Combining RMSE and R-squared for Better Forecast Model Evaluation

mdancho84 · x · 2026-09-01

In forecasting, maximizing R-squared (variance explained) is not always ideal as it prioritizes fitting variance over stability. The author suggests combining R-squared with Root Mean Squared Error (RMSE): RMSE measures model stability, while R-squared indicates how well the model captures peaks and valleys. This dual-metric approach provides a better picture of model performance.

Original post →

More from Research

Research channel →