Semi-final qualifications 2016: track record as predictive tool
Tempting though it might be to try and build a regression model to predict who might qualify from this year’s two semi-finals, that would be a lot of work. And hard. But mostly too much a lot of work. Instead, we’ve decided to look at past experiences qualifying from either of the two semi-finals for any year’s Grand Final.
Here are some of the key consideration that inform this analysis:
- We looked at each country’s record qualifying from any semi-finals, including the monster semis from 2004-2007
- However, we started with 2008, for parsing lists of qualifiers the first year of two semi-finals each producing 10 qualifiers
- We did not count being pre-qualified for a Grand Final as a qualification
- We ignored the “wild card” rules of 2008 and 2009 and used instead the 10 highest ranked entries in those semi-finals, rather than the 10 qualified finalists
Those were our guiding principles.
Next, we went back to each year’s semi-final scoreboards from 2008-2015 to see how each place in the performing sequence did. Most years these slots were randomly assigned; the last few years we’ve had the awful “producer-led draw”. However, rather than focus on scores or rankings we simply used “not qualified” (zero) and “qualified) (one). Because for this analysis, that’s the focus: getting over this first hurdle towards a potential win overall.
Finally, we did a really statistically lazy thing: we merely averaged the percentage of qualifications (by the country/broadcaster) with the percentage of entries from that spot in the sequence having qualified. Let’s jump in…
Semi-final One
Here’s our data crunching for the first semi-final:
Slot | Country | Artist | Song | Slot qual | Country qual | Average |
1 | Finland | Sandhja | Sing It Away | 25% | 60% | 43% |
2 | Greece | Argo | Utopian Land | 63% | 100% | 81% |
3 | Moldova | Lidia Isac | Falling Stars | 25% | 70% | 48% |
4 | Hungary | Freddie | Pioneer | 13% | 83% | 48% |
5 | Croatia | Nina Kraljić | Lighthouse | 50% | 44% | 47% |
6 | Netherlands | Douwe Bob | Slow Down | 63% | 25% | 44% |
7 | Armenia | Iveta Mukuchyan | LoveWave | 75% | 88% | 82% |
8 | San Marino | Serhat | I Didn’t Know | 50% | 16% | 33% |
9 | Russia | Sergey Lazarev | You Are the Only One | 63% | 100% | 81% |
10 | Czech Republic | Gabriela Gunčíková | I Stand | 63% | 0% | 31% |
11 | Cyprus | Minus One | Alter Ego | 13% | 40% | 26% |
12 | Austria | Zoë | Loin d’ici | 75% | 33% | 54% |
13 | Estonia | Jüri Pootsmann | Play | 63% | 42% | 52% |
14 | Azerbaijan | Samra | Miracle | 88% | 100% | 94% |
15 | Montenegro | Highway | The Real Thing | 63% | 29% | 46% |
16 | Iceland | Greta Salóme | Hear Them Calling | 75% | 66% | 71% |
17 | Bosnia and Herzegovina | Dalal & Dean feat. Ana Rucner & Jala | Ljubav je | 63% | 100% | 81% |
18 | Malta | Ira Losco | Walk on Water | 50% | 44% | 47% |
Next we plotted the distribution of the averages. This isn’t a Gaussian distribution, but we wouldn’t expect it to be anyways: different years had different numbers of entries in a semi-final, different countries participated, and each year is its own statistical event.
We subsequently looked for points across these averages at which we could break the distribution down with some typology of performance. In the end we settled on:
Level | Range |
Poor | 0-35% |
Marginal | 36-46% |
Good | 47-70% |
Excellent | 71=100% |
For semi-final one we get the following distribution: three (3) poor, five marginal, five good and five excellent. See below:
Slot | Country | Average | Prediction |
1 | Finland | 43% | Marginal |
2 | Greece | 81% | Excellent |
3 | Moldova | 48% | Good |
4 | Hungary | 48% | Good |
5 | Croatia | 47% | Good |
6 | Netherlands | 44% | Marginal |
7 | Armenia | 82% | Excellent |
8 | San Marino | 33% | Poor |
9 | Russia | 81% | Excellent |
10 | Czech Republic | 31% | Poor |
11 | Cyprus | 26% | Poor |
12 | Austria | 54% | Marginal |
13 | Estonia | 52% | Marginal |
14 | Azerbaijan | 94% | Excellent |
15 | Montenegro | 46% | Marginal |
16 | Iceland | 71% | Good |
17 | Bosnia and Herzegovina | 81% | Excellent |
18 | Malta | 47% | Good |
From this, our qualifiers would be: Greece, Moldova, Hungary, Croatia, Armenia, Russia, Azerbaijan, Iceland, Bosnia & Hercegovina and Malta.
