UPDATE - 25 Sept 2025: This project has reached its EOL. I will no longer work on this project after understanding more of the nature of the product being traded.
Trading NVIDIA knockout derivatives with Machine Learning techniques. Random Forest will be the main algorithm used here.
- Completed basic data pipeline from Alpaca API.
- Completed basic triple barrier labelling with 5 labels for Turbo Knockout derivatives, such as the one listed within the KID.pdf.
- Completed basic ML model that seems to overfit to small sample of data obtained.
- Completed basic backtesting to validate ML model. Model shows overfitting patterns.
- Discovered Turbo knockouts are simply against the average retail trader. At my current knowledge, it is not worth trying to challenge the giants as I will set up my own failure.
- Discovered that traditional Technical Analysis did generate high Sharpe ratios within trading Turbos with the labels generated from my adapted triple barrier system. If this repo is to be continued, an interesting point to look would be to use ML as risk manager instead of a predictor.
You may use my work in any way you like. I would simply be happy with just a mention of my GitHub account within your contribution. I am also not liable for any losses made on your side with my work. I will mention that this project had never generated any profit, nor is it intended to ever be in live production.