Longbridge Buffett Moat Stock Screener

# Screening criteria — Layer-1 hard filters, Layer-2 moat scoring, verdict matrix, holding-period mapping

Loaded on demand by `longbridge-buffett-moat-stock-screener`. Numeric defaults align with Buffett's publicly stated reference points (Letters to Shareholders, *Berkshire Hathaway 1977–2024*, and *The Essays of Warren Buffett*). Do not change without updating SKILL.md examples in lockstep.

---

## 1. Layer-1 — Hard quantitative filter (pass / fail gate)

Apply as a binary pre-filter. A symbol that fails ≥ 2 filters drops out by default (user can relax to "fail ≤ 3" on request). The thresholds reflect Buffett's own framing: a great business should clear all of them under normal conditions — failing two simultaneously is a strong signal the business itself isn't good enough for the framework.

| # | Filter | Default threshold | Why this matters (Buffett anchor) | Source field |
|---|---|---|---|---|
| 1 | ROE (5y average) | ≥ 15% | *"The primary test of managerial economic performance is achievement of a high earnings rate on equity capital employed."* Sustained high ROE = the business itself is good. | derived from `financial-report --kind IS` (NI) ÷ avg equity from `financial-report --kind BS` |
| 2 | Debt-to-asset ratio | ≤ 50% | High leverage adds risk that's invisible until it isn't. Buffett: *"Leverage is the only way a smart guy can go broke."* | BS — total liabilities ÷ total assets |
| 3 | Free cash flow (OCF − Capex) | positive each of last 3 years | *"Accounting earnings can lie; cash cannot."* FCF is the "real money" test. | `financial-report --kind CF` |
| 4 | Years listed | ≥ 5 years | Need a track record to judge management & business quality; pro-rate the 5–10y window if 5–10 years available. | `basicinfo.listing_date` |
| 5 | Gross margin (TTM) | ≥ 30% | Low gross margin generally signals no pricing power, often commoditised. Buffett's favoured businesses (See's, Coke, Apple) all clear this comfortably. | IS — gross profit ÷ revenue |

### Sector-adjusted filter substitutions

Banks / insurance / brokerage business models have intentionally different balance-sheet shapes; the raw rules above produce false rejects. Substitute as follows and **note the substitution in the row's Data Note**:

| Sector | Substitution |
|---|---|
| Commercial banks | ROE rule kept; **debt rule** → CAR ≥ regulator-required + buffer · **FCF rule** → ROA ≥ 1.0% with stable NIM · **gross margin** → cost-to-income ratio ≤ 50% |
| Insurance | ROE kept; **debt** → solvency margin ≥ 200% · **FCF** → underwriting combined ratio < 100% over 5y · **gross margin** → omit |
| Brokerage / asset management | ROE kept; **debt** → capital-adequacy ratio per regulator · **FCF** → omit (replace with operating-cash-flow proxy from net commission income); **gross margin** → omit |

If the symbol's sector classifier is missing or ambiguous, default to the raw rules — better to flag a borderline name with a "sector data missing" note than silently mis-apply the wrong substitution.

### Pro-ration rules

- **Listed 5–10 years** (i.e. less than a full 10-year window): pro-rate filters 1 and 3 over the available years. Mark row "history-limited".
- **Sector mid-cycle dip** (e.g. one bad year in an otherwise clean record): the user may *manually* relax filter 1 ROE to "5y avg with worst year removed" — must be echoed in the Market Summary.

---

## 2. Excluded cohorts (no scoring, dropped from the universe)

These businesses are structurally outside Buffett's framework. The screener does **not** run substitute models for them — they are removed from the universe and counted in the Market Summary "Excluded" line. If the user explicitly asks for one of these, return an honest empty result with a redirect to the nearest non-excluded sector.

| Cohort | Reason | Display in Market Summary | Suggested redirect |
|---|---|---|---|
| Airlines | Buffett 1997 Annual Meeting: *"airline industry is a value-destroying industry"*; high capex, cyclical demand, no pricin