Background
The ones who know me in person probably know about my background in Math, Statistics and Data models.
As part of ESCBEAT time, I have a great passion for Eurovision.
So the natural question would be: Why not combine both?
Model – Main Ideas
What does the model include? I have included several different variables:
- Digital Platforms Data – How popular are the songs on the 2 main digital platforms:
- YouTube – sum of views of the videos. We consider the videos on the official Eurovision Song Contest YouTube channel, as well as the ones on the artist’s channel and videos on the national broadcaster channel. If an act is chosen by a national selection, the video from the selection will be included. In the case of an act that was chosen internally, we would take one live performance into consideration. Pre-parties videos are not included.
All views have been updated to Saturday, April 23rd.
- Spotify – number of streams (updated to Saturday, April 23rd)
- YouTube – sum of views of the videos. We consider the videos on the official Eurovision Song Contest YouTube channel, as well as the ones on the artist’s channel and videos on the national broadcaster channel. If an act is chosen by a national selection, the video from the selection will be included. In the case of an act that was chosen internally, we would take one live performance into consideration. Pre-parties videos are not included.
- Betting Odds – We consider the average position for each country in the semi-final betting odds. The average is calculated on the positions between April 1st and April 23rd.
- Qualifying Rate For Country – The Eurovision semi-finals were introduced since 2004 (one semi-final in 2004 – 2007, and 2 semi-finals since 2008). We calculated the percentages of qualifying for the final for each country. For example, if a country has qualified 6 times out of 12 times, the value will be 50% = 0.5. We would like to remind you that during the semi-finals of Eurovision 2008 – 2009, the jury had the option to choose a 10th qualifier which wasn’t necessarily in the 10th place. We have taken this under consideration
- Average Rate for Country – We calculated the average positions that each country achieved in the history of the semi-finals. For example, Australia has participated in 5 semi-finals (2016 – 2019, 2021). The places they achieved in the semi-finals were 1, 6, 4, 1, 14 and therefore the average will be = (1+6+4+1+14)/5 = 5.2.
- Qualifying Rate for Position – We calculated the probability of qualifying from each spot in the semi-finals. For example, we have had 30 semi-finals events so far. In 13 cases, the country that performed in the 5th position has qualified for the final. Therefore, the probability of qualifying from the 5th position is 13/30 = 43.33%. We would like to remind you that during the semi-finals of Eurovision 2008 – 2009, the jury had the option to choose a 10th qualifier which wasn’t necessarily in the 10th place. We have taken this under consideration
- Average Rate for Position – We calculate the average place in the semi-finals for each optional spot. The most successful spot in the semi-finals is the 14th position.
- Voting Patterns – Based on the results of each one of the shows since 1975, we built a list of the top 10 countries that tend to vote for each country. Insights are based on each vote: semi-final, final, televote and jury voting. Thereafter, for each country, we counted the number of “friendly countries” that will vote in the same semi-final.
Together they permit us to have a wide point of view of the contest
The Method
Based on preliminary research, I’ve given a weight (0-100) to each one of the 8 suggested variables. The sum of the weights together is 100.
| Variable (Data Feature) | Weight (percentage in final score) |
| YouTube Views | 16.666% |
| Spotify Streams | 16.666% |
| Semi-Final Betting Odds | 33.333% |
| Qualifying Rate Per Country | 6.666% |
| Average Rate Per Country | 6.666% |
| Qualifying Rate Per Position | 6.666% |
| Average Rate Per Position | 6.666% |
| Number of “Friendly Countries” voting in the same semi-final | 6.666% |
We have ranked the obtained values in each variable and scored the country based on a uniform range. This means that if we have 17 countries particiapting in semi-final 1, the one with the highest value in a specific feature will get the score of 17, while the lowest one will get 1 point only.
Thereafter, we multiplied the score with the weight as explanied in the table above.
Semi-Final 1 Analysis
As a reminder, this is the running order of the first semi-final:
Remember: The highest score you get, the highest your chances to qualify.
