Thursday, 16 May 2013

Bonus Question
Predicting Eurovision finalists: would you beat a monkey?

If this isn't the sort of monkey who would
like Eurovision, I don't know what is
For many, Eurovision remains a one-night stand with all the sequins Europe has to offer. The hardcore, meanwhile, know different. Saturday's final will always be the main course of this annual European feast, but the preceding week with its two semi-finals, serves as a delicious starter (complete with the odd really weird canapé that nobody likes).

What's more, with no British interest to worry about, UK fans can sit back and enjoy the delightfully difficult game of guessing who will make it to the main event. After the first semi-final on Tuesday social media (well, the media I socialize with, at any rate) was awash with people proudly declaring how many of the 10 finalists they correctly predicted, with 7 out of 10 widely considered a 'good' performance. But is it really?

On face value 70% does sound pretty respectable: that's a first class degree at most universities, after all. Reframing the problem, however, can change our perspective. Predicting who will go through is equivalent to predicting who will crash out, and with only 16 countries competing on Tuesday, spotting 7 out of 10 qualifiers is the same as getting just 3 of 6 losers correct. All of a sudden things aren't quite so impressive.

Score Monkey %
10 0.01
9+ 0.8
8+ 9
7+ 39
6+ 78
5+ 97
4+ 100
I thought I'd put my lifetime of statistical training to good use and see how well someone picking at random (a monkey being the traditional example) would do at predicting Eurovision finalists. I've written up the mathematical details over on my stats blog, in case that's your thing, but most of you are probably just interested in the final numbers. The table on the right, then, summarizes how well my hypothetical Eurovision loving monkeys would have fared at predicting Tuesday's success stories. The Monkey % indicates how many monkeys would score at least that well (so you may notice that 100% of monkeys would manage at least 4 out of 10 - the lowest possible score).

Overall, the monkeys would get on surprisingly well. Around 40% of them would, for example, manage that 7 out of 10 many people seemed quite proud of. Even my own performance of 9 out of 10, which I was really rather happy with (despite my copy book being blotted by the Netherlands, of all things), only puts me in the top 1% of monkeys. Not so impressive after all.

Score Monkey %
10 0.005
9+ 0.4
8+ 5
7+ 27
6+ 65
5+ 92
4+ 99
3+ 100
The semi-final fun isn't over, of course, so what would be a good score for Thursdays's second heat? With 17 countries to consider, things are slightly different, and this table summarizes our updated monkey performances.

Unsurprisingly, with an extra country in contention, prediction gets a *lot* harder. As such, I think 7 out of 10, placing you in the top quarter of monkeys, could be considered a reasonable performance. 8 out of 10 would get a statistician interested, while 9 or 10 would suggest you should spend less time reading this and more time at the bookies (or maybe just outdoors). Or you could just sit back, relax, and succumb to Eurosong Fever.

No comments:

Post a Comment