade, comparing the action to the actual price movement 12 months later. It then generates a self-contained interactive HTML report that visualizes trade timing, independence from company peers, and "god-making" detection (checking for unethical patterns like restricting subscriptions at market peaks).

Who it’s for

• Retail investors who want to verify the actual skill of a fund manager beyond surface-level ratings.
• Analysts looking for a data-driven, behavioral audit of fund management.
• Users of AI agents (like Claude Code, Cursor, or OpenClaw) who want to integrate financial auditing skills into their workflow.

Highlights

• Behavioral Scoring: Rates managers on timing, control, and excess quality rather than just final returns.
• Trade Autopsy: Verifies the win rate of buy points and the "dodge rate" of liquidations against 12-month future trends.
• Interactive Visuals: Includes K-line replays, cost lines, and "disaster movies" that simulate potential losses.
• Independence Audit: Measures how much a manager's unique decisions differed from their colleagues and whether those differences led to outperformance.
• God-Making Detection: Scans for "black history" such as window-dressing at year-end or restricting fund entry at market peaks.