people signals


talent leadership change innovation

What kills prediction markets

Recently I have been amazed again at the accuracy of prediction markets – and at their uselessness and irrelevance.

In the last weeks we had a few major decision coming up in Germany. The first one was who will win the soccer Bundesliga; the second who will be Germany’s Next Topmodel. Both competitions were very tight calls and had huge media attention. It was no suprise then that there were numerous polls on who will win these races. In the first case, it has been the closest race in recent Bundesliga history. Two games before the end, there were 3 teams that were in close reach and could all win one of Germany’s most important trophies. In the leading spot was Schalke 04, the team with the biggest budget of the three and close runner-ups in recent years. Second came VFB Stuttgart, the surprise team of the season with young players, relatively small budget and not a cup in 15 years. Third came Werder Bremen, the team that dominated the first half of the season and also the Bundesliga champ in recent years. Close match, open outcome. What would the masses predict?

Germany’s biggest newspaper (Bild Zeitung) opened a poll on who would come in first. As people voted they seemed to favor the underdog: VFB Stuttgart. Two weeks later, Germany had a new national champion – the young, inexperienced players from Stuttgart. Wow! That was a great call – no clarity in that decision and the masses were correct. James Surowiecki must have liked that.

The second show-down came with the casting show of Germany’s Next Topmodel. Heidi Klum took the Tyra-Banks role and eliminated a beautiful ambitious girl each week. The final was made up of three hopefuls: Hana (the dark-haired czech with Angelina-lips and previous model experience), Ani (the blonde who worked in her parents boutique) and Barbara (the redhead studying math in Bavaria). Who would wear the crown at the end of this competition?

Again, Bild set up a poll. To my suprise (and some other bloggers as I have read), the predicted Barbara as the winner. Last weeks show came down then with a big suprise: Barbara won the competition. Another Surowiecki-moment!

Both calls have been sort of odd to me. There was nothing clear in any of those calls. But both times the masses trumped the experts. There were two additional observations though that kind of killed the Surowiecki-glory of those polls:

What lacked in both polls was the number of participants. In the Bundesliga draw there were 2000 people engaged and in the Topmodel vote about 1000. That is almost nothing. The Bundesliga is followed closely by maybe 20% of the populiation (would result in 15 million individuals). .The Bild Zeitung is the biggest daily publication in Germany and is filled with Bundesliga news daily. Of all those people only 2000 voted. I am too lazy to do the math, but it doesn’t strike me as a lot.

The picture fort he Topmodel-competition is similar. In pre-final episodes they had a market share of 25%, which is around 3 million people. Of all those media-savvy young people who blog and youtube about this event, on 1000 cast their vote in prediciting the outcome. These marginal percentages in participation are similar to what I have witnessed with the pilot at our company. At ouf the 125 people signed up for the market, only 4-5 really traded. This is a lowsy participation.

It is all the more striking since the accuracy proves so true. Also at the internal market, the active people were quiet good and the returns were nice. People genuienly like the idea of bottom-up information gathering and no-bullshit predicitons. But then no one participated. If I look on the web at prediction scenarios, they don’t look much better: bizpredict is lame etc. So even though the results are beautiful, they are obviously of no use to the people.

Which brings me to the final killer on the use of prediciton markets: irrelevance. What could Werder Bremen do about being traded as a non-winner? What could Schalke or Hana do about it? In a company we might say that these information can filter in to the correction process or uncover problems early. May be. But so far the prediciton markets don’t have a mechanism to feed up the right ideas to address a loosing trade in the market. And that is a problem. A big one. Transparency only helps if it can trigger some correction actions. And if the masses are not involved in solving the problem then the ball is back in the hand of the few experts.

While Surowiecki seems relevant to our thoughts on how to gather information and our ideology of the positive effect of involving people, the reality shows that it is no easy step to do what matters most in business: being useful and relevant.


Filed under: change, organization, prediction markets