Ramanujan Stock Screening

Problem being solved

A-share fundamental screening often falls into two bad modes: reading every page of a financial report without compressing it into the few load-bearing relationships, or reacting to announcement headlines and vague wording such as expected to and framework agreement as if they were evidence. This skill frames the task as a more engineering-style filter: use a small set of checkable financial fields and disclosure wording to reject companies with hard structural problems, such as cash flow divergence, goodwill overhang, pledge risk, or earnings quality that depends on non-recurring items.

How the skill works

• Structure first: inspect the balance sheet, income statement, and cash flow statement as one system, with attention to ratios and direction rather than isolated numbers, such as operating cash flow / net profit excluding non-recurring gains, goodwill / net assets, and major shareholder pledge ratio.
• Anomaly detection: treat mismatched signals as high-priority red flags, including high cash plus high interest-bearing debt, repeated auditor changes, and non-recurring gains that dominate reported profit.
• Disclosure validation: evaluate announcements as event inputs, checking whether they contain concrete numbers, whether positive wording is diluted by expected-to language or active-promote language, and whether key information such as contract amount, funding source, or dividend coverage is missing.
• Closed-loop verification: run a coarse screen, then a deeper stare-down pass, and finally force a short-seller view to identify the strongest attack point before accepting the thesis.

Boundaries and caveats

This is better used as a fundamental risk filter and watchlist builder, not a short-term timing tool. It depends on real data and industry benchmarks; without comparisons for gross margin, cash-to-profit, and valuation percentile, the judgment becomes much weaker. For high-leverage sectors such as financials and real estate, simple thresholds need adaptation, and historical patterns cannot be blindly extrapolated after structural industry shifts.