Eurovision 2015: Can YouTube data indicate who will win?

Using YouTube data collated on Saturday 18 April 2015, is it possible to predict the winner of the 2015 Eurovision Song Contest? 

I don’t generally do maths. Or at least, I don’t especially enjoy it. And while Microsoft Excel may well take some of the pain out of doing calculations, it still remains the case that for the most part data analysis doesn’t normally get me excited.

But, a conversation with an old friend who is struggling to understand the YouTube success of the UK’s song compared to his personal favourites – Heroes (Sweden) and Rhythm Inside (Belgium) prompted me to think how all the songs are performing on YouTube in terms of statistics. Was there a runaway winner I wasn’t aware of? Would we be able to have a vague idea idea of who the winner would be in May by looking at the YouTube stats today?

There then started a two and a half hour session of number-crunching of the kind I’ve never embarked upon before. Not only is what follows unprecedented for me, the fact I enjoyed the process is also rather odd. None of what follows is especially scientific – I fully expect that if you held my workings up to the light it would all disintegrate. (If it does, please endeavour to point this out privately to me, if you’d be so kind.)

The analysis

You can see the raw data I collated in this PDF.

The initial method seemed fairly straightforward: compare YouTube views of all of the competing songs and see who comes out on top. Why YouTube views now? A couple of reasons. The first is that this YouTube data is the most readily available and by using that from Eurovision.TV’s YouTube account, everyone’s on a (sort of) level playing field. Second, the weekend before the official Eurovision album comes out, for a lot of fans, watching the songs on YouTube is the quickest (and cheapest) way of hearing all the songs in the contest.

Belgium's Loic Nottet awaits the conclusions of my statistics-based Eurovision odyssey.
Belgium’s Loic Nottet awaits the conclusions of my statistics-based Eurovision odyssey.

I grouped the songs into Semi-Final One and Semi-Final Two songs based on YouTube playlists Eurovision.TV have produced. Total page views for Semi-Final One were 11,579,862; Semi-Final Two was approximately 49% down at 5,699,655 page views.  The top most performing song in Semi-Final One was Russia’s A Million Voices with 2,972,153 page views; in Semi-Final Two was Azerbaijan‘s song 889,560 views.

The songs which have an automatic place in the final (Australia, Austria, France, Spain, Germany and the UK) had been viewed in total 4, 444,530 times, comparing well with the second semi-final. The UK’s song Still In Love With You had been viewed 1,471,081 times.

Worth noting that the data was collated on Saturday 18 April 2015.

Pulling in Extra Data

As I was going through the songs to get total views, I quickly realised that there were significant fluctuations in the number of likes. In terms of user behaviour, was there a need to capture the number of times people had clicked on a like button on YouTube (and the dislike button). Did a like/dislike count as a real expression of whether a song had registered well or not, as opposed to views which could – potentially – be regarded as an ambiguous measure?

Thinking further, was it necessary to subtract the dislikes from the likes to get a more accurate measure? And then, what about likes/dislikes as a percentage of the total views?

That’s when I realised things had spiralled out of control and the brakes needed to be applied.

Using YouTube data to extrapolate the Eurovision 2015 qualifiers

Next, I drew up a list of countries for each semi-final and calculate rankings based on four categories:

1. YouTube views
2. YouTube Likes
3. Likes adjusted by Dislikes (subtract dislikes from total likes)
4. Adjusted Likes as a percentage of views.

For each category I came up with a ranking (largest to smallest) of countries for each category. Every country in every category got a point relating to where in the ranking they were. The top country – 27 points; the last – 1 point. I then totalled up each countries points and carried forward the top ten best countries on to the final.

