Linear regression's most misunderstood metric

The R-square (Pearson's coefficient of determination) is a metric used to evaluate "how good" a linear regression model is. Its possible value is between 0 to 1, looks deceivingly like probability. But it tells a different a story.