GMAP Predicts the 2024 Eurovision Winner

GMAP Predicts the 2024 Eurovision Winner

As experts in geographical modelling, wouldn’t it be possible to predict who will win the Eurovision Song Contest? We first developed a model to do just this many years ago and used it to predict Ukraine’s first ever win in 2004. Since then its success has been more limited, as we frequently failed to account for the Je ne sais quoi that makes a Eurovision winner.

Now, 20 years later, it’s time to see if we can build a model to predict who will win this year.

The Eurovision Model

We will start this year with a simple model that looks for geographical biases in our scoring preferences. To do this, we have ranked every entry for the past 10 years. These historical rankings were actually carried out each year in advance of the Contest, so there has been no adjustment for hindsight. Then, when compared to the actual results, the model builds index scores to show which countries we have been biased towards or against. The final step is to apply these index scores to our rankings for 2024. This results in our prediction of who will qualify from the Semi-Finals, and then the scores for the Grand Final on 11 May.

Anomalies in the Data

A good model is only as good as the data that goes into it. Anomalies in the data need to be smoothed out, otherwise the results will be based on distorted data. Anything we got wrong in our judging would otherwise be amplified in future years. Thus, after some of us failed to predict the win for Portugal in 2017, the index score for Portugal rocketed, meaning every low score for Portugal would get translated into a high final result after that. So, to minimise the effect of Portugal being the predicted winner each year, we smooth out the distortions using as many years’ data as possible. We have gone back as far as 2016 – prior to this, the final results weren’t split between the Jury and Public votes, so it seems a good place to draw the line. 

First Semi Final

If you didn’t know, 37 countries compete this year in two Semi-Finals followed by a Grand Final. The top 10 from each Semi-Final qualify, plus the Big Five (France, Germany, Italy, Spain, UK) and the host country (Sweden). Spoiler alert: these are GMAP’s predictions for the first Semi-Final. The top 10 qualify, the next five don’t.

The model predictions are generally in line with our scoring this year, except some such as Cyprus and Moldova have jumped up the rankings. This is because we missed the appeal of Cyprus with Fuego in 2018 and Moldova with the saxophone in 2017. Obvious how we got it wrong when you see them live on the night; they did much better than we predicted. We’ve smoothed out the anomalies, but these are still countries we consistently underestimate.

What a shame for poor Ireland. Some of us had Ireland as their favourite out of the entire list of 37 countries. The model has done its job, taken our geographical bias into account, and listed it as a non-qualifier. It’s borderline however, so fingers crossed they do get through on the night.

Note: this is the first year Luxembourg has entered since 1993. Index scores for Luxembourg have been modelled by using an average of the rest of the scoring for all other countries. We’re not basing it on the fabulousness of France Gall.

Second Semi Final

This one contains the overall winner, and it’s a landslide victory in terms of the winner of this Semi-Final. It’s borderline around the qualification mark, so there may be hope for the likes of Austria yet.

Figure 2: Second Semi Final

The Grand Final

According to the model, Switzerland wins for the first time since 1988. Back then it was a two-horse race with just one point in it against the UK. The closeness in scores was a bit embarrassing, given that Switzerland had a 20-year old Céline Dion (dubious outfit, but we’ve all been there). This year, we predict Switzerland as a clear winner, no doubt about it.

Finland, the Netherlands and Croatia are comedy entries. The Finnish one is actually laugh-out-loud funny most of the way through. And the Croatian one tells of the brain drain from his poor country to the rich West. But comedy ones don’t generally win, whether they do an homage to 2-Unlimited or not (if you’re reading this, you can get down off the table now, Jibs), they just do better than expected. That might mean these come a bit higher than we’ve predicted.

The twins from Sweden are interesting here. If we assume Switzerland will win (the bookies also think so, and they are rarely wrong these days), then the one we are most at odds with the bookies with is Sweden. Our group scores it second. We normally score Sweden high, so the model has pushed it down into fifth place. The bookies have Sweden in 15^th place, so this is one worth watching. Perhaps we have got this completely wrong (no surprise based on past experience) or the host country might have got it right again this year.

Figure 3: Grand Final

Conclusion