2017 Eurovision Final results (televote!) estimate: Can Italy (or Portugal or Bulgaria…) stop a seventh Swedish win?

Adrian Kavanagh, 12th May 2017

In these past, I used this model to successfully predict the Azerbaijan win at the 2011 contest,  the Denmark win in 2013, the Sweden win at the 2015 contest and Russia’s win in the televote at the 2016 Final, while this same model correctly identified 17 of the 20 qualifiers from the 2015 semi finals (although it proved decidedly less effective in predicting the 2016 qualifiers and predicted only 15 out 20 qualifiers for the 2017 semi finals). Now that we know the running order for the 2016 Eurovision Final I am going to use this to try and tease out who the likely winners will be of the 2015 contest will be. There are, however, a variety of factors (including the problems in terms of predicting the 2016 and 2017 semi final qualifiers) that suggest that the 2017 Final model may not be as accurate as in previous years (or at least the years leading up to the 2016 contest). The changes being made to the voting process in 2016 effectively mean that each country’s jury vote score and public vote/televote score will be treated as separate entities for this year’s contest – i.e. each country will award two separate scores – a jury vote score and a televote/public vote score. It is the latter of these two different scores/rankings (i.e. the televote score) that this model should be most effective in predicting.

With the numbers crunched, Italy, Sweden, Bulgaria and Portugal stand on top of the pile. Some of these countries have relatively good positions in the contest running order, some have a tendency to do well in terms of “friends and neighbours” and “diaspora” voting and all of these enjoy very high rankings in the bookies odds.  Other countries/finalists, such as Belgium, Romania, the United Kingdom, Armenia and Croatia, also figure strongly in relation to these factors, or some of these factors. But be wary!

  • This model cannot take account of the impact of the actual performances on both Final nights (including the Jury Final on the Friday night and Public/Televised Final on the Saturday night).
  • As the voting history statistics are based mainly on past televoting trends, the model cannot take account for the voting decisions of the highly influential professional juries, who have as much bearing on the Final result as the televotes have.
  • The voting history statistics for Australia are quite limited and based on just three contests (2015 Final, 2016 Semi Final 2 and 2016 Final) – contests that Australia finished 5th, 1st and 2nd in respectively – meaning that the Australia vote estimates are somewhat over-estimated as regards this particular factor (especially with countries such as Russia and Serbia not taking part in this year’s Final).


In terms of working out who may win this year’s Eurovision, various factors can be looked at, but the three most significant ones (that can be – sort of – quantified) are the Eurovision betting odds, past voting histories (involving the 26 finalists and the 16 other countries that lost out in the two semi-finals) and position in the running order. Performance quality and the staging of an entry (both in the show itself and the previous night’s dress rehearsal/jury final, which is the contest that the Eurovision juries get to vote on) are also vitally important, of course. As only a crystal ball can predict the quality and impact of the different acts’ Eurovision Final performances at this stage (although the betting odds will pick up on/reflect the quality of semi-final performances and perceptions/impressions of the other rehearsals), this analysis can just focus on past voting history, draw position and betting odds as a means of determining which countries are likely to do well in the 2017 Eurovision Final.

Past voting histories: During the televoting era, countries have shown a remarkable consistency in terms of the other countries that they vote for. This is probably most evident in the case of Greece and Cyprus’s tendency to award each other douze points in Eurovision contests – a similar relationship had been evident between Turkey and Azerbaijan in some of the more recent contests before Turkey withdrew from Eurovision. Similar trends can be observed for most other European contestants, including Ireland who showed a remarkable consistency during the “full-blown” televoting era in awarding its high Eurovision points to Lithuania, Latvia, Poland and the United Kingdom, especially from 2005 onwards. Such has been the consistency in Eurovision countries’ voting patterns during the 2000s that one can easily suggest the existence of different, geographically-based, voting blocs, which tend to award especially high numbers of points to certain countries (not necessarily always other bloc members) and from which bloc members can expect to attain their highest Eurovision points tallies.

