The amount of bad data on last.fm and their lack of interest in cleaning it up makes me sad. You can see a small example just by looking at the current top songs: http://www.last.fm/charts/track
Morning Mr Magpie is in position 8 instead of position 4 where it should be, because 8105 people scrobbled Morning Mr. Magpie with a dot. Tip of the iceberg.
Really egregious are the albums with no scrobbles because every track on the album has "(Album Version)" or some such affixed to its name.
Once on a Pandora station they played me a track by a band called Burial, instead of the dubstep artist Burial I had added to the station. So I sent them an email and they thanked me and told me the issue was very important to them and they cleaned their data. Last.fm doesn't bother, even when it causes their top-20 charts to be completely wrong.
Actually, we do clean that data up with our corrections system. We have a large fingerprint database which allows us to work out when a track is the same and we update the charts across the site to reflect this. We even expose this as a service in the API:
What you're probably seeing there is tracks which are so new that we've not determined similar fingerprints for them yet, or those changes have not propagated to the charts (we use a lot of caching as you might imagine). In fact, checking the chart now it seems that that particular problem has cleared up.
Also, artist disambiguation is a _hard_ problem, but it's something that we'd very much like to address.
Monster should be in place 3 with 14,757 listeners. Instead it's in places 12, 16, 27, and 61. It was a top 20 track for some number of weeks, but it didn't show up in the top 20.
In fact, checking the chart now it seems that that particular problem has cleared up.
I'm not sure what you're referring to. I've never seen this problem clear up.
Also, artist disambiguation is a _hard_ problem
I understand that, and I'm not criticizing that. I'm just pointing out that Pandora is making an effort to work on that hard problem while Last.fm ignores very easy problems. Specifically the easiest problem: albums in which all the track names are wrong because they all have "Explicit" or "Album Version" or some other string appended to their names.
The real killer for auto-correction is that it seems to require many users to work - artists in the thousands don't seem to ever get corrected.
It also doesn't help that Pandora has terrible tagging. I have a number of "Horner, James - Braveheart, Film Score"-esqe scrobbles that originate from them.
Morning Mr Magpie is in position 8 instead of position 4 where it should be, because 8105 people scrobbled Morning Mr. Magpie with a dot. Tip of the iceberg.
Really egregious are the albums with no scrobbles because every track on the album has "(Album Version)" or some such affixed to its name.
Once on a Pandora station they played me a track by a band called Burial, instead of the dubstep artist Burial I had added to the station. So I sent them an email and they thanked me and told me the issue was very important to them and they cleaned their data. Last.fm doesn't bother, even when it causes their top-20 charts to be completely wrong.