Footnote explaining how I compiled the Billy Joel popularity ranking:

I found Billy Joel's charts for the past six months ( http://www.last.fm/music/Billy+Joel/+charts?rangetype=6month&subtype=tracks ), and then did some manual aggregation to compensate for inconsistent metadata. For example, Last.FM counts "She's Always a Woman" and "She's Always a Woman To Me" and "She's Always a Woman - Live" as different songs: I wanted to count them as one song and sum their total listeners to evaluate how popular the song actually is. The only way I could do this was to sum each of these by hand -- rather, in a spreadsheet. I was able to assemble a list of the 119 most popular Billy Joel songs this way, which you can find here ( http://earnthis.net/wp-content/uploads/2012/11/billyjoelchart.txt ).

That is the summary of how I found and ranked the most popular Billy Joel songs. For more thoughts on the way I collected the data and how it affected the ordering, keep reading:

---

Some quirks emerge from the list I generated because of how I collected the data. One is that it is a "rolling" list based on the past six months from that exact moment, so the order if I generated the list now may be different right now than what I have written.

Another flaw is that the list I used is not complete. It shows the top 500 most popular "songs," where more than three quarters are either repeats under different titles, or ineligible for my list (see criteria below). This presents a few wrinkles: First, any song or spelling that didn't crack the top 500 (at least 42 listeners for my calculation) did not show up. This means some of the ordering on the list, particularly the bottom half, could ultimately be a bit different.

For example, let's say there's a bootleg live version of the song "Roberta." Maybe the title of that track is listed on a few computers as "Roberta - live" instead of "Roberta." And let's say that this happened on 41 computers, which would be just enough to not crack the top 500, and would thus be unaccounted for by me.

This alternate version would add to its listener total and bump the song up a few spots, from 110 to 106 on the popularity list. It's a minor change, but if there are several different spellings of the same song all hovering just below the list I used, it could considerably alter its placement on the list.

Another consequence of the way I calculated the list is that it could allow for a listener to be counted twice. For example, someone listening to "You May Be Right" and "You May Be Right - Live" would be counted in my spreadsheet twice for "You May Be Right." However, I think this is justifiable and rather trivial.

The final interesting twist resulting from my method of calculation is that some songs on my top 100 did not make Last.FM's top 500 at all! That's somewhat surprising on the surface, but it makes sense: The songs that did not make the list are all demos, and thus would rank among Joel's least popular songs.

Plus, these demos might not have consistent formatting. For example, 160 people might have listened to the song "Cross to Bear." This would be enough to appear on the list. But suppose these 160 people used different song titles on their computer: 40 used "Cross to Bear - Demo," 40 used "Cross to Bear," 40 used "We All Have Our Cross to Bear," and 40 used "Cross to Bear (Demo)."

Because of this, the song wouldn't show up on the top 500! I feel this is an especially large problem for demos which were never released with standard titles on a studio album. Their only release was in a collection of rarities, My Lives.

One final note: I excluded from the list covers, songs not written by Billy Joel (except Why Should I Worry, included by accident), classical songs, and instrumentals.