The table is too wide, so we’ve split it into 2 images:
Example of a score: Let’s see together how Albania has scored 1210 points which puts them in 4th place:
- YouTube Views = 16.66%: Albania has achieved around 7.8 Million views. The only country to defeat them in this semi-final was Ukraine. Therefore, Albania gets 16 points (as the leading country gets 17 points). So the first component of the score is 16.66 * 16 = 266.666 points
- Spotify Streams = 16.66%: Albania has around 2.2 million Spotify streams. This puts them in 5th place (after The Netherlands, Norway, Ukraine and Austria). Therefore, 5th place means 13 points (as the first places get 17, 16, 15, 14 points). So it adds them 16.66 * 13 = 216.666 points
- Semi-Final Betting Odds = 33.33%: The average place of Albania in the semi-final betting odds is 5th. Therefore they will get a score of 13 points. Let’s add 13.33 * 13 = 433.333 points
- Qualifying Rate Per Country = 6.66%: Albania has participated in 16 semi-final shows to date. They have qualified in 9 times out of 16. So their qualifying probability is 56.25%. This puts them (together with Lithuania and Iceland) in the 9th place regarding this variable. So let’s add another 6.66 * 9 = 60 points to the final score.
- Average Rate Per Country = 6.66%: If we calculate the average place of Albania in the semi-finals, we get 10.63 which puts them in the 10th place which equals to 8 points. Therefore, it is time to add 6.66 * 8 = 53.333 points.
- Qualifying Rate Per Position = 6.66%: Let us focus on the cases where a country has competed in the 1st slot in the semi-finals. In 16 cases out of 30, the country qualified for the country. So the qualifying rate from this slot is about 53.33%. This puts the 1st position in the 11 place, therefore Albania will get the rank of 7 points. This leads to adding of 53.33 * 6.66 = 46.66 points.
- Average Rate Per Position = 6.66%: Let us focus on the cases where a country has competed in the 1st slot in the semi-finals. The average place the country has finished in this case is about 10.17. This puts the 1st slot in 11th place. Therefore let’s add 6.66 * 7 = 46.66 points.
- The number of “Friendly Countries” voting in the same semi-final = 6.66%. For each country, we’ve built a list of the top10 countries that tend to give them points (both televote and jury vote, in semi-finals and finals). In Albania‘s case, the country will have the option to rely on 5 of its friendly countries in this semi-final: Greece, Italy, Switzerland, Croatia, Austria. There are 5 countries in the same situation in this semi-final (having 5 out of 10 friendly countries): Albania, Portugal, Denmark, Switzerland, Austria. This puts each one of them in 5th place regarding this category. So each one of them will get a score of 13 points. Therefore, we can add another 6.66 * 13 = 86.66 points.
- Putting it all together, our model estimated Albania at 1210 points. Comparing this to the other countries in the semi-finals, Albania has the 4th highest ranking.
As you can see, our prediction for the 10 qualifiers is (not to confuse the points with the scale of points at Eurovision!)
- 🇺🇦 Ukraine – 1473 points
- 🇳🇴 Norway – 1340 points
- 🇳🇱 The Netherlands – 1257 points
- 🇦🇱 Albania – 1210 points
- 🇬🇷 Greece– 1197 points
- 🇲🇩 Moldova – 1077 points
- 🇦🇹 Austria – 1000 points
- 🇵🇹 Portugal – 967 points
- 🇦🇲 Armenia – 913 points
- 🇱🇹 Lithuania – 803 points
Appendix: You can see the list of friendly countries that voted in the same semi-final
| Country | List of Friendly Countries voting in the same semi-final |
| Albania | Greece, Italy, Switzerland, Croatia, Austria |
| Latvia | Lithuania, Norway |
| Lithuania | Latvia, Norway, Ukraine, Italy |
| Switzerland | Austria, Albania, The Netherlands, Iceland, Denmark |
| Slovenia | Croatia, Portugal |
| Ukraine | Armenia, Moldova |
| Bulgaria | Greece, Moldova, Albania, Austria |
| The Netherlands | Denmark, Austria, Switzerland |
| Moldova | Portugal, Ukraine, Italy, Greece |
| Portugal | France, Switzerland, Iceland, Armenia, Latvia |
| Croatia | Slovenia, Austria, Switzerland, Ukraine |
| Denmark | Iceland, Norway, The Netherlands, Latvia, Slovenia |
| Austria | Bulgaria, Slovenia, Lithuania, Switzerland, Iceland |
| Iceland | Denmark, Norway, Latvia |
| Greece | Albania, Bulgaria, Armenia |
| Norway | Denmark, Iceland, Lithuania, Latvia |
| Armenia | France, Bulgaria, Greece, The Netherlands |
Semi-Final 2 Analysis
As a reminder, this is the running order of the first semi-final:
Remember: The highest score you get, the highest your chances to qualify.
