Investments toolkit

My personal collection of scripts and automations to help me handle my investments.

Note that this repository is also used for my own educational purposes (e.g. testing new libraries or technologies). I'm deliberately implementing myself some REST API clients and calculations (technical indicators) that are already available in other libraries. When I find alternative implementations, I try to link them throughout comments in the source code.

Summary

• Calculates the correlation matrix of the assets in my portfolio (and optionally compare it with an extra list of assets)
• Calculates the Average True Range (ATR) volatility technical indicator
• Uses ATR to calculate a trailing stop loss
• Unified (HTTP) endpoint for getting the latest price from different sources and calculating the stop loss

Generic features

• Caches the results of HTTP requests locally (file system backend from requests-cache). Useful for local development
• Caches assets metadata and price history remotely (Google Firestore). Useful for production deployments
• Exposes some of the graphs through HTTP (FastAPI). Useful for creating links from my Google Sheets
• Exposes some of the raw outputs as CSV through HTTP (FastAPI). Useful for embedding the results in Google Sheets
• Containerized as a Docker image and ready to be deployed in the cloud (Google Cloud Run)
• Interactive graphs (Plotly)

Portfolio correlation

Motivation

Calculates a correlation matrix between multiple assets. As part of my trading strategy, I am interested in knowning the current level of correlation between the assets that I already own. Additionally, when I am opening new positions, I want to select the assets with the least correlation possible to my current portfolio.

Specific features

• Calculates a correlation matrix and plots it on an interactive graph
• Uses clustering to order the correlation matrix, making it easier to interpret it
• Serves the output via HTTP either as a graph or as a CSV

Usage examples

Programmatically (Jupyter notebook)

This graph below was generated from one of the example files (Jupyter notebook).

Via HTTP server (standalone graph)

The same graph as above, but returned in a standalone HTML page. Useful for creating a link with your assets from Google Sheets while still keeping the interactivity.

Other assets with portfolio

During the screening process before starting new positions, it is possible to provide an extra list of assets other than the ones in your portfolio. Those assets will be appended at the end of the correlation matrix, after the clustering step is done. This facilitates identifying which one of the extra assets have the lowest correlation to the assets in your current portfolio.

Via HTTP server (CSV)

Useful for embedding the results in Google Sheets (using IMPORTDATA()).

Latest price

First I started using =GOOGLEFINANCE() in my Google Sheets spreadsheet to get the latest price and a % variation for the day for each asset. At some point, some of my new stocks were not available in Google Finance and I had to use =IMPORTXML() to parse the public page from my broker to retrieve the same information. And then, I started using multiple brokers and I had to introduce =ImportJSON() for those as well, which made it quite cumbersome to maintain within the spreadsheet.

In this project, I hide this complexity from my spreadsheet by just exposing a /price/<asset_id> to handle all the different data feeds.

In other words, I could replace this:

=GOOGLEFINANCE(CONCATENATE(D20, ":", SUBSTITUTE($F20, "_", "-")), "changepct")/100
=INDEX(ImportJSON("https://oaf.cmcmarkets.com/instruments/price/X-AASOB?key=<KEY>", "/movement_percentage"), 2, 1)/100
=VALUE(SUBSTITUTE(SUBSTITUTE(SUBSTITUTE(IMPORTXML("https://www.avanza.se/aktier/gamla-aktiesidan.html/666686/swedencare","//span[contains(@class, 'changePercent')]/text()"),",","."), " %", ""), "+", ""))/100

With this:

=VALUE(ImportJSON(CONCATENATE("https://<HOST>/price/", A3, ":", C3), "/change_pct", "noHeaders"))/100

Trailing stop loss

Calculates a trailing stop loss based on the Average True Range (ATR) indicator. I use this stop loss both for position sizing and for calculating my exit point for my positions. Automating it here is my first step towards having this project automatically update my stop losses in the brokers I use. Also exposed as an HTTP endpoint to be easily embedded into Google Sheets.

This graph below was generated from one of the example files (Jupyter notebook).

Technical indicators

Average True Range

The average true range (ATR) is a technical analysis indicator, introduced by market technician J. Welles Wilder Jr. [...] (it) is a market volatility indicator.

Source: Investopedia

---

Architecture

Data feeds

Clients to retrieve asset information and historical prices from different sources. Currently, the following ( minimalist) clients are implemented:

• Avanza: a Swedish broker
• CMC Markets: an international CFD broker

Persistence

Clients to interact with Google Firestore and for setting up HTTP requests local cache.

Models

Python classes to represent assets, price bars and data sources.

Formulas

A module to hold formulas, such as the calculation of technical indicators.

Deployment

I host an instance of the HTTP server for my personal use on Google Cloud Run. Both my uses of Google Cloud Run and Google Firestore for this project falls under their free tier.

Check .env.example for the environment variables necessary to spin up an instance of the server.

Development

I use ngrok when I need to connect Google Sheets with a local instance of the HTTP server.

Useful commands:

• make lint: runs mypy
• make type: runs black and flake8
• make serve: starts a uvicorn server with live reload enabled

License

This project is licensed under the Apache License, Version 2.0: http://www.apache.org/licenses/LICENSE-2.0.