Tomorrow is Eurovision, the biggest non-sporting event in the European TV Calendar, and we want to know who’s going to win. We’ve looked at collusive voting patterns to see if it can give us a clue.

Our model takes voting data from 1975 to the present day and tests for any unusual voting patterns.1 The first evidence of collusion is between the United Kingdom and France during 1975-1980, followed by an even more surprising alignment between Sweden, Germany and Israel from 1981-1986. During 1986-1990 we begin to see more obvious partnerships, with the Cyprus-Greece alliance appearing alongside a more natural pairing of Sweden with Denmark. Into the 1990s, Cyprus and Greece continue to wind up Wogan by exchanging ‘douze points’ while other pairs soon evolve into small groups – there is a “Viking Empire” centred on Sweden and a “Balkan Bloc” of former Yugoslav republics (see diagram). Since 2000 these small groups have continued to expand whilst there is the creation of a “Warsaw Pact” between Russia, Poland and Ukraine.

These groups could be down to distinct regional variations in tastes of music but this does not explain the late emergence of such blocs in the history of the contest. It’s also possible that the people of Europe are using the contest as a chance to express their opinions on the political behaviour of their neighbours – Greece and Cyprus have obvious historical ties whilst the original “Balkan Bloc” was centred on Croatia and the other republics of the former Yugoslavia. Potentially the most likely reason for the increase in collusive voting is that voters realise that it increases their own country’s chance of winning the contest. Successful partnerships have been copied by voters in other countries simply as a means of obtaining votes in return.2 The introduction of a public vote is another factor behind the acceleration in collusion levels in recent years.3 The smaller “Balkan Bloc” and the “Warsaw Pact” groups have expanded into a massive Eastern European conglomerate and this was instrumental in giving Russia victory last year. The “Viking Empire” is growing too but countries in Western Europe are finding things harder. Germany and France, for example, don’t have any friends at all. The diagram below shows a bit more detail.

So does any of this help with the all-important question of predicting this year’s winner? Well, there is a big change this year with the return of national judges. Whether this makes any difference to levels of collusion remains to be seen. However if voting follows the patterns seen in recent years it is hard to look beyond an Eastern European country. Our top tips are Elena from Romania singing the Balkan Girls, Sakis from Greece with This is our Night and an outside shout for Alexander from Norway hoping for a Fairytale victory. With thanks to Derek Gatherer’s paper “Comparison of Eurovision Song Contest Simulation with Actual Results Reveals Shifting Patterns of Collusive Voting Alliances”, published in the Journal of Artificial Societies and Social Simulation vol. 9, no. 2 in March 2006. Footnotes 1. Due to the considerable variability in the number of participants and regular changes to the voting rules we’ve used a Monte Carlo approach to simulate the entire history of the contest. Monte Carlo simulation is ideally suited to this type of problem when a lack of data, a high level of complexity, or the short timescales mean analytical solutions are not possible. We’ve generated thresholds for statistical significance (at the 95% level) and compared against observed results from the real contests. 2. Collusion could therefore be considered a meme (Dawkins, The Blind Watchmaker) – a horizontally spreading cultural behaviour that has progressively colonised the contest. It has an evolutionary nature with one-off replications of colluding countries (occurring for whatever reasons) surviving due to a greater chance of victory and natural voting selection. 3. The graph below shows how many pairs of countries have a strongly positive coefficient of collusion in each five-year period. The increase in apparent collusion over the last 10 years is 10-fold.