The Spearman rank-order correlation coefficient (SROCC) measures monotonicity: does the metric order clips in the same sequence that people did. It is the most forgiving validation question because it ignores the exact numbers and looks only at ordering, so a metric that ranks ten clips the way viewers did is perfectly monotonic even if its scores live on a wildly different scale. SROCC is computed by replacing every value with its rank and then correlating the ranks; it runs from -1 to 1, with 1 meaning identical ordering. A key practical advantage is that, because ranks survive any order-preserving stretch, SROCC needs no fitting curve at all, whereas PCC and RMSE always do. It also has a blind spot: it ignores spacing and is insensitive to a constant bias, so a metric can rank clips correctly yet sit far off-scale. SROCC, PCC, RMSE, and the outlier ratio are the four ITU-T P.1401 statistics that grade a metric together.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
SROCC answers one question about a metric: does it rank clips in the same order viewers do. Our seven-volume handbook covers the full video stack; for SROCC start with Volume 1, which explains VMAF, SSIM and PSNR, subjective MOS, and how objective metrics are validated against human ratings. Written by Nikolay Sapunov, CEO at Fora Soft.
Choosing a quality metric and need SROCC results on your own content?
We will validate candidate metrics against viewer scores on your content and report SROCC, PLCC and RMSE side by side. 250+ video products shipped since 2005.
