April 2021: Income Report & Retrospective
How much I made in April 2021 participating in Numerai tournaments and as an early stage solo-entrepreneur.
I’m officially earning more money participating in the Numerai tournament than in my day job.
Profitable enough to leave it and dedicate full-time to the tournament. A tournament where I invest less than 10 hours per week.
This train of thinking drives to one conclusion, why not bootstrapping a business as a solo-entrepreneur, try to make it profitable, and after having two extra incomes start seriously thinking about leaving the day job.
Here is what I have learned and earned in April 2021
Income Report for April 2021
Numerai Classic Tournament Payout
Initial stake: 790 NMR ($51,350)
Payout: 370 NMR ($24,050)
Average weekly return: +10.78%
Monthly return: +50.77%
April has been the most profitable month so far. I had received an approximate payout of 370 NMR ($24,050). This is +226 NMR (+157.90%) than last month.
The average weekly return has been around 10.78% with a monthly return of 50.77%
Numerai Signals Tournament Payout
Initial stake: 4.65 NMR ($302.25)
Payout: 0.16 NMR ($14,10)
Average weekly return: +0.85%
Monthly return: +3,44%
It’s been the first month staking in the Signals tournament. I only have one model and it’s heavily based on the example Numerai provides.
What happened in April
Automatized all Numerai predictions submission
I had almost all models automatized but a couple of them. This month I invested few hours to fully automatize the process. Now I don’t have to do anything, every Saturday all the models are submitted while I’m having a beer and chilling out with friends.
I have decided to stop one of my best models at Numerai
The results were promising averaging a 13.26% weekly for a monthly performance of 64.35%. But the approach was too complex. The model should be run in a GPU instance which makes it difficult to automatize using Numerai Compute. I was running everything in Google Colab but after some thinking, I decided to stop the manual process and move one.
Open source contribution to yfinance
Participating in Signals means downloading your own financial data. I’ve been using yfinance to download Yahoo Finance data in Python from the very beginning. The threads implementation was quite buggy so I submitted a PR expecting to get it merged soon: https://github.com/ranaroussi/yfinance/pull/685
Also, I created a package called ysignals to make it easier to build models using data downloaded from Yahoo Finance. It’s pretty simple but has potential.
I have started investing time bootstrapping a business
During the last years, my interest in finances has increased significantly. After some research looking for a tool to keep track of my assets, I ended creating a spreadsheet.
You can see here one of the plots I use to track asset growth. (Yes, I know, I have to rebalance it but is expensive in term of taxes)
After more than a year of playing with the spreadsheet and fighting against some limitations. I have decided to start building a service to replace that old spreadsheet. The service is still early stage but I’ve been sharing it with close friends and the feedback has been quite positive so far.
May 2021 goals:
Try a couple of new approaches for Numerai models. — I want to retry an approach I tried months ago but with some improvements. I hope this time will improve the performance.
Invest some hours in feature engineering for Signals tournament. — As I said, my approach is quite simple and I’d like to start creating new features to see how difficult is to get a good model there.
Finish the main screen of the assets tracker. — Continue moving forward with the service to track assets.