How to Predict the Performance of NBA Draft Prospects

Abstract

The authors describe a new mathematical system for predicting outcomes of NBA draft prospects based on a statistical concept called relevance, which gives a mathematically precise and theoretically justified measure of the importance of a previously drafted player to a prediction. They also describe fit, which measures a specific prediction’s reliability, thereby offering guidance about how committed a team should be to a draft prospect. And they show how fit, together with asymmetry, identifies the uniquely optimal combination of previously drafted players and predictive variables for each individual prediction task. The authors argue that their new relevance‐based prediction system addresses complexities that are beyond the capacity of conventional prediction models, but in a way that is more transparent, more flexible, and more theoretically justified than widely used machine learning algorithms.

Keywords: Adjusted Fit, Asymmetry, Central Limit Theorem, CKT regression, Codependence, Fit, Gaussian Kernel, Information Theory, Informativeness, Lasso Regression, Machine Learning, Mahalanobis Distance, Model-based Algorithm, Model-free Algorithm, Nearest Neighbor Partial Sample Regression, Relevance

JEL Classification: C00, C01, C02, C10, C13, C18, C45, C50, C51, C55, Z20, Z29