I mean, it didn’t pick the winner or anything, and it was dead-wrong about Ireland, but it allowed me to correctly judge that Italy was going to do the best out of the Big Five countries (which means I very nearly broke even, thank you very much), and it also allowed me to identify that Azerbaijan were going to do considerably better than most experts seemed to think.
The votes award 1-8 points to a country’s 10th to 3rd favourites respectively, 10 points to the second place, and 12 to the first place, which tends to give a runaway victory, so for Denmark to win with 281 points and Azerbaijan to come second with 234 (with Ukraine, another of my faves, third with 214) is really bloody good. If I’d bet on them to place, rather than to win, I’d have come away richer (slightly, because it was only a £5 bet after all).
Conclusions: The software, as it currently works, is never going to reliably pick the winner, even with proper weighting. The survey is too heavily biased away from Eastern Europe, so it’s not taking the temperature over there effectively enough, and that’s partly due to Western Europe being a far greater presence on Twitter. Nothing I can easily do about that. Also, Ireland. Not sure what the fuck happened there, but I suspect I need to introduce some semantic interpretation to figure out if people are being positive or negative. Even so, just on the strength of the song, I’m baffled that Ireland did so badly. It’s like everyone just sort of forgot about them. A weird result by anyone’s standards, and one to keep in mind.
Even with these problems, the software was still a bloody incredible tool for assisting in making an informed decision. It showed me that after the second semi Azerbaijan was suddenly getting a lot of attention. Bearing in mind that they could rely on the Eastern Bloc to vote their way, this showed me that over here in the West we were gonna give them votes too. Certainly not a foolproof method, but now I know how it works, and it gave me one hell of an edge over everyone else, especially considering that this was just a silly little trial-run.
I’m going to fucking town with this, boys and girls. This changes everything.
Well, my software has been scanning the Twitter chatter about the Eurovision entries every five minutes (timing problems permitting) for the past week. It’s half past eight on the morning of the live final, and I have collected over 1850 sets of results. I want you to picture a spreadsheet of data, 41 columns wide, and getting close now to two thousand rows long. That is what I have made, and it’s all based entirely on what people are talking about on Twitter.
As far as I’m aware, this is the single most detailed study of Eurovision Twitter buzz ever conducted (though if you know of another, I’d be seriously fascinated).
Unfortunately, since there’s more people here in western Europe, my data is biased towards the Big Five countries (UK, France, Spain, Germany, Italy - all share the majority of the cost for putting on the contest), and since I’ve never run a study like this before I have no baseline to work out weighting. It’s all intuition, experience, and best-guesses who might win, but here’s what I reckon:
Pretty much all of the former USSR countries qualified, so their vote will be split and diluted. Still, Mother Russia has a solid song and is tracking well. Probably not a winner, but a respectable placing is on the cards.
The Ukraine has a giant. It’s an ingenious ploy, because everyone’s talking about the giant. That counts for a lot. The song isn’t half bad, and the singer is a pretty lady. They’ll also do well.
None of the former Yugoslavian countries made it through to the final. Bit of a shocker, that. They normally vote for each other, but now those votes are in the wind. Still, expect them to favour the former Soviets, and I suspect they’ll quite like Ireland and The Netherlands.
Everyone in Malmö is talking Denmark for a win, but it looks to me like Finland’s gambit of having a kind-of demeaning post-feminist bit of pretend lesbianism is giving them some last-minute traction. When you’re choosing who to vote for from so many countries, memorability counts for a lot, and a girl-on-girl snog (awkward and fake as it may be) is a pretty good way to get everyone to remember you. Suspect they’ll finish above Denmark in the end.
Romania has a falsetto-singing operatic vampire with a giant parachute skirt from which emerge dancers. If you are from America and have never watched Eurovision, this sort of shit is pretty much standard. I like them. Bit of a dark horse, as are The Netherlands. I don’t much like the Dutch song, but they’re getting people talking, and that’s what counts.
Finally, who’s gonna win? I’ve weighed up the options, looked closely at all my data, and finally decided who to put money on. I make no guarantee about the quality of my tip, but I reckon Azerbaijan have a good song, clever staging, and the weight of their neighbours behind them, plus they’ve been getting a sudden boost in attention since the rest of us saw them in the semi on Thursday. I’ve got a fiver on them to win, at 12:1. I’ve also put another fiver on Italy to get the top score out of the Big Five, at 1.83:1. I’m confident about the latter, less-so about the former. Ukraine and their giant could easily steal it out from under Azerbaijan, or maybe the bookies are right to make Denmark the favourite. But that’s where I’ve put my money.
