Play-money prediction markets can be just as accurate, if not more accurate, than their real-money counterparts. As far back as 2003, NewsFutures helped prove this point by running a large scale real-world experiment comparing its play-money NFL football predictions with those of Tradesports. Esteemed scholars David Pennock (Yahoo!) and Justin Wolfers (Wharton) analyzed the data and concluded that both sets of predictions were just as accurate (and both were more accurate than the vast majority of individual “experts”). The resulting peer-reviewed article (here in pdf format) caused quite a stir because of the deeply ingrained real-money bias harbored by most economists.
It is worth revisiting this issue once more in the light of the recent U.S. mid-term elections. Tradesports and NewsFutures were again running comparable prediction markets on the subject. In the comparative chart below, it appears that NewsFutures play-money traders called the Democratic-House outcome earlier and more decisively than Tradesports’ real-money traders. (This snapshot was taken on the morning of Nov 6, on the eve of election day.)
Based on this single observation , we wouldn’t want to claim that NewsFutures’ political predictions are better that Tradesports, but we certainly would conclude that if you’re looking for some serious political predictions, you just cannot dismiss play-money markets out of hand.
NewsFutures operates both play-money and real-money markets (where legal, of course), and we are very aware of the different sets of challenges and rewards that are inherent in each type of implementation. But we’ve also learned that if you’re just looking for accuracy, everything else being equal, money just doesn’t matter.
For companies who want to internalize a prediction market but are not allowed to use a real-money implementation, this is very good news!