R-squared

The fraction of a dataset's total variation a least-squares fit explains.

$R^2 = 1-\mathrm{SSE}/\mathrm{SST}$ , where $\mathrm{SSE}=\sum r_i^2$ is the sum of squared residuals left over and $\mathrm{SST}=\sum(y_i-\bar y)^2$ is the total variation before fitting. $R^2=1$ means every point sits exactly on the fitted line; $R^2=0$ means the fit does no better than predicting the mean $\bar y$ for everything.