Wine and Math by Lars Verspohl
Wine and Math is a visual story trying to understand if a statistical model can help predict the quality of a wine based on its physiochemical properties like Acidity, Alcohol, pH, ...
It started out with the possibly naive assumption that - given the right data - it should be possible to gain somewhat accurate quality estimates. It walks through an example dataset in that spirit, introducing the data at hands as well as the inner workings of a Logistic Regression - with the help of a sloth.
The story eventually reaches an interactive wine quality simulator allowing the user to play with the data's physiochemical properties in order to compose the highest quality wine.