Currently listenbrainz has problems dealing with recordings that appear on multiple releases which leads to statistics that do not reflect what a user is actually listening to (for example: a track that was released as a single and then later included in an album may match to either the single or the album, regardless of where the user listened to it, and then this is used in the stats).

One proposed way of handling this situation is allowing users to select the release/release group when linking a listen to allow them to correct mistaken matchings. This would likely best be paired with factoring in submitted release mbids etc into the stats as well