We’ve been frustrated by a few things about music charts recently. One is the issue of chart integrity, the lines between manipulation & engagement and how to keep such a system honest while rewarding the right people. Another is the terrible quality of most music services powered by Twitter.
Most of them have serious problems. Some don’t reveal the methods or sources in which they’ve gathered their data, just publishing the chart and hoping people would buy it. That prevents the web at large from auditing and finding errors in their methods and work.
Others don’t tell us anything new and don’t account for the social and interactive nature of Twitter, lumping it all together as ‘activity’. So we end up with U2 as the top artist (or Michael Jackson, for the foreseeable future). Finally, the remaining set of tools just implements a file sharing layer on top of Twitter, then allowing others to search for uploaded media.
Given all this, we did what we usually do when there is music/web stuff out there that sucks: we built our own!
Check it out here: Twitter Music Chart
How does it work?
We monitor Twitter for links pointing to tracks on the Hype Machine. We then give each of those tweets a number of points based on the number of followers (and the ratio of friends & followers) that person has. Finally, we add up all the points and figure out which tracks are tweeted by either the most influential twitter users, or by the largest group. Simple.
Just tweet a link to any track on the Hype Machine and it will alter that track’s position in the chart. You can also check your Twitter score on the page to see how many points your tweet would add. Try it!
How is this chart different?
Our interactive Twitter music chart addresses the problems we discussed above. It’s easy to audit as you can see how many points were added by what users and when. It also accounts for the social nature of Twitter by carefully weighing more influential users vs. groups of less influential users. Finally, it presents an easily digestible overview of what’s going on in an interactive way. We welcome everyone tweaking and changing the chart with their tweets – that’s the point!
How is the Twitter user score determined?
We use a formula designed to spotlight tracks tweeted by influential Twitter users as well as those tweeted by groups of people overall. Here is the actual formula:
round(( 1/3 * (twitter_followers / 10) ^ 0.5 ) * (twitter_followers / twitter_friends) * 10))
The follower/friends ratio in this case has a floor of 0.25 and a ceiling of 3.0 to limit certain extremes. We really like the results we’ve been seeing with this approach, but we also know there are many serious math geeks out there. If you have comments on creating a better formula that would better capture the different kind of tweeting activity, let’s chat.
Oh yeah, and one last thing: @hypem will now tweet whenever a new artist or track makes it to the top of any of our popular ranks, so follow us to stay in the loop.
Would love to know what you think!
Is it just tweets that came via “loved XYZ on HypeM”? If that’s the case, why not just rank by loved on the site and leave twitter out of it? If that’s not the case and you’re looking for references on Twitter that don’t originate on HypeM – that’s cool.
July 8, 2009 11:50 am
All tweets that point to tracks on the Hype Machine are eligible, so yeah, we are welcoming everyone to participate.
July 8, 2009 12:11 pm
If you’re a Hype Machine aggregated blog, can you link your Twitter account to your blog’s page on HM?
July 8, 2009 12:14 pm
Hmm, what would you expect that to do?
July 8, 2009 12:21 pm
I think the twitter chart is a great idea, but it seems weird that one guy with 4k followers can single handedly place a track in the top 10!!!
(See #9 – Harlem Shakes)
Could number of followers have some sort of a diminishing return cap on it as you increase beyond the 500-1000 range?
July 8, 2009 12:23 pm
Congrats on the launch! Looks like a useful addition to HypeM. At first glance, it seems more relevant to me personally than the other popularity lists on your site, which tend to skew towards electronic and club music whereas the Twitter list is more my taste. It will be interesting to see how the Twitter chart evolves over time as more people tweet HypeM links (great viral marketing for HypeM, by the way).
One possible concern is that the people with the most followers on Twitter could dramatically alter the results. For example, the top song on the Twitter chart currently has 469 points. One tweet from Ashton Kutcher (@aplusk) is worth 5108 points today. So, it looks like Ashton could single-handedly take over the list. I’m not sure I want the star of “Dude Where’s My Car” to have that much influence over my listening options. Then again, maybe you already account for this in the details of the algorithm. Either way, I’m psyched for the new option.
Nice work!
July 8, 2009 12:24 pm
The same sort of value that having Delicious and Technorati counts up on each blog’s page. I’d actually argue, potentially more.
Though those are both services that don’t require the blog to partake in it — more passive tracking vs an active engagement.
THOUGH aren’t active relationships much more valuable for blogs ultimately? Encourages conversation which is the biggest difference between “publishing” and “blogging,” no?
July 8, 2009 12:26 pm
Hey didn’t you listen! There’s a floor to the extremes! So no worries!
