Attention Web World Cup: Weekend 1 Update


Last week was the start of the World Cup, which meant the kickoff of Chartbeat’s Attention Web World Cup. We’re just 11 matches in and we’ve already seen some pretty awesome games. Some of my favorites include: The Netherlands soundly defeating the defending champs; Costa Rica surprising Uruguay with a 3-1 upset; and Switzerland scoring in the 93rd minute to defeat Ecuador.

But, for those of you who were disappointed in the performance of your teams over the weekend, here’s your chance for redemption. Below are the scores for how the teams fared this weekend in the Attention Web World Cup, and they are quite different than the outcome of the “real” World Cup.

Engagement between countries is very similar … this is truly anyone’s cup!

(Winning score highlighted  in green, draws in yellow.)




Wait, how does a draw work in the AWWC?

Many of you will notice that in some games a two-second differential, for example, will result in a win for one of the teams, yet in another game, a two-second differential will result in a draw. Take, for example, the Cote d’Ivoire/Japan matchup. Japan had a median Engaged Time of 26.0 seconds, and Cote d’Ivoire had only 20.0 seconds. A six-second differential, but we had a draw? What’s with that?

As I said in the last post, I determine the winner in a statistical manner. Over the course of the game, I sample Engaged Time for users from each country for the top 20 articles on each of Chartbeat’s sites. This results in a distribution of times for each team. To determine a winner, I ask, statistically, whether these two distributions are different. In other words, I try to determine that if I had a large enough sample of Engaged Times for each country, would it turn out that one country consistently had a larger median Engaged Time? The problem—and this is a fundamental concept in statistics—is that the size of a sample is directly related to the precision with which you can judge your statistic of interest. In our case, this amounts to the fact that the more data we have, the narrower the margin can be for us to determine a winner.

And here’s the rub: For countries like Cote d’Ivoire and Japan, we didn’t have many samples to look at. With these distributions, there is too much variability in the data for us to precisely determine whether the 26-second median we measured for Japan is, in actuality, truly larger than the 20-second median we measured for Cote d’Ivoire. We just can’t know if Japan had such a large median only because of the particular sample we drew in comparison to Cote d’Ivoire’s sample.

In this way, the Attention Web World Cup is quite democratic. Those countries whose web presence across our sites isn’t very large don’t automatically get relegated to the bottom of the heap, they have a good chance at getting 1 point through a draw.

Keep checking back for updates and tweet about your favorites using #AWWC.

Boa Sorte e Divirta-se!

More in Research