Education
Teach backtesting on data that doesn't lie
FinzData is free for quant finance and financial-engineering courses. Instructors and enrolled students get full-history access for the length of the course, on the same point-in-time, survivorship-free data used in production research.
The offer
Free for courses at degree-granting institutions
For the length of the course, the instructor and every enrolled student get full-history access to the datasets that matter for teaching rigorous backtesting:
| US equities | Survivorship-free US stocks with 8,500+ delisted names and 231 ETFs |
|---|---|
| Prices | 1-minute to daily bars since 2000 |
| Fundamentals | Point-in-time SEC fundamentals, as filed |
| 13F holdings | Institutional holdings since 2013 |
| COT | CFTC Commitments of Traders since 1986 |
| Macro | 20,000+ FRED/ALFRED macro series with vintages |
| Access | Python client (pip install finzdata) and an MCP endpoint for AI-agent coursework |
import finzdata as fd
df = fd.Ticker("AAPL").history(start="2000-01-01")
print(df.tail())Why it matters for teaching
The biases that quietly fail student backtests
Most student strategies look better than they are because the data quietly cheats. FinzData is built so the three classic biases can be taught with real data instead of warned about in a slide:
| Survivorship bias | Delisted names stay in the universe, so a strategy that held losers to the end shows it. |
|---|---|
| Look-ahead bias | Fundamentals and macro are point-in-time: a query as of a past date returns only what was public that day. |
| Overfitting | Students can grade strategies with the deflated Sharpe ratio instead of the best backtest they happened to find. |
Every response carries a snapshot id, so every student's results are reproducible: the instructor can re-run any submission against the exact same data snapshot and get the exact same numbers.
How to apply
One email, one course key
The instructor emails [email protected] from their university address with the course name, term dates and expected class size. We reply with a course key within two business days.
Students without a course can use the free tier (last 12 months) or buy per-ticker history from $19 in the shop.
Assignment ideas
Five assignments that only work on honest data
- Momentum with and without delisted names.
- Event study around earnings filings using filing timestamps.
- 13F crowding factor.
- Macro regime study with ALFRED vintages vs revised data.
- Agent-built strategy graded by its deflated Sharpe ratio.
FAQ
Common questions
Who qualifies?
Instructors at degree-granting institutions. Research projects qualify too.
Can students keep data after the course?
They keep their downloaded files for coursework. Redistribution is not allowed.
Is there a cost?
No.