Welcome to our Magic: The Gathering card prediction models with Artificial Intelligence!
Our vision is to bring AI to Magic and help evolving this amazing game!
This site serves as demonstration to many Machine Learning models that are trained using the information from cards the Scryfall Magic The Gathering API.
The process goes something like this: First we get all Magic cards from the Scryfall API, we then transform those cards into data prepared for machine learning, converting each card into hundreds of different data points. That's when we proceed to show that data to differe tmachine learning models alongside with the target we want to predict. The model will start learning patterns hidden within the data and improve with the more data we feed it, until it's able by itself to provide accurate predictions.
As of today, we have 3 types of models and use 3 different algorithms:
- BINARY models, MULTICLASS models and NUMERIC models, to define the type of prediction we want to do (a YES or NO question vs a specific category prediction)
- Logistic/Linear Regression, Gradient Boosting and Deep Learning (Embeddings and bidirectional LSTM network)
If you want to learn more about the process, here's an article
- Install the libraries required from
requirements.txt
- run the app using
streamlit run app.py