Re: UKRAINE 2018
Did some calculations earlier today, based mostly on semifinal scores, to try and predict tonight's winner.
For jury scores, I looked at each individual juror's scores for each song in the semis to figure out their most likely rankings of the songs that reached the finals (giving averaged half-points to break ties fairly), then combined them to figure out the likely jury ranking for the finals. Points that will actually be awarded are to the left, while their tally in the unconverted jury scores is in brackets to the right:
Jury Points:
6P Lelya
(15P)
5P Laud
(13P)
4P Melovin
(12.5P)
3P Kadnay
(11.5P)
2P The Erised
(8P)
1P Vilna
(3P)
Predicting the televotes was a little more complex, since I wasn't sure which was a better metric: televote scores in the semis or live performance views on YouTube. Both have their ups and downs: in particular, looking at semifinal points is more likely to capture the group of actual voters without including e.g. foreign views (which won't influence the final score on the night), but looking at YouTube could give a better idea of how the hype around songs has grown over time, and is less influenced by the chance that one semi is stronger than another.
So for the televotes, I've made three tallies: one based purely on semifinal televote scores (with ties broken by the percentage of the total semifinal vote that they managed to attract), one based purely on YouTube, and one that combines the rankings of those two systems with a 50/50 weight.
SF Televotes Only:
6P Melovin
(9P, 25.49%)
5P Vilna
(9P, 17.66%)
4P The Erised
(8P, 17.49%)
3P Kadnay
(8P, 15.40%)
2P Laud
(7P, 14.84%)
1P Tayanna
(7P, 13.23%)
SF Televotes and YouTube Views Combined:
6P Melovin
(12P)
5P Vilna
(9P)
4P Tayanna*
(6P)
3P Kadnay
(6P)
2P The Erised
(5P)
1P Laud
(4P)
(Note: Favoured Tayanna over Kadnay in the tiebreak here, as her lead over him in the YouTube views - 362K vs. 270K - seems more significant than Kadnay's lead in the televote percentages, i.e. 13.23% vs. 15.40%.)
YouTube Views Only:
6P Melovin
(SF2, 466K)
5P Tayanna
(SF2, 362K)
4P Vilna
(SF1, 338K)
3P Kadnay*
(SF2, 270K)
2P Laud
(SF1, 299K)
1P The Erised
(SF1, 214K)
(Note: Kadnay's performance has only been uploaded for a week, but almost has as many views as Laud's, which has been up twice as long. This made me bump him up by a position on the assumption that he would've topped Laud if the two had been uploaded for an equal length of time).
And so, of course, the final result in my estimation depends on which of these televote models we use. As a result, I took a look at each possibility to get an idea of different ways that things could go. Note that, for these systems, the televote takes precedence over the jury vote in case of a tie, as I'm assuming that rule hasn't changed from last year.
Jury Points + SF Televotes:
1st Melovin
(10P: 4J + 6T)
2nd Laud
(7P: 5J + 2T)
3rd Tayanna
(7P: 6J + 1T)
4th Vilna
(6P: 1J + 5T)
5th The Erised
(6P: 2J + 4T)
6th Kadnay
(6P: 3J + 3T)
If SF televote scores are a better indicator than YouTube views, then Melovin should have an easy win. He'll be the only one to get a good score from both the jury and the televote, with the televote killing off other jury favourites Laud and Tayanna while the jury kills off televote runners-up Vilna and Erised (with Kadnay not threatening the top scores in either group).
Jury Points + Combined SF/YouTube Scores:
1st Melovin
(10P: 4J + 6T)
2nd Tayanna
(10P: 6J + 4T)
3rd Vilna
(6P: 1J + 5T)
4th Kadnay
(6P: 3J + 3T)
5th Laud
(6P: 5J + 1T)
6th The Erised
(4P: 2J + 2T)
If we assume SF televotes and YouTube counts to be equally useful indicators, then things look much closer at the top: nobody comes close to Melovin and Tayanna, but those two tie on 10 points. Again, though, Melovin wins in this scenario, this time by being favoured in the tiebreak as the televote winner.
Jury Points + YouTube Views:
1st Tayanna
(11P: 6J + 5T)
2nd Melovin
(10P: 4J + 6T)
3rd Laud
(7P: 5J + 2T)
4th Kadnay
(6P: 3J + 3T)
5th Vilna
(5P: 1J + 4T)
6th The Erised
(3P: 2J + 1T)
If YouTube votes resemble the final televotes more closely than the SF televotes do, then Tayanna looks likely to take the win, topping Melovin by a point by virtue of her jury lead being slightly larger than his televote lead.
Comments and summary: Melovin looks rather strong going into the final, as the semifinals suggest he'll come around third on jury scores and then will top the televotes with ease regardless of which indicator is stronger. In most circumstances, that'll see him win overall. He's also got a good chance of topping Laud on jury votes to come second with juries: he did almost as well with the juries in SF2 as Laud did in SF1, and SF2 seems to have had tougher competition. If he
does manage that top-two jury score, then he's very likely secured himself the ticket to Lisbon.
Tayanna has a chance to win, but to do so, would probably require three factors in her favour. These are that a) the juries continue to support her as their favourite (relatively likely, but not a guarantee), b) Melovin stays in third or below with the juries (seems around a 50/50 chance), and c) YouTube is a more accurate bellwether than SF televotes, meaning Tayanna takes second place or above in the televote (no idea on the probability here, but I'll call it a fifty-fifty).
With that in mind, I'd say we've got around a 65% chance of Melovin winning as expected, a 25% chance of Tayanna winning with the conditions above, and then a 10% chance of something unexpected.
This third category could include a whole range of things. Melovin could grow on the juries and top both categories; Tayanna could make a surprise televote win if she's had a swell of support in Ukraine; other acts could impress on the night, improving on their semifinal vocals and staging by enough to overtake the two favourites; or those two could just make an unforced error and flop, deciding on a weird staging overhaul or hitting a series of bum notes. Anything can happen! I just thought I'd try my hands at an analysis here - and I look forward to figuring out which model was closest to the actual results