Business Analytics Blog

Business Analytics Blog

Opinions expressed on this blog reflect the writer’s views and not the position of the Capgemini Group

Did crowdsourcing the Euros actually work? By Charles Douse and Tom Heath

Category : Figure It Out

We asked Capgemini colleagues to predict the scores of all the games throughout the group stages of Euro 2016 and over 7200 replied. Now the group stages of the Euros have concluded, we can see how accurate the Capgemini-crowdsourced predictions were.

Can we predict the final standings?

 The final standings in the group stages were calculated using the most probable outcomes from the crowdsourced data. With all the group games now having been played we can examine the difference between the predicted crowd sourcing results against the actual results, as can be seen below.

By comparing the predicted table against the actual table we can quickly see which teams were over- and underestimated in performance according to the crowdsourcing data. Wales and the Republic of Ireland are the teams that exceeded expectations of the crowdsourcing, finishing with 4 points higher than predicted. By far the team to have underperformed the most was Portugal; finishing 6 points lower than predicted by the crowdsourcing.

Taking the most likely results from the crowdsourced data we see that only 4 teams were correctly placed (France & Switzerland from Group A, Germany & Poland from Group C) and 11 out of 16 correctly predicted to go through.

While predicting with 69% accuracy as to which teams will progress may seem a reasonably strong prediction, this is merely on progressing. If we then look at the placement of the teams, this is when the accuracy of the crowd sourcing data truly departs from the actual results.

Looking at whether teams would progress or not, we now look at whether or not the crowdsourcing data can correctly predict where teams will finish in the groups. Relative to the ‘progress or not’ predictions the results are far less accurate when looking at where teams will place; the crowdsourced data only correctly guessed where teams would finish with 17% accuracy.

Can we predict the results of games?

The first 12 results (win/draw/lose) of the competition were correctly predicted by 50.03% of the crowdsourced data. The last round of group games was significantly lower with an average of 25.61% correctly predicting correct results of games. The game which was predicted with the most accuracy was Germany vs Northern Ireland with a whopping 95.03% correctly predicting that Germany would be victorious. The next closest result was Belgium to beat Sweden with 51.45% of the crowdsourced data correctly predicting this followed by Poland drawing with Ukraine at 40.31%. In contrast to this, the lowest predicted outcome was in the Republic of Ireland’s victory over Italy with 3.37% envisaging this unlikely win.

Can we predict the scores in each game?

Now to the scores on the board. With no longer simply predicting a Win/Draw/Lose scenario the accuracy of the crowdsourcing is lower- much lower. The scores of games were correctly predicted a miniscule 2.93% of the time.

The result that fewest people predicted correctly was the thrilling 3-3 draw between Hungary and Portugal at 0.17%, narrowly claiming that title from the almost as equally unlikely 3-0 victory Wales posted over Russia at 0.18%. The easiest game to predict was the 1-0 victory by Poland over Ukraine predicted by 8.8% of the crowdsourcing participants.

The difficulty in predicting individual games scales up to the inaccuracy of predicting the final standings of the group stages.

As the competition advances, it will be interesting to see how the crowdsourced predictions match up (pun intended) to the results we see. You would expect that with the only option for a team is to either progress or be eliminated from the tournament the accuracy of this type of prediction will rise. Teams now have a clearer picture of what is in front of them in the sense that it is a winner from either Team A or B. Unfortunately this will put an end to hours of pub conversation of   “If A beats B then we play C. But if D beats E by 3 or more goals then we play F....” and so forth and so on.

 Scores may be more accurately predicted further into the competition as the margins between teams narrow and they become closer in talent, tactics and class.

However, the magic of the Euros are its shocks, its unpredictability and its underdog stories. To finish on a cliché: It’s anyone’s game.

About the author

Charles Douse
Charles Douse
Charles is a consultant within the Customer Experience and Analytics team at Capgemini Consulting. He has experience working in the Government and Retail sectors and with a background in Mathematics, has a strong interest in using analytical techniques to gain insights from data.

Leave a comment

Your email address will not be published. Required fields are marked *.