Classical statistics originated centuries ago, mainly for purposes of gambling and astronomical navigation. Today we are challenged to predict outcomes of more complex systems, such as social behavior. Machine learning can aid in these tasks, but suffers from many challenges of its own. In this presentation, Mark Kritzman (State Street Associates Founding Partner and CEO of Windham Capital Management) and David Turkington (head of State Street Associates) – both coauthors of a new book “Prediction Revisited: The Importance of Observation” – revealed the building blocks of data-driven prediction from a fresh perspective whereby data represents “experiences” rather than variables, starting with the information theory foundation for linear regression and building to more sophisticated nonlinear and conditional enhancements to predictive modeling.