Before we begin, let’s all take a moment to appreciate the full video for San Marino’s dearly departed entry. Check out the monocle pluck at 1:30.
Anyway, a few big surprises in the first semi-final. Greece and Bosnia-Herzegovina both failed to qualify for the first time ever, while the Czech Republic made it to their first final. Both Iceland and Finland are out, despite the boost of having Sweden voting. Austria and Netherlands were both ranked as unlikely qualifiers by the model, but it’s a little unclear to me why—they’ve both had quite good records recently.
I can think of three possible explanations for this kind of upset. First, it might be a fluke. There are lots of possible unlikely events in any contest, and some of them are going to happen every time. I don’t think this is what’s happening here, but it’s hard to estimate how likely it is that something this weird would happen.
A second possibility is that times have changed. The model currently has no time-dependence, so a performance last year is considered as relevant as one in 2004 (which is the current cut-off year). That would certainly explain Austria and Netherlands being ranked so low, and possibly also the overconfidence about Greece and Bosnia-Herzegovina.
The final possibility is that this is a consequence of the new voting system. The EBU don’t release any voting information from the semi-finals until after the final, so we can’t see where the votes came from. It could be that the model needs a more radical rejigging to account for the new system.
In any event, I don’t have time to do anything drastic before Thursday, so we push on with the second semi-final using the current model. But first, a little caveat.
While running the updated simulations for the second semi-final, I uncovered a bug in my model code. The model was overweighting performances from countries with very little data in the system1. This usually wouldn’t be a problem (and the bug had probably been there for at least a year), but it did mean that the model greatly overestimated Australia’s chances this year.
I’ve fixed the bug for everything from now on, but a prediction is a prediction, so I’ve let the pre-contest predictions stand as they are.
Updated final predictions
As you can see, Australia’s chances have taken a bit of a nosedive. This is completely down to fixing the bug I described above. They’ve still got a very respectable 6% shot at the title, but my 20-1 bet doesn’t seem so clever any more.
Somewhat confusingly, Austria have leaped up towards the top of the table. Although Austria are definitely an outside shot with the bookies (60-1 at time of writing), the same was true of Conchita Wurst a week before the 2014 contest. An Austrian win wouldn’t be so much “stranger things have happened” as “this precise strange thing happened before, two years ago”.
Sweden and Russia are also very highly ranked, as they were before the contest, and as they are every year. There’s statistically very little difference between the top three—they’re all well within the margin of error of each other.
Looking to the second semi-final, there are some strong contenders here. Despite the bugfix dragging down their chances of an overall win, Australia are still very much favoured to qualify (95%). Ukraine (91%) and Serbia (84%) are also perennial powerhouses. Despite the weak Scandinavian showing in the first semi-final, we should also expect to see Denmark (71%) and Norway (63%) through.
Also strong here are Georgia (65%) and Poland (65%). Rounding out the set, the best guess is Lithuania (61%), Macedonia (58%) and Albania (54%), but I wouldn’t rule out Israel (52%) or even Ireland (49%).
One country which would have been a very strong contender for qualification here is Romania. They were originally scheduled to perform in this semi-final, but were forced to withdraw about three weeks ago, because of unpaid debts to the EBU. Although it’s sad they’re not competing, it’s interesting to look at who benefits the most from their absence.
Poland are the biggest beneficiaries—both countries have strong diaspora votes, which would usually compete for points. Lithuania are in much the same position, but slightly less so. I’m honestly not sure what Macedonia are getting out of it, except possibly slightly more Balkan votes. For everyone else, it’s mostly just good to have one less competitor.
For the technically minded, I divided by the number of entries, when I should have divided by the number of entries plus one. ↩