You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This dataset contains licensed Level-2 market data for the Shanghai and Shenzhen stock exchanges. Access is granted for non-commercial academic research only. By requesting access you confirm that you will not redistribute the data or any derivative from which it can be reconstructed, and that you will cite this dataset in any resulting publication.
Log in or Sign Up to review the conditions and access this dataset content.
Tick-by-Tick Orders and Trades — China A-Shares (2023)
Full-depth Level-2 message data for every listed stock, fund and ETF on the
Shanghai (.SH) and Shenzhen (.SZ) exchanges, 2023.
This repository holds one calendar year. The remaining years live in sibling repositories, one per year:
| Year | Repository |
|---|---|
| 2023 | alphat01/Tick-by-Tick-Orders-China |
| 2024 | alphat02/Tick-by-Tick-Orders-China |
| 2025 | alphat03/Tick-by-Tick-Orders-China |
| 2026 | alphat04/Tick-by-Tick-Orders-China |
At a glance
- Trading days: 242 (20230103 – 20231229)
- Archives: 242
- Total size: 606.9 GiB (compressed)
- Granularity: every order and every trade, millisecond timestamps
- Layout: one 7-Zip archive per trading day
Layout
202301/20230103.7z
└── <WIND_CODE>/ e.g. 000001.SZ, 600519.SH — one directory per instrument
├── 行情.csv snapshot quotes: trade prints + 10 levels of depth
├── 逐笔委托.csv order-by-order: every submission and cancellation
└── 逐笔成交.csv trade-by-trade: every execution, with both side order ids
| Month | Archives | Size |
|---|---|---|
202301/ |
16 | 35.2 GiB |
202302/ |
20 | 49.7 GiB |
202303/ |
23 | 55.9 GiB |
202304/ |
19 | 51.8 GiB |
202305/ |
20 | 50.1 GiB |
202306/ |
20 | 52.0 GiB |
202307/ |
21 | 42.8 GiB |
202308/ |
23 | 52.3 GiB |
202309/ |
20 | 48.6 GiB |
202310/ |
17 | 46.3 GiB |
202311/ |
22 | 63.0 GiB |
202312/ |
21 | 59.1 GiB |
Columns
All three files are CSV with a single header row in Chinese. 自然日 is
YYYYMMDD, 时间 is HHMMSSmmm (millisecond resolution). Prices are integers
scaled by 10000 — 114100 means 11.41 CNY.
逐笔委托.csv — order-by-order feed
| Column | Meaning |
|---|---|
万得代码 |
Wind instrument code, e.g. 000001.SZ |
交易所代码 |
Exchange-native symbol |
自然日 / 时间 |
Date / time of the message |
委托编号 |
Exchange order id |
交易所委托号 |
Exchange sequence number |
委托类型 |
Order type |
委托代码 |
B buy / S sell / D cancel |
委托价格 |
Limit price (×10000) |
委托数量 |
Order quantity, shares |
逐笔成交.csv — trade-by-trade feed
| Column | Meaning |
|---|---|
成交编号 |
Exchange trade id |
成交代码 |
C cancel / F fill |
BS标志 |
Aggressor side |
成交价格 / 成交数量 |
Execution price (×10000) / quantity |
叫卖序号 / 叫买序号 |
Resting sell / buy order ids — these let you rebuild the book |
行情.csv — snapshot feed: trade print, cumulative volume and turnover, OHLC,
ten levels of bid and ask price and size (申买价1..10, 申买量1..10,
申卖价1..10, 申卖量1..10), and index-level aggregates.
Reader code and the full field dictionary ship in docs/ —
l2_read.py parses the CSVs, 各代码的意义.docx documents every code value.
Usage
from huggingface_hub import snapshot_download
# one month
snapshot_download("alphat01/Tick-by-Tick-Orders-China", repo_type="dataset",
allow_patterns="202301/*", local_dir="./2023")
7z x 20230103.7z -o./20230103
Access
Gated. Requests are reviewed manually. Academic, non-commercial use only; no redistribution. See the terms on the access form above.
- Downloads last month
- 172