vanekpetr / investment-funnel

Investment Funnel 📈 is an open-source python platform designed for an easy development and backtesting of outperforming investment strategies.

Hey, Robots!

Currently, I'm working as a Machine Learning Engineer @ Go Autonomous in Copenhagen where I deep dive into topics of NLP, Computer Vision and MLOps.

In a free time, I am helping as a teaching assistant at DTU and KU, co-supervising master and bachelor theses within OR, ML and asset allocation as well as applying this knowledge to create an open-source algorithm-based investment advisor.

🔧 Technologies & Tools

🚀 GitHub Stats

🛸 Check this out

investment-funnel's Issues

Min & Max asset weight constrain

Start a pyproject.toml file

Add some extra statistics on the backtesting page

Enable to view the full name of the fund

Load downloaded eToro data from file

Markowitz model with MOSEK (Or different risk measures)

Change page naming

Rolling MonteCarlo simulations

Add a free license option for the CVaR model

Better description for each page

Implement Infront Data

Let's use Infront's data instead of the one from Yahoo finance

Add 1 over N model

"Loading" note when computing backtest

Use dependabot for your dependencies...

copy

# To get started with Dependabot version updates, you'll need to specify which
# package ecosystems to update and where the package manifests are located.
# Please see the documentation for all configuration options:
# https://docs.github.com/github/administering-a-repository/configuration-options-for-dependency-updates

version: 2
updates:
  - package-ecosystem: "pip" # See documentation for possible values
    directory: "/" # Location of package manifests
    schedule:
      interval: "weekly"

into .github/dependabot.yml

Change to MIT license

Rewrite/Add CVaR model with MOSEK

Navigation bar in the AlgoStrata colours.

Highlights should be in the AlgoStrata's orange if possible.

Names instead of ISIN codes on the backtesting page.

Simplify CVaR Targets generation methods

Find assets and Download data

1. Go to investment-funnel/investment_club/benchmarks.py
2. Find all assets used in Lysa and NORD Investments portfolios

• assets used in Lysa portfolios are already found
• for NORD, find assets in portfolios 5, 13 and 20 (Kourosh can provide screenshots)

4. Find corresponding tickers for all assets
5. Download data for these tickers from Yahoo or any other source (we need daily prices after adjustments, ideally in USD)

Just one graph/one dropdown on the second Feature selection page.

Search function on the first page

Better loading sign for backtesting page

Add horizontal red line representing 0% for all graphs

Adjust description of risk classes to be readable

Use Algostrata's data

Create NORD and Lysa benchmarks

1. Go to investment-funnel/investment_club/benchmarks.py
2. Use data provided by @jorgeebg96 and portfolios provided by @VanekPetr to create data for NORD and Lysa portfolios

• start with a datapoint for which we have data for all downloaded assets
• in this datapoint, create portfolio based on weights provided in portfolio dictionaries
• use weekly returns and construct the portfolio value in 4 weeks
• after 4 weeks rebalance the portfolio to get again the weights provided in portfolio dictionaries (or use any other strategy before you rebalance back to starting weights)
• repeat this until the end of dataset

3. If case of any question you can contact @VanekPetr or Kourosh

Possibility to select specific dates on the second page

Show optimal portfolio on the graph with efficient frontier

More clear/dark font for all graphs' descriptions (for example "risk class" font for the first graph)

Two Graphs for the second page

Updated README with all the previous research and relevant MOSEK cookbook links

Remove global variables from Plotly callbacks

Possibility to change rebalancing periods

Adjust title and axis of our graphs

Could be:
Annualised average returns
Annualised standard deviation of returns (risk)
Relationship between risks and returns for a time period from "Date" to "Date" (date without a time)

Double check Monte Carlo and investigate if possible top merge with Bootstrapping

Check data pre-processing

Delete all tickers which has not available data in the beginning but also at the end of the selected intervals.

Revisit ci/cd

@VanekPetr I don't think you should auto-tag and publish releases with every successful run of main. That's getting too much?

We should release if two conditions are fulfilled

1. All tests etc. have passed (I will make them part of the ci/cd process)
2. A user has introduced a tag

I take inspiration from https://github.com/cvxgrp/.github

Experiment with backtest and colours

Github action which builds docker image and pushes it into registry after a new tag is introduced

@VanekPetr You should get an account on docker hub to publish the images you construct (every time you tag)

Code Clean-up

Let's clean up the code to a higher standards

Add all possibilities for backtest

Add current mix into the results

Inside a table. Possibility to change a period?

Possibility to select training and testing periods on backtesting page

Use dates instead of numbers for the composition graph

Split the code-base into public and private

Split the project into the open-source part with all our models and algorithms and into a private part with our deployment and data.

Deployment to AlgoStrata's Kubernetes

Create portfolios with weights

Create 3 NORD investments and 3 Lysa portfolios for our benchamrking

1. NORD portfolio 20
2. NORD portfolio 13
3. NORD portfolio 5
4. Lysa portfolio 100% in stocks
5. Lysa portfolio 75/25 in stocks/bonds
6. Lysa portfolio 50/50 in stocks/bonds

Change colours in the composition graph

Save results on the backtesting page when switching to another one