Generally speaking when I first use Picard on a set of newly ripped tracks (where there are no MB albumid tags), clustering and lookup find the right album and match the tracks to the album very well.
But in a few cases where the ripper has named all files 01 Track 1 etc., it doesn't work well. This is because it cannot use track-title and relies on track-length more than track-number.
Track-length has a weighting of 8 and Track-Number a weighting of 6.
Changing the weightings in the Picard code is easy to do, but might have unintended consequences.
Has anyone got any knowledge of how the current weightings were determined and any suggestions as to how to tweak the weightings to make track number have greater importance than an almost perfect track-length?