Second Semi-final
Using the same approach for the Thursday night show, we get:
Slot | Country | Artist | Song | Slot qual | Final qual | Average |
1 | Latvia | Justs | Heartbeat | 63% | 36% | 49% |
2 | Poland | Michał Szpak | Color of Your Life | 38% | 33% | 35% |
3 | Switzerland | Rykka | The Last of Our Kind | 38% | 20% | 29% |
4 | Israel | Hovi Star | Made of Stars | 75% | 42% | 59% |
5 | Belarus | Ivan | Help You Fly | 25% | 33% | 29% |
6 | Serbia | Sanja Vučić ZAA | Goodbye (Shelter) | 75% | 71% | 73% |
7 | Ireland | Nicky Byrne | Sunlight | 38% | 50% | 44% |
8 | Macedonia | Kaliopi | Dona (Дона) | 50% | 42% | 46% |
9 | Lithuania | Donny Montell | I’ve Been Waiting for This Night | 38% | 55% | 46% |
10 | Australia | Dami Im | Sound of Silence | 50% | n/a | 50% |
11 | Slovenia | ManuElla | Blue and Red | 50% | 33% | 42% |
12 | Romania | Ovidiu Anton | Moment of Silence | 50% | 100% | 75% |
13 | Bulgaria | Poli Genova | If Love Was a Crime | 88% | 11% | 49% |
14 | Denmark | Lighthouse X | Soldiers of Love | 75% | 73% | 74% |
15 | Ukraine | Jamala | 1944 | 63% | 100% | 81% |
16 | Norway | Agnete | Icebreaker | 50% | 78% | 64% |
17 | Georgia | Nika Kocharov & Young Georgian Lolitaz | Midnight Gold | 63% | 75% | 69% |
18 | Albania | Eneda Tarifa | Fairytale | 50% | 55% | 53% |
19 | Belgium | Laura Tesoro | What’s the Pressure | 38% | 27% | 32% |
Next we plotted the distribution of the averages, using the same typology:
Level | Range |
Poor | 0-35% |
Marginal | 36-46% |
Good | 47-70% |
Excellent | 71=100% |
For semi-final two we get the following distribution: four (4) poor, six marginal, four good and four excellent. And one wildcard: Australia, which hasn’t yet competed in a semi-final. See below:
Slot | Country | Average | Predict |
1 | Latvia | 49% | Good |
2 | Poland | 35% | Poor |
3 | Switzerland | 29% | Poor |
4 | Israel | 59% | Good |
5 | Belarus | 29% | Poor |
6 | Serbia | 73% | Excellent |
7 | Ireland | 44% | Marginal |
8 | Macedonia | 46% | Marginal |
9 | Lithuania | 46% | Marginal |
10 | Australia | 50% | Good* |
11 | Slovenia | 42% | Marginal |
12 | Romania | 75% | Excellent |
13 | Bulgaria | 49% | Marginal |
14 | Denmark | 74% | Excellent |
15 | Ukraine | 81% | Excellent |
16 | Norway | 64% | Good |
17 | Georgia | 69% | Good |
18 | Albania | 53% | Marginal |
19 | Belgium | 32% | Poor |
From this, our qualifiers would be: Latvia, Israel, Serbia, Romania, Denmark, Ukraine, Norway and Georgia. That’s only 8. Given how well Australia did across Europe with both televoters and juries in 2015, we grade them as a Good*. Still only gives us 9 qualifiers.
We have two “marginal” countries – Lithuania and Macedonia – at the top of their band (46%). I would err on the side of weighting the country score—which is significantly reflective of what sorts of resources go into selecting and promoting an entrant in any given year—over the slot score. That puts Lithuania ahead of Macedonia.
However…
This analysis does not take into account anything else, such as song quality, profile of artist, or diaspora or regional voting blocs. Fear not: we are also crunching those numbers. We will look at numerous attributes of entries and artists, including what we can glean from national finals and pre-Contest live performances.
Stay tuned!
Leave a Reply