Over a fortnight in the Eurovision “bubble”—the press room, which is mostly fan press until the second week—there are all sorts of folks trying to predict results. Who win will? Which pre-Contest favourite will do a Kate Ryan: crash and burn out of its semi-final? Which wee country will do a Rossinelli: finally qualify for the Grand Final, only to finish last on Saturday?
And the question that pre-occupies many of us in during Eurovision week: who will qualify from each semi-final.
In addition to most fan sites have online polling, sites like XTRA do straw polls in the press room itself. And then there is the serviette of death: amongst our Eurokin there is the traditional prediction on the back of a piece of scrap paper: jot down your predictions.
But prediction in this community is a process, rather than a single event. Like many proper Eurovision obsessives, we do a prediction:
- Before the draw performance order is set (March)
- After the draw performance order is announced (April)
- Before the first rehearsals (May)
- Just before each jury final
If that seems excessive to you…you’re in the wrong room!
In the years before I started this site I would use a spread sheet as a sort of predictive model of semi-finalists. As if testing a regression model, I assigned values for song genre, performer age and gender, whether we’d yet seen an actual live performance (and how good it was), whether the delegation had a diaspora or were in a voting bloc (big or small), and where in the draw—because back then it was a random draw—each performed.
The best I ever got via that method was 85 per cent across two semi-finals. I’ve got 9/10 a handful of times; I’ve done as poorly at 6/10.
For quite some time I’ve been ruminating on developing a more streamlined way of predicting Grand Final qualifiers. Anyone who works with data for a living knows that a simple, elegant method is always better than a complicated one—if it works.
I might have cracked this. Read on.
First I aggregated all the data from the two semi-final year. We’ve had two semi-finals every Contest since 2010. This was the first year where semi-final jury and televote scores were released.
However…rather than looking only at the overall rankings I looked at the ordinal rankings for televotes and jury scores too, hopeful of some patterns emerging. Why rankings rather than scores? Because 2013.
From the 2013 Eurovision split jury and televote scores haven’t ever been released. The EBU only released ordinal rankings for each component and “average rankings” within each component. For all the other years we have both the ordinal rankings and the split scores. Therefore we can go based on scores, as its not “like for like” *shakesfist at the EBU*
But bear something in mind: across all these years the relationship between jury scores, televote scores, and overal scores and rankings has varied. In some years these two components were combined/integrated (based on ordinal rankings, either of top 10 or a complete ordinal ranking), aggregated scores (added up, since 2016), or used to shift the final rankings for any jury ranked top 10 semi-finalist that wasn’t in a semi-final top 10 (in 2008/2009). But we have ordinal rankings across all the years between 2010 and 2018.
It turns out that ordinal rankings are good enough. Whew!
The complete data are found here. Before I reveal my proposed method, here’s an interesting bit o’ trivia…
It turns out that all sorts of countries have “won” a Eurovision semi-final in some way. We have the official winners each year, but we can also see who’s “won”—finished first—in either the televote or jury vote in a given semi-final.
Here’s the list:
Twenty one countries have topped at least one semi-final ranking between 2010 and 2018. If we included the earlier double sem-final years we could add Iceland to the list.
That’s roughly half the number of delegations having ever competed over those years!
But let’s get back to predicting semi-final qualification: is there any relationship between the televote and/or jury rankings as a predictor of Grand Final qualification?
It appears there is!
What emerges quickly is that entries ranked in the top 5 of either the jury or televote rankings of their semi-final qualifies for that year’s Grand Final.
In fact, there have only been one exception. In 2013 Montenegro’s epic Igranka finished fourth in the televote of semi-final one, but Who See ended up 12th overall.
Except…in 2013 the jury and televote rankings were combined to produce an official ranking. In that system, a low ranking in the other component could effectively erase a strong ranking in the other—as it did to Montenegro.
Da-da-da-da Igranku (Source: YouTube/Eurovision Song Contest)
If we were to apply today’s system—which adds the two scores to produce an overall ranking based on combined points (not combined rankings)—here’s the revised scoreboard for 2013 semi-final one:
Oh wait, right. We can’t. Because 2013 is the only year the EBU’s not released the actual split scores—only the ordinal rankings. *scowl*
In every other year, any entry that was ranked in a televote or jury top 5 qualified out of its semi-final.
Based on reports on the ground, I would surmise the following top 5 in each semi-final. These are not ordinal rankings—just the top 5. Though I have put my guess at who will win each component in bold.
- SF1 Jury: Slovenia, Serbia, Iceland, Greece, Portugal
- SF1 Televote: Greece, Australia, Serbia, Cyprus, Czech Republic
- SF2 Jury: Azerbaijan, Russia, Netherlands, Austria, Sweden
- SF2 Televote: Russia, Switzerland, Sweden, Malta, North Macedonia
That produces two partial lists of qualifiers:
- SF1: Slovenia, Serbia, Iceland, Greece, Portugal, Australia, Cyprus, Czech Republic (8 of 10 qualifiers)
- SF2: Azerbaijan, Russia, Netherlands, Austria, Sweden, Switzerland, Malta, North Macedonia (8 of 10 qualifiers)
Which leaves two more qualifiers from each semi-final. Check back with me after the Grand Final.