The impact of what is often referred to as “bloc voting” – but I prefer to use the terms friends and neighbours and diaspora voting – has been tempered in recent years by the reintroduction of a professional jury voting element, as part of a 50-50 voting system involving televoting and jury votes. But given that televoting still accounts for half of the total votes being awarded by countries, the past voting histories of countries can still give a strong indicator as to the destination of their Eurovision votes at the 2017 Final. (And some juries may reflect the voting patterns of the general populace, also.) This suggests that some countries – due to an ability to score highly within one, or more, Eurovision neighbours/diaspora voting blocs – will start the contest at somewhat of an advantage to countries such as France, the United Kingdom, Spain, Germany, Portugal and Cyprus, which have struggled to win Eurovision points at various contests over the past decade and half. The change in the rules for the 2013, 2014 and 2015 contests as to how televotes and jury votes would be combined had the potential to even further blunt the impact of bloc voting. This was particularly evident in the fact that a low jury ranking from Ireland for Poland at the 2014 meant that country won no points from Ireland at that contest, despite the fact that the Polish act won the Irish televote. The change in the voting process for this year, however, negates the ability of a low jury vote score to blunt the impact of a high televote ranking.

To analyse this, I calculated, and ranked, the average number of points awarded by all the countries voting in the Final since the introduction of televoting in 1998 to the 26 countries that are taking part in this year’s final. (I used the televoting ranks also for the 2014 contests.) This ranking is then used to determine the finalists that the different voting countries are likely to award their 12 points, 10 points…and 1 point to in these semi-finals, based on past voting histories.

For example, the Irish 12 points would be expected to go to Denmark (with an average points tally of 6.9 from the Irish televote to Danish acts since the introduction of televoting in 1998). The next highest ranked country, Poland (6.7 points average), would be predicted to win 10 points from Ireland, followed by the next highest ranked countries, Australia (5.7), which would be predicted to win 8 points from Ireland, Sweden (5.1), which would be predicted to win 7 points from Ireland and the United Kingdom (4.9), which would be predicted to win 6 points from Ireland, and so on.

The predicted points’ destinations for all the other 41 voting countries in the Final are estimated in a similar manner. These points are then combined to calculate the overall predicted point tallies for the finalists based on past voting histories for these contests.   

Position in the Running Order: As discussed in greater detail elsewhere on this site and in The Eurovision Handbook 2014, vote patterns for past contests suggest that a country’s draw position can (in part) determine how well they do in that contest, with certain positions in the contest running order (generally later, although certain earlier draw positions such as 10th and 11th in the running order have proven to be relatively good ones also) being associated with a stronger likelihood of success/higher average points level as against other draw positions (generally those in the earlier part of the show). In most cases a good draw position does not convey an unsurmountable advantage (or disadvantage), which cannot be balanced by the quality of the entries involved and strong/weak performances on the night(s) in question or past voting histories.

This was evident in 2014, when Austria won the contest from 11th position in the running order and Sweden finished in 3rd place from 13th position in the running order, and in 2015, when Sweden won the show from 10th position in the running order. During the era of producer-led running order allocations, the 10th to 13th slots in the First Half running order have generally been associated with higher points tallies, in any case, given the tendency for producers to locate contest favourites, who are drawn to perform in the First Half of the show, in these running order slots.

But, as discussed elsewhere, a good position in the running order has been shown to give a notable advantage to different acts in the past, all things being even. For the purpose of these predictions, the average number of points won from different draw positions in all of the Eurovision Final contests between 2003 and 2016 was calculated and assigned to the countries selected to perform in those positions at Friday/Saturday night’s final.

Betting odds: Past voting histories (or a tendency for countries to award their higher points to certain countries) and position in the contest running order does have an influence on Eurovision success rates. But these do not account for other key determinants of success – song quality and the impact of a song/the way a song is presented on the Eurovision stage. These factors are hard to measure but a variety of methods might be employed to determine rough estimates of these, including pre-contest internet polls (although these may have their limitations, as is often the case with online surveys/polls).

