2016 forecast: pre-contest
Image source: Andres Putting (EBU)

2016 forecast: pre-contest

What’s new in 2016?

Another year, another contest. We’ve got the usual revolving door of countries participating: Portugal are out because their selection procedure was shambolic, and Romania have been somewhat controversially disqualified because they didn’t pay their bill from last year.

On the other hand, big hitters Ukraine are back after missing last year’s contest due to the ongoing conflict in the east of the country. They’ll be performing a totally non-political number about the Soviet invasion of Crimea in 1944 (sample lyrics: “When strangers are coming / They come to your house / They kill you all and say / We’re not guilty, not guilty”). Bosnia-Herzegovina, Croatia and Bulgaria also return, although their absences were more financially driven.

After last year’s successful pilot program, Australia have been adopted as a permanent participant. There’s no automatic qualification this year though—they’ll be competing in the second semi-final along with everybody else.

The big change is that the voting system has been changed, in a way which arguably makes it both more transparent and more confusing at the same time. If you really want to know the gory details of the voting system, then pour yourself a stiff drink and read on. If not, please feel free to skip the next section, and join up again when you see the YouTube videos start to appear.

A brief history of Eurovision voting systems

The systems used for Eurovision voting over the years might charitably be described as baroque. For the first contest, in 1956, each country sent two jury members to Lugano, except for Luxembourg, who let the Swiss vote on their behalf. There was a secret ballot, and Switzerland emerged victorious. This voting system was not re-used in subsequent years.

From 1957 to 1966, there was a sequence of slightly different voting systems, most of which relied on a jury ranking their top three, five or ten songs, and awarding points thusly. After that, there were a few years where each jury member (ten per country) could give one point to their favourite song. This led to a four-way tie in 1969, so they went back down to two jury members per country (one young, one old), who could rank songs from one to five. Then for some reason, they went back to the previous system for one year in 1974.

In 1975, something approaching the modern system was introduced. Each jury member ranked their top five (or in later years, ten) songs, and these rankings were added together to produce a top ten from each country, which were awarded points ranging from 1 to 8, then 10 for second and 12 for first.

This system somehow stayed in place unchanged for twenty years. In 1997, five countries trialed a televoting system, and from 1998 onwards most countries used a televote to rank the songs, except in a few cases where technical issues or low vote numbers led to the use of a backup jury.

After the 2007 contest (won by Serbia), there were some complaints from Western European countries that the voting had become “political”, and in particular that the countries of the former Yugoslavia and former Soviet Union were passing out points among themselves, rather than voting for said Western European countries. The fact that the Scandinavians had been doing this for decades was conveniently ignored.

As a result, starting in 2009, a hybrid system was used, in which jury votes and televotes were combined to produce a single ranking. The system used for this combination has changed from year to year, but most recently it consisted of each jury member giving a full ranking from e.g. 1 to 25, which were then added together and reranked to give a combined jury rank, which was then added to the televote rank to give a total score, which was then reranked and translated into the scores 1-8, 10, 12.

This year, however, somebody noticed that this system isn’t very transparent, so there’s a new thing.

How does it work now?

So for this year, there’s going to be twice as much voting. Each country will have a jury score on the 1-8, 10, 12 scale, decided in the same way as the combined jury rank above. Each country will also have a televote score on the same scale, decided based on the number of televotes.

In order to save time during the scoring phase, only the twelve points from each country’s jury will be announced as we go along, with the other points magically appearing on the scoreboard. Then the total televote scores for countries which placed outside the top ten in the overall televote will be added to the scoreboard without fanfare, and finally the total televote scores from the remaining countries will be announced.

I have multiple higher degrees in mathematics, but if the EBU expect me to follow this, they are drastically underestimating how much I drink during a Eurovision broadcast.

Model changes

Once again I’ve stripped out last year’s changes, because they weren’t very effective. We’re back at the basic model, with four main terms:

  • How much the voting country “likes” the performing country
  • How strong a typical song from the performing country is
  • How strong an individual song is
  • Random noise

To this I’ve added an extra term, which indicates how “populist” a song is. A song with a positive populism score will do well in the televote, but poorly with juries, and a song with a negative populism score will do the reverse. Thankfully, the EBU have provided full split jury/televote results for the past two years, so we can estimate how strong this effect is.

