02 · Data & Research Platforms

"Watching charts relies on software; research relies on data." Most people get stuck at step one: where does data come from, does it cost money, and how do you make sure it's right once you have it?

This article organizes data platforms into four layers: quant data sources (code/terminals), channels for historical quotes, macro & industry data, news & research reports — plus a primer on "alternative data," ending with a zero-cost research toolkit. Solve the data problem first, then talk strategy (see Chapter 15 - Quant Practice↗).

⚠️ Risk Warning

Data interfaces' fields, free quotas, and terms of use change at any time (defer to each platform's latest announcements and official docs); free data can contain gaps, delays, and abandoned endpoints — cross-check before using it for live decisions. Evaluate the return on any paid data service yourself. Markets carry risk; invest with caution. Nothing here constitutes investment advice.

1. Quant Data Sources: Five Steps from Free to Paid

Selection logic:

• Just want to prove a concept, don't need stability → start free with AKShare;
• Serious A-share research that must be reproducible → Tushare Pro (upgrade points as needed);
• Don't want to maintain data yourself, backtest on a platform environment → JoinQuant online;
• Prioritize data quality and API standards, willing to pay → RiceQuant / terminals;
• Institution-level or deep research → Wind (see the professional terminal section in 01 - Market Data Software↗).

2. Channels for Historical Market Data

2.1 Exchange Official Websites

The most authoritative first-hand channel — and free:

2.2 The Tushare Pro Points System

• Registration is free; basic interfaces (daily bars, instrument lists, trading calendars) require few points;
• High-value data (financial indicators, minute-level data, dragon-tiger lists) requires more points — earned via contribution, sponsorship, or payment; see official docs for exact rules;
• Usage notes: daily call counts are capped, so fetching whole-market history means batched loops + rate limiting — never write a one-shot pull-everything script.

2.3 Crypto Exchange Public APIs

Crypto offers the easiest-to-obtain first-hand historical data of any market:

• Binance/OKX/Bybit all expose public candlestick history endpoints (e.g., /api/v3/klines), accessible without an API key;
• You can page backward from fairly early points (e.g., Binance BTCUSDT daily bars go back to 2017; check official docs for exact ranges);
• Well-specified fields, no price-adjustment issues (spot has no corporate actions) — the best dataset for personal quant practice;
• Limits: rate limited by weight, so control request frequency when pulling full history; endpoint paths and parameters change between versions — defer to official docs.

# Teaching example: fetch Binance daily candles (real fields per official docs)
import requests
resp = requests.get(
    "https://api.binance.com/api/v3/klines",
    params={"symbol": "BTCUSDT", "interval": "1d", "limit": 100},
    timeout=10,
)
data = resp.json()   # [open time, open, high, low, close, volume, ...]

2.4 Data Quality Checks: Reconcile Numbers Before Research

The first thing after getting data isn't modeling — it's reconciling numbers against market software (method detailed in Chapter 15 - Quant Practice - 02↗):

3. Macro & Industry Data

3.1 Official First-Hand Data (Free)

3.2 Commercial Macro Data (Paid)

• CEIC: global macro/industry data platform, broad coverage, institutional subscriptions;
• Wind macro module: most used domestically, macro data and quotes in one terminal, easy export;
• Eastmoney Choice / THS iFinD: also cover macro data at lower prices than Wind.

Free personal alternative: the NBS website carries every key series; combined with the "China Data" portal and the PBOC's statistical reports it covers ~90% of macro research needs — but you'll have to assemble time series yourself.

3.3 Examples of Industry Data Sources

Industry association data is usually free or low-barrier, but its definitions may differ from official statistics — cite source and methodology when referencing.

4. News & Research

4.1 Flash News & Media

4.2 Channels for Broker Research Reports

5. Alternative Data: Institutions' "Information Edge"

Alternative data means data outside traditional financial datasets that reflects real economic activity. Institutions pay heavily for it; individuals just need to grasp the concept:

• For individuals: alternative data is extremely costly and noisy, and most people lack the processing capability — knowing it exists suffices. Your informational edge more likely comes from careful reading of public data than from chasing novel datasets;
• Collecting and using alternative data has privacy and compliance boundaries (cross-border transfer of personal data, collection consent, etc.), and institutions run dedicated compliance reviews before use.

6. Personal Free Stack: A Zero-Cost Research Toolkit

You can build a usable research foundation without spending a cent (with upgrade paths included):

Principle: the entire cost of the free stack is "your own hours maintaining the data." When data wrangling time exceeds its expected value, consider paying — what stalls most individual researchers isn't lack of money for data, but never putting the free data to work.

Data Copyright & Boundaries of Use

• Quote data copyright usually belongs to exchanges/vendors: free-interface data may be used for personal study and research; commercial use, redistribution, or external services generally require licenses — read each source's terms before use;
• Scraping public websites: respect site terms and robots rules, throttle frequency, and avoid stressing target sites;
• Crypto exchange API data: commercial/redistribution rules differ per platform — follow official docs;
• One pragmatic habit: record "data source + retrieval date" in script comments — both copyright awareness and part of reproducible research.

Wiring the Data Pipeline: From Fetching to Usable

Data research isn't "download once and done" — it's a pipeline you maintain daily:

Fetch scripts (scheduled) → raw data landed (CSV/SQLite) → cleaning & reconciliation → standard-format store → research & backtesting

• Schedule fetching with cron/system schedulers — see 05 - Runtime & Automation Environment↗;
• Run a reconciliation check after each fetch (latest date, row counts, outliers) and alert immediately on anomalies;
• Separate data from code: data under data/, code under scripts/, config under config/ — directory conventions in 15 - Quant Practice - 01↗.

7. Next Steps

With data in hand, the next step is getting "data → backtest → strategy" running end-to-end — see 15 - Quant Practice↗. For where to trade these instruments, see 03 - Broker & Futures Broker Selection↗.