Not so good
After Tuesday’s success, it was perhaps inevitable that Thursday’s predictions would be less impressive. Just going by the count (7/10) things don’t look too bad. We predicted that Iceland, Ireland and Malta would qualify, but instead we got Israel, Montenegro and Slovenia.
Unfortunately, looking at the details, things are a bit worse. Israel qualifying is fairly unsurprising, and it was reasonably likely that one of Ireland or Malta would fail, although both is a bit unlucky. The really wrong stuff comes when we look at Iceland, which we had rated as over 90% certain to qualify, and Montenegro, which was only a paltry 18%. Of course, it’s possible that we just got unlucky here, but I’ll be definitely be looking into these results for next year.
The big change for the final is that the model is now predicting a Norwegian victory. Norway had a slightly tougher qualification path than their neighbour Sweden, and this means that the model has increased its estimation of their chances accordingly.
The betting markets are less bullish on Norway however. On Betfair, they’re seventh favourite, but are trading at a fairly long price of 95, for an implied probability of just over 1%. In contrast, the favourites Sweden are trading at 2.92, or 34% chance of winning.
We can also look at which countries are likely to give each other twelve points.
The usual relationships are all in full force. The single most reliable twelve is from Cyprus to Greece (Greece to Cyprus comes third, behind Macedonia to Albania). Belarus to Russia is another safe bet, as is Slovenia to Serbia.
Interestingly, the model predicts that Sweden is likely to gather most of Scandinavia’s twelves, rather than Norway, despite predicting Norway for the overall win. If the model is wrong about Norway, it will be because it overestimated the song quality, not because of voting patterns.
Finally, it’s worth looking at which countries are good predictors of the winner.
Here, we’re looking at the probability that each country gives twelve points to the eventual winner. This may be useful as an early indicator of who’s likely to win. However, there’s a caveat here that the order in which votes are announced is chosen deliberately to maximise tension, so some skepticism about the early scoring is probably warranted.
As usual, Hungary is rated highly by the model, although they have not correctly picked the winner for the last two years, instead picking the second-place finishers. A more interesting choice may be Australia. Although the Ozzie votes have been unofficial, they’ve correctly picked the winner for the last three years. It may turn out that the best predictor of how Europe will vote will come from the other side of the world.