Holger K. von Jouanne-Diedrich • 5/11/2026

Learning Data Science: Why a High R^2 Can Be Misleading

This article discusses the pitfalls of relying solely on a high R-squared value to judge regression model quality. It explains the formula for R-squared, shows how the mean model yields zero R-squared (underfitting), and how a high-degree polynomial can achieve perfect R-squared but overfits the data, performing poorly on new data. The post emphasizes that a high R-squared does not guarantee a useful predictive model, making it relevant for data science and machine learning practitioners.

0 comments

#Data Science #Model Evaluation #Overfitting