The similarity algorithm appears to be taken directly from last.fm, which uses some kind of collaborative filtering method on user listening stats to determine artist similarity.
In general, I think this map does a good job of grouping similar artists. Mapping a high dimensional similarity space down to two dimensions is undoubtedly going to result in a few data points that seem out of place, particularly when dealing with content like artist similarity where individual perceptions can vary drastically.
I don't know much, indeed, anything, about most of the bands/artists on this map, so I'm not going to suggest any rearrangement, but I'd be interested to see this concept of "similarity".
Actually, I probably wouldn't be interested, but I hereby announce my position as a skeptic.
No, the one that puzzles me is Sarah Brightman. I can understand them trying to keep her away from the cool kids, but she's got to be feeling like she got off the bus in the wrong neighborhood.
The style of music the Blues Brothers played, the Stax sound, cuts the continuum running from jazz-influenced big band risqué pop to straight-ahead risqué rock and roll. The rhythm section is from the rock world -- electric bass, guitar and piano, with foursquare beat-keeping drumming (as opposed to the sort of thing that, say, a Buddy Rich would do) -- the vocal performances are clearly in the Calloway-Prima-Charles lineage, and the horns are pure big band pop.
OK, I'll go and listen to some more and try to hear the similarities and connections you make. Thanks for taking the time and effort to explain - 'preciate it.
Somehow this map has put The Blues Brothers between Frank Sinatra and The Beatles. To me that's just ridiculous, and a clear indication that their "similarity" algorithm is seriously broken.
They don't seem to explain it either. They talk about their recursive technique for starting from one point and fanning out, but they magically talk about "similarity" without giving any definition or indication of how they compute it.
Personally, I write dozens of programs every week to explore data, and I look at the result for stand-out stupidities. To me, this is one, and it says that the algorithms need to be checked.