Anyway, personally I don’t really care for Twitter. Twitter is for people who don’t have a blog/are too lazy to have a real blog. Haha, said with a semi-ironic glimpse in the eye. 😉 You can twitter through a regular blog, just it that it’s not limited to like three words and doesn’t appear on Twitter’s servers. But please do win me over to Twitterania. I just don’t see the use of it yet.
Hence I find this chart fairly unnecessary. Having two charts on Hypem? Nah. That’s confusing and a bit of wish-thinking forced branching out that I don’t think fits Hypem. But by all means keep it. When I grow too old to rightfully be able to manage a blog I’ll start music-tweeting instead. You’ll find me on that chart then for sure!
July 8, 2009 12:47 pm
This is awesome. Well done guys. Love your approach and that you’re SHOWING the math behind it. Great move.
One question, are tracking shortened URLs, or just straight up http://hypem.com/track/whatever urls?
July 8, 2009 12:51 pm
It tracks all urls, shortened by almost any shortener. Thanks to BackType for that: http://backtype.com
July 8, 2009 12:55 pm
Anthony + HypeM team –
This looks like an awesome idea. I’m sure you’ll refine it as you move along, but so far it looks like a great start.
Looking forward to seeing what else you guys roll out.
(How about the ability to grab shortened URLs for tracks for easy Tweetin’?)
July 8, 2009 1:30 pm
This is great. One wish – could you make a chart RSS feed ?
July 8, 2009 3:16 pm
I’m an admitted mathophobe so forgive me if I’m just not getting it, but… couldn’t there be some sort of reward built in for people who are not only being followed by loads of others, but who are also following back?
No offense to the superstars of Twitterland, but I’d wager that a lot of their accounts aren’t even (wo)manned by them but by aides, reps, or managers. So what if Morrissey tweets a song? He doesn’t follow back anyone. There’s nothing really inherently “social” about that transaction. (yeah I picked one of my heroes for that example to show that I’m not just throwing stones at celebrities)
xoxo,
Tart
July 8, 2009 3:55 pm
Some following back is fine, but if you follow 3000 people and you have 3100 followers, you are likely a spammer, and some of this compensates for that.
July 8, 2009 4:11 pm
Instead of a ratio of followers to following, a better way to monitor spammers might be updates to following times Y. Some one who has significantly more updates than followers certainly is not a spammer and it also counts those that are more “social.”
July 9, 2009 12:58 am
Sorry. updates to those they are following*
July 9, 2009 12:59 am
Yes! finally a legit and awesome way to track music on twitter! You guys should redirect http://hypem.com/twitter to http://hypem.com/#/twitter/popular/1/ .. right now its going to 404 seems like a quick fix 🙂 its 2009 and the less I can type the better haha
July 9, 2009 10:38 pm
Interesting. So if I understand this correctly you only look for Hypem links on Twitter, or do you actually check how many clickthroughs result from them? Because if someone with 1000 followers mentions a track and only one follower actually clicks on that link they really do not carry much influence.
Oh and also, when linking to The Hype Machine on Twitter I’ve noticed there are often multiple entries for one track, even if artist and track name are exactly the same. Do you add those up or could it happen that the same track appears multiple times in the Twitter charts?
July 10, 2009 3:39 am
So, you’ve created a better chart by giving insane amounts of credibility to anyone on the Twitter suggested user list??
As far as I can tell, a single tweet by Ashton Kutcher will put anything at #1.
This is to DECREASE manipulation??
July 11, 2009 12:42 pm
I have done a fair bit of work in this area and generally approach this problem by first defining what it is that you are trying to achieve with such a ranking, rather than just diving into creating a magic formula.
For example, one client is doing something quite similar, but with links in general. They are very concerned about spam and as such I avoided using # of followers or following in my ranking algorithm. Both of these numbers are easily manipulable. I understand that you are currently winsorizing these numbers, but if people do attempt to spam your rankings, you will find yourself in a position of fighting an ever escalating battle by tweaking these limits and consequently increasing your false-positive rate of capping out real user behavior.
A better approach is to use graph theoretic measures that are not simply based on the degree distribution of nodes in the social graph, but rather use measures that encapsulate a users position within the graph – these measures are both much more difficult to manipulate and tend to better align with the underlying motivations of your ranking. For example, applying the Page Rank algorithm to the social graph is one way of evaluating the ‘influence’ of twitter accounts. It is a far better measure of influence than simply followers/following and very difficult to manipulate.
While many of these measures, such as centrality, betweenness, similarity, etc. are computationally intensive to calculate either sampling or approximation based approaches can be used to get much faster results.
If you’d like to learn more, I’d love to work with you guys to improve the great start that you’ve already made.
Cheers,
Josh
July 12, 2009 1:23 pm
The twitter charts have been rick rolled
July 13, 2009 12:09 pm
Your formula is (approx) equal to
Score = twitter_followers^0.5 * (twitter_followers/twitter_friends)
with the ratio constrained to [0.25, 3.0]
July 17, 2009 7:15 am
If you have X followers then your max Score = 3(X^0.5) even if you follow only 1 person.