The table is too wide, so we’ve split it into 2 images:
As you can see, our prediction for the 10 qualifiers is (not to confuse the points with the scale of points at Eurovision!):
- 🇸🇪 Sweden – 1550 points
- 🇵🇱 Poland – 1427 points
- 🇫🇮 Finland – 1257 points
- 🇦🇺 Australia – 1253 points
- 🇨🇾 Cyprus – 1193 points
- 🇷🇸 Serbia – 1190 points
- 🇪🇪 Estonia – 1147 points
- 🇨🇿 Czech Republic – 1090 points.
- 🇷🇴 Romania – 1067 points
- 🇧🇪 Belgium – 973 points
Appendix: You can see the list of friendly countries that voted in the same semi-final
| Country | List of Friendly Countries voting in the same semi-final |
| Finland | Estonia, Sweden, Poland, Ireland |
| Israel | Azerbaijan, Finland, Australia, the United Kingdom |
| Serbia | Montenegro, North Macedonia, Germany, Sweden |
| Azerbaijan | Georgia, Czech Republic, Malta, Romania, Montenegro |
| Georgia | Azerbaijan, Cyprus |
| Malta | Australia, Azerbaijan, United Kingdom, Ireland, North Macedonia, San Marino, Montenegro |
| San Marino | Azerbaijan, Georgia, Montenegro, Malta, and Poland |
| Australia | Poland, Sweden, the United Kingdom, Finland, Israel, Spain, Belgium |
| Cyprus | Malta, San Marino, United Kingdom, Georgia |
| Ireland | United Kingdom, Czech Republic, Sweden, Belgium, Estonia, Spain |
| North Macedonia | Serbia, Montenegro, Azerbaijan |
| Estonia | Finland, Sweden, Ireland, Georgia, Israel |
| Romania | Spain, Israel, Azerbaijan, Cyprus, Malta |
| Poland | Germany, Ireland, the United Kingdom |
| Montenegro | Serbia, North Macedonia, Azerbaijan, San Marino, Malta |
| Belgium | Australia, Poland, Sweden, Ireland |
| Sweden | Estonia, Finland, Australia, Malta |
| Czech Republic | Spain, Belgium, Israel, Australia, Estonia |
Remarks
- Unsurprisingly, we see a different ranking by Spotify users and YouTube rankings. The main explanation is that users from different countries prefer using different digital platforms. For example, Spotifu is very popular in Western Europe.
- It is worth mentioning that jury votes are harder to predict. We believe that they can be somehow inferred from the trends that the model’s features give.
- Different countries have selected their songs at different times. However, when songs are popular, it can be easily seen, no matter how much time has passed since they were released.
- One can think about the model results as a prediction for the actual ranking, and not only the list of qualifiers.
- There are many different approaches to solve this challenge. Some of them take more time and are harder to explain. That’s why we chose a simple model.
What do you think about the concept?
Did you find the model’s predictions logical?
Would you be satisfied with this list of qualifiers?
If you share the content, please don’t forget to give a porper credit!
Read all about Eurovision 2022 on our homepage
Great work! Thank you!
Are you planning to make the same thing about the final?
I would like to try, let’s see it time will permit!