# JinJing Shared Data - Unified OHLCV Datasets

**Version**: 2.0.0 | **Updated**: 2026-04-13

## Dataset Overview

| File | Market | Rows | Period | Notes |
|------|--------|------|--------|-------|
| `cn_unified.parquet` | CN (A-shares) | 8,502,401 | 2018-2026 | Full coverage, 4987 stocks |
| `cn_2010_2017_lite.parquet` | CN (A-shares) | 19,440 | 2010-2017 | **LIMITED**: only 10 stocks |
| `us_unified.parquet` | US | 905,803 | 2010-2026 | US stocks |
| `hk_unified.parquet` | HK | 499,076 | 2010-2026 | OHLC/zero-vol filtered |
| `delisted_unified.parquet` | Delisted | 1,207,970 | Various | Delisted stocks |

## Schema (All Files)

```
date, symbol, open, high, low, close, adj_close, volume, market
```

## ⚠️ Data Quality Notes

### adj_close Risk (US/HK)
- **US/HK `adj_close`**: For recent data, `adj_close` ≈ `close` (dividend/adjustment factors may be stale or unavailable)
- If you need accurate adjusted close, recalculate using dividend data from a financial API
- **CN `adj_close`**: Set equal to `close` (no adjustment applied)

### CN 2010-2017 Lite Data
- Only 10 Shanghai stocks (sh.600000-sh.600015 range)
- Not representative of full market - use for historical research only
- Full A-share coverage starts 2018

### HK Data
- OHLC consistency filter applied: removed rows where `high < low` or `high < max(open,close)` or `low > min(open,close)`
- Zero-volume rows removed (were 9.6% of raw data)

## Fixes Applied (v2.0.0)

| Severity | Issue | Rows Affected | Fix |
|----------|-------|---------------|-----|
| **P0** | CN 2010-2017 only 10 stocks (misleading) | 19,440 | Separated to `cn_2010_2017_lite.parquet` |
| **P1** | HK OHLC inconsistency | 5,073 | OHLC filter applied |
| **P1** | HK zero-volume rows | 53,552 | Zero-vol filter applied |
| **P1** | Schema column order not unified | All 4 datasets | Normalized to `date,symbol,open,high,low,close,adj_close,volume,market` |
| **P1** | DELISTED spurious `amount` column | 552,070 | Column dropped |
| **P2** | CN 3 rows `open=0` | 3 | Rows removed |

## Usage

```python
import pandas as pd

# Load any unified dataset
df = pd.read_parquet("cn_unified.parquet")
df['date'] = pd.to_datetime(df['date'])
```