If someone has 144X followers, his/her min Score = 3(X^0.5) even if s/he follows 25M persons.
July 17, 2009 7:39 am
I would like to see the number of times some one is retweeted taken into account to calculate the user score. It would make more sense.
You can base yourself on Dan Zarella retweetability index to get a start : http://www.retweetability.com/
July 27, 2009 2:09 pm
“Anthony says
Some following back is fine, but if you follow 3000 people and you have 3100 followers, you are likely a spammer, and some of this compensates for that.”
I’m also a little confused because I’ve never seen a spammer cultivate more followers then friends. Usually quite the opposite with a ~ .5 : 100 followers to friends ratio. If you’re actually cultivating a true social campaign with your twitter account, i.e. you in turn follow all your followers except those deemed to be spam. Then I’d expect to see the ratio you mentioned, and would deem this person’s tweets more valuable.
August 8, 2009 1:43 pm
Just FYI, the Hypem blog menu doesn’t contain the Twitter menu item like the main site.
Since you’re blogging about the Twitter chart, it might be handy to add it to the blog menu as well!
August 11, 2009 10:17 am
Do the charts not update in real time? I have been hearting and tweeting tracks I like but they dont got up in the chart rating.
September 15, 2009 6:49 pm
Great! This I dreamt of. Thank you, and I start loading up my songs.
To free the world, to change the world for good.
Madonna, were is your vocale?
Iwanjka_Lunaxme, Luna’y, Iwy
October 13, 2009 10:43 am
The problem with this model is for me that people who have a lot of influence on twitter have a very high influence on the twitter stats regardless to which space they are influential in.
For instance the current number one is basically on top because Arianna Huffington posted a link to it. All the other people that tweeted that link out add very few points and only 14 people tweeted it.
I can’t agree with that rating really. ariannahuff is not an authority on music…
January 28, 2010 10:53 am
myspace=music
twitter=nerds
May 3, 2010 3:40 am
[…] Boyd: The Interactive Twitter Music Chart. Here’s a creative way of ranking tweets by the user’s following to followers […]
May 6, 2010 11:43 am
Boyd;the great twitter music chart. Here they creative way of tweets by the user’s follow to followers…
May 13, 2010 8:38 pm
I think a Music Chart is a great idea. I think trying to work how important how many people follow you or you follow can be tricky. If you are a big artist you might have alof of followers and you would follow less. If you are an unknown artist you may have equal amount of followers/followed because you may have tried to meet people so they can hear your music. When working out the charts re-tweets should also be taken into consideration, but I suppose some people might try to manipulate. I write music and I would love people to listen to my music and I am unknown and I can do it is keep an online prescence and hopefully people will listen to music and come back. Personally I would love to get on the charts. I suppose its okay to dream. Vic
May 29, 2010 7:25 pm
Just seems a bit meaningless really, especially if what people above are saying is true. If one person who is simply followed by a lot of people can tweet once and skew the chart then who cares about it.
Plus what does it mean. I guess i’m a purist at heart and charts should be about sales. The Billboard chart which i believe includes radio play and other non-sale things has always been irrelevant to me because of that.
If you want to make a relevant twitter chart then it should be about tweets alone, or re-tweets in fact (because someone cared enough to pass it on), and one person shouldn’t have more value than another because of the twitter popularity contest, that’s just a bit tragic.
Anyway, as long as you had fun and someone enjoys your list then fair enough, but i looked through it, it didn’t touch me at all and it means nothing, sorry.
September 17, 2010 1:49 pm
Hey,
I have my twitter connected to my hype account but I want to disconnect the two, how do I go about that?
Thanks!
October 5, 2010 8:24 pm
Can you vote for the same song more than once?
October 6, 2010 8:00 am
ye.. it is possible for you to vote for the same song and more than once…? and how does the link is enable …
October 29, 2010 2:18 am
I think this a really great idea!
November 12, 2010 2:03 pm
It is a great idea, but I think it needs some refinement before it will be fully accepted by the hype community. Like a user posted above, this should not be a twitter popularity contest, making it easy to skew the charts. Maybe basing the chart on ‘re-tweets’ would give a fairer perspective.
March 10, 2011 8:53 am
[…] Machine started this chart back in 2009 as a way to better measure what online audiences were into. One of the biggest problems they wanted to address dealt with […]
March 19, 2013 7:26 pm
Great to see this idea in operation – really clever thinking!
May 2, 2013 10:58 am
[…] HypeMachine (an aggregator of music blogs) has recently released its Twitter Music Chart, a music chart which ranks songs based on 1) number of tweets made for each song and 2) who has […]
March 13, 2014 3:24 am