Bye Bye Borda?

Since 1975 the Eurovision Song Contest has used the same point allocation system: a version of a Borda count. For the first 20 years of the Contest, all sorts of other scoring—and point allocation—systems were tried. The quadruple tie in 1969 led to a mini-boycott the following year, in fact. Thus in Vienna we saw the douze points era enter its fourth decade.

Could it be time for a change?

In the last decade, we have had several run-away results : no one would have topped Loreen or Lena or Alexandar Rybak, each of whom roared to victory. But when has the douze points allocation of points impacted the results—particularly in a close year, where the winner has won by a handful of points?

I decided to take a look–and hold that thought. But first, let’s take a quick look at voting systems and the Eurovision Song Contest.

We Are the Winners

While we might have been using douze points for a long time, how (and who) allocates points from each host broadcaster has changed. Sometimes there have been minor tweaks or local adjustments; other times we’ve seen wholesale shifts and restructures. Since our Dutch friends Teach-In Ding-a-Dong-ed their way to victory in Stockholm, we have seen a transition from jury-only scores (1975 to 1996), through a gradual introduction of limited televoting (1997 to 2001). For a couple of years in the noughties, a handful of broadcasters elected to use a 50/50 mix of televotes and jury votes for their points allocations. And throughout the televote era there has always been a role for juries: as a backup for televotes, in case of too low a televote turnout or if the televote system (phone calls, SMS or both) fails.

Beginning with the 2008 Contest, we saw the purposeful, generalised re-integration of jury voting to moderate televote results—results that were seen by many ESC observers to have little to do with the calibre of the song or presentation. In particular, concerns with certain diasporic, post-colonial or bloc voting patterns drove this change. Initially this was operationalised by replacing the entry ranked 10th in each semi-final’s televoting results with the highest ranked jury entry that was not in the televoting top 10. This has sometimes been known as the “Macedonian rule” since it was Macedonia that often seemed to qualify 9th or 10th thanks to ex-Yugo support–and it was their entries in 2008 and 2009 entries that were bumped out of Saturday night’s final as a result.

From 2010 onwards a combination of televote and jury scores (or rankings–there is a difference) has been used to allocate which 10 entries receive points from each participating host broadcaster. While the different versions of this mélange extraordinaire have all arguably had own limitations, broadly speaking the impact has been a broader range of countries making it to each year’s Grand Final. We still see a consistently high number of Scandinavian and ex-Soviet countries supporting their “friends’” entries, but strong entries from anywhere have managed to get out of the semi-finals, often getting into the top 10 of Saturday night’s table too. This seems to have incentivised countries like the Netherlands and France to send strong entries by established artists. Perhaps they won’t win; at least they won’t be humiliated if they deliver strong performances.

Ultimately the Eurovision is both a contest and an evening of light entertainment. The calibre of the entries determines the overall quality of that experience for viewers. All the LEDs, drone cameras and gimmicks in the world cannot cover up pedestrian songwriting and bum notes. At least 26 times in one night. So there needs to be an element of meritocracy–a result based on quality–or the Contest isn’t really a Contest.

Runaway

Point allocation in the douze points system is what’s known as an interval scale. While each host broadcaster allocates a top 10 ranking, they are not scored equally between ranks (that would be an ordinal scale). Under the current system, songs ranked third through tenth progress in increments of one: the 9th song gets one more point that 10th and one point less than 8th. An effective weighting that the difference in standard between these entries is equal.

However the top two entries both get a significant bump of two points compared to their nearest ranked peers: 12 points for the top entry and 10 points for the runner up. As a result the highest rank entry gets 50% more points than the third place one (8 versus 12 points, with an extra 4 points being 50% more than 8). The second ranked entry gets 25% more points than the third place one (8 versus 10 points, with an extra 2 points being 25% more than 8). Multiply this over a few dozen votes (or, as in more recent years, 30 plus votes) and we rather easily get runaway scores and runaway winners–even when there are two or three entries with similar numbers of first, second and third place rankings.

This isn’t what a Borda count is supposed to do: in fact, it’s supposed to eliminate ambiguity when there’s a close result. So how do we unpack the current system, to get a sense of how much this is skewing the results?

Together We Are One

Given that in most years since 1975 we have had a clear winner. We could have recalculated all the scores from 1975 onwards (OK, I actually did do that, thanks to some nifty Excel programming and Johnny Chan’s amazing Eurovision data archive). But that’s rather uninteresting. Instead, we looked at three Eurovision Song Contest finals where the results were very close (1988, 1991 and 2003).

