I try to use MB Picard to detect a release with very similar filenames.
Unfortunately “Lookup” doesn’t work for this release and I have to use “Scan”.
Some of the filenames will be assigned to the same track, even if the track number in the filename is clearly different:
In the above picture you can see that the first track “Are you Ready? (full)” has two entries.
The first one with track number 01 is the correct one.
The second entry has the tracknumber 13 and should be assigned to track 13 “Are you Ready? (underscore)”
How can I adjust the “detection” settings to prioritise the track number in the file name?
Especially if a track with the correct track number and a similar filename is available - and not yet assigned - for such a release?
You can’t currently. I would just match files manually in this case. This release in combination with scan is a rather special case.
For scan Picard queries AcoustID for matches, which will give a list of recordings¹ with metadata. Picard selects the best matching recording based on metadata similarity, which in this case doesn’t include the track number. It then matches the file to the release based on the found recording ID. This matching will use track length or number to decide between tracks that have the same recording ID (same recording on same release happens, but is a rare case).
The problem with your release is that there are two separate MB recordings, but they share an AcoustID. E.g. Track "9b32fcc0-f2be-4519-9cb6-2e567c381af4" | AcoustID for both versions of “Are You Ready?”. The metadata comparison from AcoustID result gives the same similarity, so it picks the same recording ID in both cases. Matching to the release then is fixed to the recording ID, which always gives the same track.
I need to check, but I think AcoustID also gives the position of the recording on linked releases. This probably could be used to improve the selection of the proper recording ID from the AcoustID result.
UPDATE: Yes, we should be able to improve this with the data provided by AcoustID. I added
1) To be precise it’s a list of AcoustIds and the linked recordings for each
I’ve voted for it, although the current behavior also has an advantage. I would never have otherwise discovered that a 10-CD Irish folk release contained two duplicate recordings (different artists and slightly different titles). A scan assigned the duplicate recordings to the same track, which is how this fact was even noticed.
