Collaboration in ESC finals from 2003 to 2015
Choose a country on the map
Choose dataset
What is this map about?
This map is an attempt to depict the collaboration between Eurovision Song Contest competitors using the points awarded in the finals from 2003 to 2015. Televoting was not mandatory before 2003, so I would compare apples to oranges if I included the contests before 2003. In 2016, the voting system was changed again, so I didn’t include contests after 2015, either.
Where does the raw data come from?
It’s obtained from the Wikipedia articles about the contests from 2013 to 2015.
You can download a JSON file containing the points every single country received from the other countries here. All countries are represented by their ISO 3166 code.
This JSON file lacks the information whether two certain countries ever participated at the same time, though. So trying to reproduce the data in the map with just this file will lead to different results, because I didn’t “punish” countries who never contributed to other countries when they never even had the opportunity to do so.
The map is obtained from Wikimedia. I altered it a little bit, though.
How was the raw data processed?
I wrote a Python program to process the data and I will probably publish it some day. It’s a huge pile of spaghetti code I definitely want to clean up first.
Basically, I used the following formulas:
-
p(A,B)
(total points given by A to B) -
G(A)
(total points given by A) -
R(A)
(total points given to A) -
S(A,B) = p(A,B) / G(A) * 100
(A’s support for B)
Example: Almost 14% of all points Germany ever awarded went to Turkey. -
D(A,B) = p(B,A) / R(A) * 100
(A’s dependence on B resp. B’s contribution to A’s points.)
Example: Almost 17% of all points Poland ever got were awarded by Germany.
To determine the overall collaboration C(A,B) between two countries, I standardized the respective values for S and D (i.e. S(A,B), S(B,A), D(A,B) and D(B,A)) and calculated the average of these four values.
In case one or more of these values didn’t make sense, I didn’t include them, but reduced the divisor by one instead.
Example: If A never participated in the finals, it never got any points, so it wouldn’t make much sense to calculate its D(A,B) or S(B,A). Therefore I would only standardize S(A,B) and D(B,A) and divide their sum by two.
Finally, I rounded all values to two decimal places.
About the color: The map uses an HSL representation, where the lightness value is set according to the respective values for S, D or C (see above). When S or D are depicted, the lightness value equals 100-4x, where x is the respective value for S or D. For example, if the map shows the support of country A for other countries and country A gave 10% of its total points to country B, then the color of country B has a lightness of 60%.
Since the value for collaboration is basically an average z-score and these values are (1) usually smaller than the S or D values and (2) in part negative, I had to adjust the color formula in order to gain a half-decent contrast in the map. So when depicting the overall collaboration C, the lightness value equals 75-8C.
This feels arbitrary, but it just means that you cannot compare the collaboration data to the other data sets by just comparing the color. It’s perfectly fine to compare the countries within one data set by comparing the color, though.
Found a bug, got any feedback or just want to ask a question?
Got any political concerns regarding the map?
Discuss them on Wikimedia.