Get started

Quickstart

On this page4

This page takes you from nothing to your first backtest: install → load data → write a rule → backtest, in about five minutes. Every snippet runs as is.

1 Get data
data.get('price:adj_close')
Dates down, stocks across
2 Write a condition
close > close.average(60)
True = hold that day
3 Backtest
sim(position, resample='M')
Returns, drawdown, trades
4 Go live
PortfolioSyncManager
- BuyAAPL
- BuyMSFT
- SellNVDA
Updates and orders every day

	AAPL	MSFT	NVDA
10/27	385.93	610.66	223.87
10/28	382.29	610.30	221.24
10/29	387.94	609.30	219.47

	AAPL	MSFT	NVDA
10/27	True	False	True
10/28	True	False	True
10/29	True	True	False

Pick where to run it

For a first try, use Google Colab: it runs in your browser, with nothing to install.

Open a new notebook in Google Colab and run this in the first cell:

!pip install finlab

You need Python 3.10 or later. In a terminal:

pip install finlab

If package versions conflict, use Docker or install into a fresh virtual environment (python -m venv).

The image comes with FinLab and JupyterLab:

docker pull finlab/jupyter-finlab
docker run -p 8888:8888 finlab/jupyter-finlab

Open JupyterLab at the address the terminal prints.

Run your first backtest

Load closing prices

data.get() loads a dataset by its "table:field" name. The first run opens your browser to sign in to FinLab; a free account is enough.

from finlab import data

data.set_market('us')
close = data.get('price:adj_close')
close.tail()

date	AAPL	MSFT	NVDA	AMZN
2026-10-26	382.38	612.12	221.88	249.09
2026-10-27	385.93	610.66	223.87	253.1
2026-10-28	382.29	610.3	221.24	254.6
2026-10-29	387.94	609.3	219.47	243.79
2026-10-30	383.18	599.84	210.21	240.35

You get back a table: one row per trading day, one column per stock (only a few are shown). Every FinLab dataset has this shape, including fundamentals and filings.

Write a stock-selection rule

Compare whole tables with Pandas. This line means "the close is at a 300-trading-day high":

position = close >= close.rolling(300).max()

The result is a table of the same shape, each cell True or False; True means hold that stock on that day. A comparison works on the whole table at once, as this example of "close above its 60-day average" shows:

close Closing price

	AAPL	MSFT	NVDA
10/27	98	51	120
10/28	101	50	118
10/29	104	49	121
10/30	103	52	117

close.average(60) 60-day average

	AAPL	MSFT	NVDA
10/27	100	52	115
10/28	100	51	116
10/29	101	51	117
10/30	101	51	118

position What to hold

	AAPL	MSFT	NVDA
10/27	False	False	True
10/28	True	False	True
10/29	True	False	True
10/30	True	True	False

Each cell is one stock on one day. A comparison works on whole tables at once, matching dates and stocks for you; each True is a stock to hold that day. Numbers are illustrative.

Backtest

Hand the rule to backtest.sim(). resample='M' checks the rule at each month end and trades on the next trading day:

from finlab import backtest

report = backtest.sim(position, resample='M')
report.display()

report.display() draws the full report. This is what the code above produces; you can explore it right here:

Backtest report · interactive Open full size

Read the report

The top of the report shows the equity curve and the figures people check first:

Metric	What it means
Annual return	The whole backtest's return, as an average per year
Max drawdown	The deepest fall from a peak: the hardest stretch to sit through
Sharpe ratio	Return per unit of volatility; higher is steadier
Win rate	The share of trades that made money

To get these figures in code, use report.get_stats(); to see every trade, use report.get_trades().

Next steps

Write better rules Take the top names by rank, combine conditions, set entries and exits.
Load more data Fundamentals, filings and how to find a dataset's name.
Tune the backtest Rebalance frequency, stops, fees and the trade log.
Run it daily Rerun the strategy every day and get rebalance alerts or place orders.

Not sure what a dataset is called? Search fields in the data catalog, or install the FinLab Skill in an AI coding assistant (Claude Code, Cursor and others) and ask "which datasets cover revenue?".

Was this page helpful?