We specifically looked at two different scenarios, each of which flattens out the scoreboard. We calculated the final results if the voting system was switched either to:

  • A reduced Borda count, where the top two scores are 11 and 9 points (referred to as the onze points system) rather than 12 and 10 points, or
  • A non-Borda count (an ordinal scale), where the top two scores are 10 and 9 points (referred to as the dix points system)

The results are rather interesting. Let’s take a look!

A Little Bit, A Little Bit More

In 2003 Sertab Erener earned Turkey’s first (and sole) victory with Every Way That I Can. Belgium’s Urban Trad finished two points behind with Sanomi, closely followed by Russia’s t.A.t.U. and Ne Ver’ Ne Boysia a mere single point further back.

Here’s a comparison of these two entries’ scores under the douze, onze and dix points systems, with the highest score under each system in bold:

  Douze Onze Dix
Turkey 167 157 153
Belgium 165 157 154
Russia 164 155 150

Under douze points Turkey won by two points. However, under the onze points system Turkey and Belgium are now tied—though Sertab’s four top scores trump Urban Trad’s three, meaning Turkey would have retained the crown. But under a dix points system Belgium would have one by a single point.

With any of these systems, this would have meant a super-exciting end to the evening. Ah, but what’s another year…

Tie-Break

In 1991the current Borda count resulted, at least initially, in a tie. Carola’s Swedish entry Fangad av en Stormvind and Amina’s French entry C’est le dernier qui a parlé qui a raison each ended up with 146 points.

Under the 1991 rules, the countback looked at each entry’s number of top scores (or douze points; both had 4), then dix points (Carola had 5; Amina 2), thus making Sweden victorious. Under current rules whichever entry received the most douze points wins; if they are still tied whichever entry received more votes from more countries”becomes the winner. Meaning Amina would have won rather than Carola.

Here’s a comparison of these two entries’ scores under the douze, onze and dix points systems, with the highest score under each system in bold:

  Douze Onze Dix
Sweden 146 137 133
France 146 140 136

Under either of these alternate systems France would have scored an outright victory over Sweden. Regardless, still a very exciting voting sequence, non?

Era Stupendo

The 1988 Contest also featured a two-horse race. In the end Céline Dion gave Canada Switzerland its second-ever victory with Ne partez pas sans moi, by a single point over the UK’s Scott FitzGerald with Go. It remains the narrowest non-tie victory in the douze points era.

Here again is a comparison of these two entries’ scores under the douze, onze and dix points systems, with the highest score under each system in bold:

  Douze Onze Dix
Switzerland 137 129 126
UK 136 127 124

For 1988, the ranking does not change, although the differential actually would increase slightly between under the onze or dix points systems.

Ambiguous or Exciting?

In years where the difference between first and second was two (or less) points, we end up with a close result regardless of what system is used to calculate total scores. But in two instances, if we were to switch from the Borda interval scale (douze points to a wholly ordinal scale (dix points) we would still get a close result—with a different winner.

The rationale for the current Borda count is to decrease the chances of a tie score. However, we have still had a tie (in raw points accrued) in 1991. The detailed tie-break rules determined who won: any voting system where a mathematical tie is possible needs to have such rules regardless. So the douze points system does not protect against ambiguity any more than any other system. Anytime you aggregate scores from a diverse sample there is scope for a tie. If the goal of the Borda count is to avoid ties, it has not always worked.

Further, by weighting the first ranked entry 50% higher points-wise than the third place entry, the system is quantifying the first ranked entries as 50% more popular. Neither voters (by casting single votes for individual entries) or jury members, (who rank songs ordinally) are using a Borda count.

This year’s voting in Vienna was the first genuinely exciting voting sequence since the semi-final system was introduced in 2004. An exciting voting sequence is much more entertaining. Here’s how things would have looked with one of these alternative voting systems:

  Douze Onze Dix
Sweden 365 335 323
Russia 303 282 277

There’s no change in the rankings, but the gap narrows between Sweden and Russia: from 62 to 53 to 46 points.

2 Responses

  1. So in effect the 12-10-8… system is still working well! As you say, most years there is a clear favourite and winner and whatever system we have would make no difference…

    Nice to see it explained though!

    One point I would dispute is this year’s voting. I have put forward elsewhere (and no-one has challenged me) that artificially placing the juries by an algorithm made this years scoring look more obviously staged than before just to create a sense of excitement. Once we got to jury 25, from what appeared to be a tight Contest suddenly changed to Sweden getting top scores and Russia suddenly hitting lower scores and zeroes. As we got to jury 30, it was obvious who was going to win and I suspect a lot of viewers tuned out due to the shows overrunning. A tight Contest would have us getting to the last few juries before the winner was obvious, not 15 out!

  2. John says:

    True, but compared to every other year since 2004 it was exciting with 3 in the mix for the first half, then a battle between two until the last 10. In the Arena it was electric.

    So hurrah for the algorithm!

Leave a Reply