etf4u
etf4u copied to clipboard
π Python tool to scrape real-time information about ETFs from the web and mixing them together by proportionally distributing their assets allocation
etf4u
A python tool which scrapes real-time information about ETF holdings from the web, and can blend multiple ETF together by proportionally distributing their assets allocation, optionally clamping the maximum amount of assets for the blended fund.
Installation & Requirements
Make sure you're in your virtual environment of choice, then run
-
poetry install --no-dev
if you have Poetry installed -
pip install -r requirements.txt
otherwise
Usage
etf4u [-h] [--clamp CLAMP] [--minimum MINIMUM] [--exclude EXCLUDE [EXCLUDE ...]] [--include INCLUDE [INCLUDE ...]] [--out-file OUT_FILE] [--no-cache] [-v] funds [funds ...]
positional arguments:
funds A list of ETF symbols (or a single one) to scrape
optional arguments:
-h, --help show this help message and exit
--clamp CLAMP Clamp the number of maximum assets to this value, redistributing weights
--minimum MINIMUM Remove all assets with allocation smaller than this number after redistribution
--exclude EXCLUDE [EXCLUDE ...]
A list of tickers to exclude from the scraped portfolio. Pass the tickers directly to the argument (e.g. --exclude AAA BBB CCC) Or pass the path to a text file containing the tickers
--include INCLUDE [INCLUDE ...]
Only include assets whose ticker appear in this list. Pass the tickers directly to the argument (e.g. --include AAA BBB CCC) Or pass the path to a text file containing the tickers
--out-file OUT_FILE Exports the holdings list to this comma-separated (.csv) file
--no-cache Don't use cache files to load or store data
-v, --verbose Increase output log verbosity
Explanation
The tool exports the list of assets as a simple { [asset_symbol] : [weight] }
dictionary format. Use the --out-file
option to export this to a .csv.
When going through the provided ETF symbols, the script checks if there's a bespoke adapter defined to fetch information for that specific fund (some ETFs provides the full list of holdings on their website) - if not found, it uses a generic adapter that scrapes https://etfdb.com/ - the public version of the website only publishes the top 15 holdings for a fund but the script mixes and matches several requests with different sorting criterias to try and get the largest amount of data possible.
All data is cached on a daily basis, meaning that using the same fund in multiple commands will only scrape the real-time information once a day and then re-use the data from disk afterwards. Use the --no-cache
flag to always query real-time data.
You can also use the tool to scrape a single ETF by passing only one symbole to the --funds
parameter and not supplying the --clamp
option.
Creating an adapter
Simply create a new .py
file in the adapters
folder, implementing the following:
- A
FUNDS
variable in the module's scope containing a list of ETF symbols that should be processed with this adapter - a
fetch()
method which takes afund
parameter being the ETF symbol, and returns a dictionary of assets and their weights in the{ [asset_symbol] : [weight] }
format
No need to add anything else, the script automatically checks all modules in the adapters
folder when processing funds. For a practical examples, check the existing adapters.
Example usage
python etf4u ARKK ARKW ARKQ ARKF ARKG --clamp 50 --out-file blend_ark.csv
Adds together all holdings the 5 ARKβs Active ETFs, keeps only the top 50 holdings on the list, rebalances all weights proportionally and exports the assets list to the blend_ark.csv
file
Support 
If this tool has proven useful to you, consider buying me a coffee to support development of this and many other projects.