Semi-Final One Semi-Final Two
Russia 10, 10, 9, 4: 33 Azerbaijan 10,10, 10, 10: 40
Albania 9, 9, 10, 9, 7: 44 Israel 9,9, 9, 5: 32
Armenia 8, 8, 7, 0: 23 Iceland 8,8, 8, 2: 26
Georgia 7, 7, 8, 10: 32 Montenegro 7,7,4,0: 18
Belgium 6, 6, 4, 2: 18 Malta 6,6,7,0: 19
Estonia 5, 5, 3, 0: 8 Norway 5, 5,5, 4 : 19
Belarus 4, 4, 5, 5: 18 Slovenia 4,4,6, 9: 23
FYR Macedonia 3, 3, 6, 6: 18 Lithuania 3, 3,0,0:6
Serbia 2, 2, 0,0: 4 Czech Republic 2,2, 2,0: 6
Moldova 1, 1,0,0: 2 Latvia 1, 1, 1, 3: 6
Romania 0, 0, 2, 8: 10 Sweden 0,0, 3, 8: 11
Netherlands 0, 0, 1, 3: 4 Poland 0,0,0,7: 7
Hungary 0,0,0,1: 1 Ireland 0,0,0,6: 6
Qualifyers ranked in order with (points):
Albania (44) Azerbaijan (40)
Russia (33) Israel (32)
Georgia (32) Iceland (26)
Armenia (23) Slovenia (23)
Belgium (18) Norway (19)
Belarus (18) Malta (19)
FYR Macedonia (18) Montenegro (18)
Romania (10) Sweden (11)
Estonia (8) Poland (7)
Serbia (4) Lithuania (6)

A surprise emerges from this process. Sweden doesn’t appear as the runaway winner (nor Belgium for that matter) I expected it to be. My assumption had been that even being uploaded to YouTube later than the other competing countries, that Sweden would be at the top of their semi-final.

Sweden's Mans Zelmerlow singing Heroes with his animated pal at Melodifestivalen 2015
Sweden’s Mans Zelmerlow singing Heroes with his animated pal at Melodifestivalen 2015

Part of this is down to the low numbers of views for Sweden’s song Heroes. It’s a cracking number which has garnered a lot of enthusiasm amongst Eurovision fans, in part because of its rousing anthemic qualities and also because of the clever stage presentation used in Sweden’s selection show MelodiFestivalen. The preview video on Eurovision.TV plays the whole thing incredibly straight in comparison just featuring singer and lyrics. I’m reliably informed by a pal that concerns over plagiarism have resulted in an amended version being uploaded to after there claims made that the animations used in Melodifestivalen broke copyright. A comparable figure of over a million views of ESCSweden15’s Melodifestivalen upload could be counted here (see my thoughts at the bottom of this post).

The Final

I applied the same process to the competing countries in the final, combining their figures with the qualifiers from semi-final one and two. Rankings were drawn up and points assigned (27 to 1) according to how countries ‘performed’ in four categories. For the final I incorporated an additional category: adjusted likes as a percentage of total views. I’m struggling to work out exactly why at this stage (its been a long process), but I was interested in seeing what the results might be.

As a reminder, the five categories are:

1. YouTube views
2. YouTube Likes
3. Likes adjusted by Dislikes (subtract dislikes from total likes)
4. Likes as a percentage of YouTube views, and
5. Adjusted Likes (see point 4) as a percentage of views.

So, here they are. Countries ranked according to their total points assigned. First for Albania, twelfth for Belgium, sixteenth for the UK and twentieth for Sweden.

Rank Country Total
1 Albania 129
2 Georgia 126
3 Armenia 121
4 Azerbaijan 119
5 Russia 111
6 Italy 100
7 FYR Macedonia 95
8 Israel 92
9 Spain 92
10 Belarus 82
11 Slovenia 78
12 Belgium 73
13 Iceland 65
14 Norway 62
15 Australia 61
16 UK 59
17 Estonia 57
18 Malta 55
19 Montenegro 51
20 Sweden 49
21 Latvia 43
22 Germany 37
23 Austria 34
24 Lithuania 26
25 Serbia 24
26- Moldova 15
26- France 15

Some thoughts arising from this Heath Robinson approach to picking out a Eurovision winner. They’re noted below.

1. Some countries had selection programmes on TV; others revealed their songs online. Is there a correlation to be drawn between numbers of views on YouTube and whether or not there was a TV programme to decide on the song? If there was a TV show, would that result in fewer YouTube views?

2. There seems to be a lot of clustering in the top four countries in the final ranking. I’m not entirely sure why this is.

3. Are record comapnies buying YouTube views and likes? It’s possible to do. Is it being done here? Does it matter?

4. Had Sweden been able to keep their innovative TV presentation from Melodifestivalen up on Eurovision.TV as a preview video, would we see Sweden higher up the table? The ESCSweden15 account posted Mans’ Melodifestivalen performance, a video which has totted up over a million views.

5. Is there a correlation to be drawn between when a video was uploaded and the total number of views. That is, if a video is uploaded on March 12 for example, it is likely to get more views over a video uploaded on April 7, by virtue of the former being available for longer.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.