RMSE
Root mean square error; a common regression metric that penalizes large mistakes, used as the Netflix Prize scoring rule.
Root mean square error (RMSE) measures how far predicted values sit from observed targets. For ratings, it is the square root of the average squared error across held-out user-movie pairs.
The Netflix Prize required a 10% RMSE improvement over the Cinematch baseline to win $1 million. The winning BellKor’s Pragmatic Chaos entry reported test RMSE 0.8567. RMSE is standard in supervised learning regression benchmarks alongside other loss functions.