CAN DATA & STATISTICS PREDICT THE EUROVISION SEMI-FINAL QUALIFIERS?

Probably not, but it's worth a shot! Over the past month, I've developed a model for predicting Eurovision qualifiers. In this thread, I will give details on the model, its findings and of course, predictions.

The model is based on 5 variables: running order, past voting patterns, a country's qualifying history, views/streams and odds. While many people have tried to predict qualifiers using one of these variables, I think this model is really interesting because it combines them.

RUNNING ORDER: I calculated the amount of times countries have qualified from a certain position in the semi-final since the producer-determined running order was introduced, giving countries in the same position this percentage as a score.

PAST VOTING PATTERNS: I calculated whether the countries in someone's semi-final are likely to vote for them, to account for trends in voting and potentially political voting. I examined these trends over the past 10 years and calculated the percentage of votes given.

QUALIFYING HISTORY: To distinguish between countries which make an effort or investment in Eurovision and those that are less likely to do so, I calculated a score based on how much a country has qualified since semi-finals were introduced.

VIEWS/STREAMS: To get an idea of a song's (televote) popularity, I divided its views on the Eurovision Youtube and streams on Spotify by the days the song has been out, and ranked the semi-final songs by popularity, awarding them a corresponding score.

ODDS: No model is complete without the odds! I looked at countries' chances of qualifying according to the bookies and awarded them a corresponding score.

Of course, we are on the eve of rehearsals, and much can still change. Therefore, I will update the model (in particular, the views/streams and odds variables) after 1st rehearsals, 2nd rehearsals (adding rehearsal video views after 24 hours) and on the day of the semi-final.

In this way, the model can account for changes brought on by rehearsals and we might also be able to tell at what point in time the model is the most accurate.

Using all these variables and calculating an average score, I determined the 10 countries which are most likely to qualify in both semi-finals using the model. It will be interesting to see how many the model gets right and how accurate it is!

Before I reveal who the model views as qualifiers, two disclaimers are in order:
1) I am not a statistician (I took 1 stats course at university); the methodology is far from perfect and there are probably some inaccuracies in the model. However, I just did this for fun;

I wanted to add some variables which I find important together and see how accurate this might be, as a little test. This is just for my own entertainment and you probably shouldn't take it too seriously.

2) The predictions are just what the model says. Actual results could be completely different. I do not want your fave to NQ even if the model says they will. So please don't come for me, I am just having some fun with the data.

So, without further ado, let's get to the predictions...

For SEMI-FINAL 1:
QUALIFIERS -
NON-QUALIFIERS -

Analysis: In the semi of death, this could very well be the result. One surprising finding is the prediction that Lithuania will NQ: The Roop are favourites this year, but I think what harmed them here was the early release of Discotheque, lowering their views/streams count.

In any case, it would be a shame to see fan favourites Croatia and Ireland, and Belgium (for me) go. An Australia qualification might be considered by some to be quite surprising as well.

FOR SEMI-FINAL 2:
QUALIFIERS -
NON-QUALIFIERS -

Analysis: It would be a shame to see the Czech Republic go, especially after the thousands of tweets saying they will do surprisingly well that I see on my timeline every day. The other NQs are not very surprising.

Despite their qualifying history being stacked against them, I am happy to see San Marino be predicted as a qualifier (even if no Flo Rida ). Denmark might be considered a shock Q by some; but I also feel like locals will eat it up, boosted by its position in the RO.

So, those are the predictions. I am really excited to see to what extent they are accurate and it will be an interesting experiment! As I said, I will update this after the 1st and 2nd rehearsals and on the day of the semi-final.

You can find the data in detail over here: https://drive.google.com/file/d/1Gdi7eXwalZyouzsZAsN0jk02o3qgvU-2/view?usp=sharing

I really enjoyed doing this and I hope you found this thread interesting. Thanks for reading!