The closest approximation that can probably be offered in terms of contest betting odds. As punters will want a return on their investment, it makes sense that they will be backing (what they perceive to be) the best entries in order to get this return. The link between betting odds and song quality/impact is not in keeping with the principles of exact science, but betting odds do at least offers some means of quantifying song quality. A number of contest analysts would argue that betting odds often offer the most potent means of estimating likelihood of contest success, as evident in the number of recent contests that have been won by the pre-contest favourite, or countries that rank high in the betting odds. This also allows for the identification of highly tipped entries, whose likelihood of doing well in the competition would not be evident from a study of prior voting patterns for those countries. The means of transforming the Final betting odds into Eurovision points was to identify the rankings (as of 2.00pm on 12th May 2017) of the different countries across a range of betting websites. Points were assigned to different rankings based on the average number of points earned by countries finishing in that position in Eurovision Final contests held over the 2004-16 period.

Combining the voting bloc/history patterns with the impact of draw position and the betting odd weightings, the following result would be predicted:

Draw Country Betting Odds Running Order Voting History Total
9 Italy 310 78 152 213
24 Sweden 170 162 212 179
25 Bulgaria 223 101 117 166
11 Portugal 249 110 34 161
23 Belgium 184 90 58 129
20 Romania 129 113 100 118
18 United Kingdom 152 152 15 117
13 Croatia 122 115 73 108
5 Armenia 99 76 153 107
14 Australia 52 76 230 103
12 Azerbaijan 65 96 150 94
22 Ukraine 19 112 222 93
10 Denmark 72 120 108 93
26 France 116 108 9 87
7 Moldova 88 84 59 80
6 Netherlands 78 86 71 78
15 Greece 34 69 162 75
8 Hungary 58 80 69 66
17 Norway 38 101 82 65
21 Germany 43 124 32 61
1 Israel 50 67 48 53
2 Poland 29 36 78 43
4 Austria 23 57 67 42
3 Belarus 14 66 68 40
19 Cyprus 12 78 39 35
16 Spain 3 80 28 28

Contest favourites, Italy, are predicted to win the contest televote by margin of 34 votes, especially given that the voting history factor (but not the running order factor) is also strongly favouring the Italian entry. The model estimates a close contest between Sweden, Bulgaria and Portugal to be the second ranked country in the contest/televote, with the model estimating that Belgium will fill out the Top 5. But if/when the Jury Vote scores are factored in, this could lead to certain jury-friendly acts, such as Bulgaria and Portugal – and maybe also the United Kingdom – possibly narrowing the gap between them and Italy, or possibly even overtaking Italy. It is worth remembering that Italy would actually have won the televote in 2015, but only finished 3rd overall in the contest because Sweden (and also Russia) fared significantly better than Italy as regards the jury vote rankings! And of course Russia won the televote in the 2016 Final, but Russia too only finished 3rd overall in that contest because Ukraine and Australia fared significantly better in terms of the jury vote.

Update: As of 6.00pm on 12th May, Portugal has overtaken Italy in the betting odds. If the changed pattern in the betting odds was factored into the above model, then the Italy point estimate would fall to 182 (and down to 2nd in the rankings) while the Portugal point estimates would increase to 191 (moving them up to 1st in the rankings). In this scenario, there would be a margin of just 25 points separating the two four countries in the model rankings.

This does not fully confirm with the bookies’ take on the contest, with countries such as Sweden, Australia and Ukraine being pushed up the rankings, either due to a good history of strong support at Eurovision from the different voting countries or due to a highly favourable position in the contest running order, or a combination of both these factors, as in the case of Sweden.

