Source: task5-ashare-leader/output/task5_data_layer_status.md
Last modified: 2026-05-17 13:03:29 · 1,659 bytes

Task #5 Data Layer Status

Status: complete for RMB A-share leader v0 raw data.

Raw Data Pulled

- Universe: 5,447 A-share stocks active during 2015-01-01 to 2026-05-16.

- Pre-adjusted OHLCV + limit prices: 28/28 parquet chunks.

- Daily fundamentals/factors: 28/28 parquet chunks.

- Northbound holdings: 28/28 parquet chunks.

- Suspended flags: 11/11 parquet chunks.

- ST flags: 11/11 parquet chunks.

- Benchmarks: CSI 300 + CSI 800 daily bars.

- Shenwan industry snapshots: 2015-2026 annual snapshots.

Data root: ~/data/mini-cypress/rqdata/equity/.

Fundamental Field Mapping

- EPS MRQ: basic_earnings_per_share_mrq_0.

- EPS TTM: basic_earnings_per_share.

- Revenue TTM: operating_revenue_ttm_0.

- Gross profit TTM: gross_profit_ttm_0.

- OCF TTM: cash_from_operating_activities_ttm_0.

- Net profit TTM: net_profit_ttm_0.

- Equity: total_equity.

- OCF/NI: cash_from_operating_activities_ttm_0 / net_profit_ttm_0.

- ROE proxy: net_profit_ttm_0 / total_equity.

- Gross margin: gross_profit_ttm_0 / operating_revenue_ttm_0.

Industry Mapping

RQData shenwan_instrument_industry(..., date=..., level=1) supports dated queries and returned old/new classifications for historical vs current dates. v0 raw layer stores annual snapshots. Backtest should use previous available snapshot, or fall back to 2018-12 fixed snapshot if stricter PIT validation fails.

Engineering Note

The strict all-market wide-matrix baseline OOM-killed at ~5.7GB RSS on this host. Chunked/candidate-first implementation is required before final strict baseline numbers. Offshore work is deferred per latest queue; Tiingo T1 remains complete locally.