Dave Conway Piano





Guild Wars 2:

Tomato Yesterday




Some Stream Stats

22 April 2017 08:00 EDT under
Music, Twitch

Hot damn, I'm behind on my blogging. March must have been so bad that I didn't even take time for a brief month-in-review post. Oh well, time to do a bit of catching up.

The other day in stream I was talking about how it would be cool to do recordings of the most-requested music on the stream and throw 'em on YouTube or something, so I ran some queries. I no longer think that that would be a cool idea. At least not yet.

I know that my meager little Twitch followership has far-reaching tastes in music, but I didn't realize how badly the breadth exceeded the depth until I ran some of these queries. Let's take a look at some basic stuff:

Query/calculationResult (all years)Result (2017 only)
Total number of requests made (all years)1,150395
Requests made via library vs. new PDF517 library vs. 633 new185 library vs. 210 new
Percentage of requests that are new PDFs55.0%53.2%
Number of works in the library528 (total inc. 2017)185 (added in 2017)
Number of unique Twitch users who have made at least one request17868
Most requests made by a single Twitch user9651
Most requests made by a single Twitch user who isn't also my girlfriend7628
Most requests made during a single stream3324
Most unique total works requested by a single Twitch user (actually not my girlfriend!)7541
Number of unique Twitch users who only request new works7024
Number of unique Twitch users who only request from the library00

Now, these numbers are obviously trivially small as far as Twitch numbers go, but there's enough here for me to recognize an obvious pattern: the library isn't getting used all that much. Over 50% of all requests are brand-new things, and just about 40% of viewers who make requests never touch it at all. There's no problem with that, in and of itself, but it does take a lot of time to catalog so many scores, and if it's a system people aren't interested in, I may have to scrap it and come up with something else, instead. What it boils down to is that there are tens of thousands of individual pieces of music out there written or arranged for piano, and having 500-something of those in the library -- even though it seems like a decent number -- doesn't scratch the true surface.

One idea I'm toying with is allowing requests to be made directly from Twitch chat. A lot of people come into the stream, completely ignore both giant "this is how you request shit" indicators, request something in chat, and then leave the instant I tell them they have to go to another page. I'd write a bot that recognizes "!songrequest" or "!sr" or "!request" commands followed by a URL from an approved domain (Google Drive, Dropbox, MuseScore, Ichigo's, etc.), and it'd pop those right into the queue. I could make my PDF catalog publicly readable on Google Drive so that instead of maintaining a library in my website's database, people could just get the direct links to Drive files and request using those. The only concern I have with that is that I'd have to be careful to mark certain scores as private, because I think a lot of pirated sheet music (yeah, that's a thing) ends up on my iPad, and I'd prefer not to freely distribute that.

I'm curious to know what viewers think about that versus the current system.

Regarding making decent recordings of some of the most-requested music, well...you people simply aren't repeat requesters. Out of 395 requests made this year, the two most popular, Hiromi's "Green Tea Farm" and my own "Toccata of Storms" arrangement, have each been requested a whopping four times. If we look at the top ten most-requested pieces, it's a two-way tie for first place, and an eight-way tie for third, which somehow includes the Diddy Kong Racing main theme:


So any kind of little mini-concert or recording series based on that is gonna be a no-go. People just like a huge range of stuff, and don't really have favorites (or at least don't want to hear their favorites all that frequently).

When I look at source media instead of individual work, though, I can paint a better picture. These are the top eight sources from which works are requested:

  1. The Legend of Zelda: Ocarina of Time
  2. Final Fantasy VII
  3. Final Fantasy VI
  4. Final Fantasy VIII
  5. Chrono Trigger
  6. Pokémon
  7. Undertale
  8. My Neighbor Totoro

The bottom seven are all close competitors, but TLoZ:OoT was in first place by a mile (66% more requests than FF7 had).

I want to do something like a monthly "fun night" on stream, just to change up the ol' formula every now and then. Something like a "Top 8 Night," where all requests must come from those eight sources; or a "Random Night," where you just click one button to make a request and it pops a completely random work into the queue.

Do you guys have any fun ideas?