There are a number of countries’ whose strong ranking in the bookie odds does not translate into a similarly positioning in terms of the overall result estimate, but this is mainly due to poor results in past contests (meaning they score poorly under the voting history aspect) for these countries or due to an unfavourable position in the contest running order. However, it is worth remembering that Finland won in 2006 despite a series of extremely poor Finnish results across the decades prior to that win with the same trend being evident with Austria and The Netherlands at the 2014 contest and, to a lesser extent, with Belgium, Latvia and Israel at the 2015 Final and with Bulgaria and France at the 2016 Final. So scoring lowly under the voting history factor should not rule out a country’s prospects entirely (especially with the growing influence of the voting juries helping to balance the friends and neighbours voting and diaspora voting factors out). A model such as this (not withstanding some successes in more recent contests, ahem) would not have been able to predict that Finland win in 2006 and was not able to fully predict the strong showing by Austria and The Netherlands in 2014. Ultimately when it boils down to it – any country can win this contest with a combination of the right song, performer and staging.

It is also worth noting that this model suggests that only two of the first seven positions may be filled solely by countries performing in the first half of the contest. The contest had been won only by countries performing between the No.17 and No.24 positions between 2005 and 2013 (with the bulk of these wins falling to countries performing in the four slots between No.17 and No.20). This did not prove to be the case in 2014, obviously, when Austria won from 11th position in the contest running order, or in 2015, when Sweden won from 10th place in the running order. But if you think the running order will have a greater bearing on the result than is being suggested by this model, then countries such as Bulgaria, Sweden, Romania, Belgium and the United Kingdom may warrant some further closer consideration.

What about France? Well performing last on the night would appear to be a very good draw given that acts performing later in the contest have traditionally done better, but this does not have the same value as performing last in a semi-final had for Jedward in 2011 and 2012, mainly because of the larger number of countries in a final! Only five finals previously have involved as many as 26 countries and the scores/positions attained by the countries performing in these positions has been highly disappointing to say the least:

  • the 2003 contest in Riga – in which the country performing last on the night, Slovenia, finished in the lower placings
  • the 2012 contest in Baku – in which the country performing last on the night, Moldova, finished in a mid-table position
  • the 2013 contest in Malmo – in which the country performing last on the night, Ireland, finished in last place
  • the 2014 contest in Copenhagen – in which the country performing last on the night, the United Kingdom, finished in 17th place, despite having been among the pre-contest favourites.

Countering this argument, it is ironic to note that, when the contest was extended temporarily (to facilitate Australia) to 27 countries in 2015, Italy actually won the televote (but not the overall contest) from the last position in the running order. Armenia also finished in the Top 10 in the televote last year when performing from the last position in the Final running order.

Statistically the best draw position to get in a Eurovision Final is to perform third from the end, which is the position occupied by Sweden In terms of actual running order positions, the best slot to get would be the 18th position in the final running order, which is occupied by the United Kingdom.

So to conclude, this study tries to offer a prediction as to how the 26 countries in the Final may fare in the televote. Due to a lack of a functioning crystal ball, this analysis cannot assess the impact of a vital element of Eurovision success or failure, namely the quality and impact of the performance on the Eurovision stage, whether at the main show on Saturday or at the equally-important Jury Final on Friday night.

Certain types of jury-friendly songs have been doing better in the contest over the past two years, as the result of the 2014 Final particularly showed. This suggests that certain jury-friendly songs may do significantly better in the actual contest than the estimated result that suggested by the model, which largely can only attempt to estimate the televote scores (and not the jury vote score). Such countries would, hence, be expected to fare well/better than the model is suggesting in the first phase of vote results on Saturday night, but then drop back somewhat when the televote scores are added in at the end of this process.

A well-thought out stage act (and it doesn’t have to be gimmicky…) and a strong performance can push previously unfavoured acts into contention, but weak live performances can, in turn, nix the prospects of contest favourites, as has happened in the past! This is a very long-winded way of saying that if you decide to bet on the Final results based on this study and you make a few bob, then that’s great and mine’s a TK Red Lemonade (with the fizz taken out – the bubbles go to my head). On the other hand, if your bet goes horribly wrong, then… hey, you have been warned in advance by me…


Tags: , , ,

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: