Strategies
One logic, three markets.
The core logic is market-agnostic: detect dividend events, score them, rank them. What changes between markets is data source, ex-date conventions, and calibration thresholds.
| Strategy | Status | Approach | Universe |
|---|---|---|---|
| Dividend ERP Ranking — A-Shares | Live | Event-driven. Declared cash dividends scored at registration date by annualized excess return over the risk-free rate. | Large-cap dividend payers, 3-year track record, DY > 3% |
| Dividend Yield Ranking — Hong Kong | In development | Ex-date based. Higher yield thresholds reflecting structurally higher HK payout yields. | HKEXnews declared distributions |
| Dividend Growth — United States | In development | Growth-oriented rather than high-yield, given lower US payout yields. | SEC EDGAR / Finnhub |
Layer separation
Each market implementation separates into three layers, so that a new market is an adaptation rather than a rewrite:
- Logic layer — event capture, scoring formula, ranking, rotation. Shared across all markets.
- Data layer — the specific source and schema for declared distributions. Market specific.
- Market layer — ex-date conventions, yield thresholds, risk-free reference. Recalibrated per market.
Core formulas
Dividend yield is computed per event, using each event's own registration-date closing price:
DY = Σ (DPS_i / P_registration_i)
Excess return potential is the annualized yield over the risk-adjusted reference rate:
ERP = annualized DY − (r10Y + λ · σ)
Events are ranked against their trailing 252-day history. Qualifying events enter available slots; a higher-ranked new event replaces the weakest current holding.
Known limitations
- Look-ahead bias — signals use already-implemented payouts, which overstates realized returns.
- Survivorship bias — delisted issuers' dividend histories are not fully verified.
- Ex-dividend gap — price drops at ex-date are assumed to recover; not guaranteed in practice.
- Odd-lot drag — capital below 1M CNY suffers rounding losses from indivisible share lots.