A strongly populist song (populism 1.3) looks something like this:

Meanwhile a very un-populist song (populism -1.3) looks like this:

In general, populist songs seem to feature groups rather than individuals, high tempos, catchy hooks, sexy butter churning, etc. Un-populist songs tend to be slow, powerful ballads, almost without exception performed by lone women.

The effect of the populism variable on an individual vote is reasonably large. The swing from televote to jury can be up to 2.6 points, the equivalent of the relationship between Latvia and Lithuania, or a little more than the typical difference between entries from San Marino and Russia.

However, this swings both ways, so a song with a high populism score will do well from the televote, and poorly with the juries, and it might just all average out. On the other hand, it’s possible that being polarising is a good strategy. If you’re unlikely to break the juries’ top ten, then it’s probably better to chase after televotes. I don’t think there’s anybody trying this strategy this year, but it will be interesting to watch how things develop if this voting system is kept.

Final predictions

Which is to say, predictions for the final, not predictions which are, themselves, final. There will be more predictions, some of which will also be about the final.

Because of the change in voting systems, I’ve run all the simulations twice—once with the old system and once with the new. To be honest, the results are almost identical:

Winning probabilities for 2016

Essentially there are two things going on with the new voting system, which tend to cancel each other out. On the one hand, having scores from two different groups tends to introduce more variety, and makes things less certain, increasing the chances of weaker countries, and decreasing the chances of stronger countries. On the other hand, having twice as many votes going on tends to reduce the amount of variation–the benefit of a single twelve in a sea of zeroes is effectively halved.

Anyway, the big story here is obviously Australia. They did pretty well last time on their first outing, coming in a quite respectable fifth. They’re obviously good, but we don’t yet know quite how good. The model is therefore a little bit uncertain about them, but they do now have probably the best average performance in the contest. It’s entirely possible that they’re going to be powerhouses, dominating the contest for years to come, but it’s also possible that last year was a fluke, and they’ll sink into ignominy from here out.

Stepping outside the model for a minute though, I think there are good reasons to be bullish on Australia. They’re well-resourced, with a strong domestic music industry which doesn’t regard Eurovision as a joke in the way that some countries do (*cough*UK*cough*). Last year they had some trouble getting votes in the former communist bloc, but they more than made up for it with a strong showing from the west. I’ve put a tenner on at twenty to one odds.

The bookies’ favourites are currently Russia, with Ukraine and France in second and third. Russia and Ukraine are definitely strongly ranked by the model, but France…aren’t. If you’re having trouble reading the tiny bar on the graph above, then I’ll let you know that the model has them with a 0.12% chance of winning, or odds of around 800-1. This might seem a bit harsh, but they’ve only made the top ten once since 2003, and failed to break double figures in points on their last two outings. On the other hand, that was with the “moustache song”, so who knows?

Semi-final 1 predictions

Qualification probabilities for 2016 semifinal 1

The effects of the new voting system are maybe a little bit more obvious in the qualification stages, although they’re still small. The new system gives a little boost to the weaker countries, but it’s not anything to get excited about.

This is probably the weaker of the two semi-finals on average–Russia are the only serious contender here. The gamblers have it as stronger, but that’s largely down to Russia’s position as the favourite.

Of the top five countries here, only Armenia has ever failed to qualify for the final, and even then only once, and by one point. I think it’s fairly safe to predict that Greece (96%), Russia (92%), Azerbaijan (91%), Armenia (90%) and Bosnia-Herzegovina (89%) will all qualify. There’s a bit of a Scandinavian clique in the voting here (Sweden as hosts also vote in this semi-final), so Iceland (65%) and Finland (55%) are also good bets. To fill out the ten, we have Hungary (60%), Croatia (59%) and Moldova (51%). Malta (48%) are unpredictable as always, so they’re probably a good wildcard bet if any of the others fail to get in.