did he even renormalize the dataset on the S-1 fill size?