> Is there a reason that the statistics of low satd macroblocks are not included in the predictor? Probably because below a certain SATD, at any sane bitrate, the residual would be quantized to zero (resulting in a skip). Dark Shikari