We’ll know later tonight, but either way it’s been fun getting this software working.
I’m gonna point it at the markets next, and see how that works out.
So, after this morning’s one hour test, I ran the code I wrote for six hours this evening, and it’s already gleaning pleasingly detailed information about the relative buzz around each country in next weekend’s Eurovision*.
Already, this looks like typical scoring for the contest, though I wouldn’t bet money on it yet (need more data), but it proves the core functionality works. Still need to make it more robust, in case my connection goes out or my spare computer crashes, but this is looking nice so far.
*If you’re not from one of the countries where Eurovision airs, it’s basically a giant singing contest, with hugely partisan voting, and several countries that stretch the meaning of the word “Europe” that most of the continent views as a great opportunity for international exposure, and Britain views as a great opportunity to get drunk and laugh at our European neighbours for taking this shit so seriously.
Probably not the way most people think about Big Data
And then I got thinking earlier on today - could I grab Eurovision-mentioning tweets automatically, to then run through a bit of parsing and weighting, and see if I can generate accurate predictions about who is going to win? I mean, sample how much each act is being tweeted about, correct for country bias using locative data, eliminate people talking about their own act (since countries can’t vote for themselves), and you have a nice little data project that’s almost certainly too frivolous and silly for serious data-heads, but is right up my street and would give me a nice fun bit of practice. But could I do it?
Well, yes, yes I think I’ll be able to. I can already grab tweets and turn them into something like this:
@Nahumsitooo | Wed, 01 May 2013 23:09:36 +0000
@Pitu_Noe espaÑa? Jaja esta bonita la cancion, pero no para #Eurovision
@Eurovision_Rt | Wed, 01 May 2013 23:09:24 +0000
RT @NerdyHBIC: OMFG crying as my sister goes into detail as to why Tooji was a poor man's Eric Saade in last year's Eurovision.
@viewastar | Wed, 01 May 2013 23:09:19 +0000
Watch Eurovision 2013 Singing Italy Marco Mengoni L'Essenziale http://t.co/krr3rxUn0f
That’s the current output, but of course I can grab way more in a single go, and filtering through it won’t be too tough. Leave it running in the background, creating a tally to graph, awwww yeah, this is gonna be funtimes. I’ll keep you informed.
Yup, still doing these. The strange thing is, no one else seems to have had the same idea. Certainly, searching YouTube for other unofficial commentaries is not returning much, except Nebraskan teens’ boring exposition of whatever run-around-and-shoot-strangers-on-xbox game they’re all playing this week.
Think about that for a second - with three dumbass unofficial commentaries under my belt, it would seem that I’m now the world leader in making unofficial commentary tracks for movies.
Not that anyone’s watching them, but I figure I’ll put out a long tail and see what happens.
I have started a New Thing. I’m calling it Bad Movie Commentary. Basically, I do my own drunk, rambling, dumbass audio commentary to a movie. Specifically, a movie that is not very good. So naturally my first choice was Twilight.
As I write this, the video has been up about sixteen hours, and even though I have bugger-all promotional pull and my SEO skills are a bit rusty to say the least, the thing’s been watched 115 times already.
I think if you add together the total number of views that all seven episodes of my earlier cartoon, Shapeborough, ever managed to get, it’s barely more than that.
I also have three thumbs down, and that perfect, priceless comment from EllaaCollinss which couldn’t be better as first comments go. Yes, I am being an asshole. Yes, I am a horrible horrible man and should know better. But this is brilliant and despite knowing I’m irredeemably Wrong, I’m so very very excited!
Due to time constraints it may not be possible to respond to your email. All messages are read, printed off, then eaten. Om nom nom.
® iTunes and the iTunes logo are registered trademarks of iTunes, Inc. ®
7digital and the 7digital logo are registered trademarks of 7digital, Inc. ® Bandcamp
and the Bandcamp logo are registered trademarks of Bandcamp, Inc. ® MySpace and the
MySpace logo are registered trademarks of MySpace, Inc. ® Twitter and the Twitter
logo are registered trademarks of Twitter, Inc. ® Soundcloud and the Soundcloud logo are registered trademarks of Soundcloud, Inc. ® FriendFeed and the FriendFeed logo
are registered trademarks of FriendFeed, Inc. ® YouTube and the YouTube logo are
registered trademarks of YouTube, Inc. ® Facebook and the Facebook logo are registered
trademarks of Facebook, Inc. ® Bebo and the Bebo logo are registered trademarks of