Thursday, May 22, 2008

Prediction market naysayers should learn from the recent NBA Draft

In the NBA Draft, 9 teams are selected to participate in a "lottery" to see who gets the first 9 picks. The 9 teams selected to participate are the ones that ended the season with the 9 worst records. The team with the worst record (Miami in this case) is given the highest probability to win the lottery (done with ping pong balls) and the team with the 9th worst record has the lowest probability of winning.

In this year's draft, the team with the 9th worst record was the Chicago Bulls. Going in to the lottery process, the Bulls were given less than a 2% chance of winning the lottery. A couple days ago, representatives from the 9 teams gathered to participate in the lottery process and low and behold, Chicago "beat the odds" and won the lottery. They now have the first pick in the draft coming up next month.

This improbable event is an excellent lesson for those who write off prediction markets as being "wrong." To say this, as a recent article in Wired did, about some political markets, shows a fundamental misunderstanding of the simple lessons of probability that are at play in a prediction market and were illustrated so nicely in the NBA draft.

In a prediction market, unless an answer is worth $0, we can never say the crowd thinks that answer CANNOT happen, only that there is a low probability of it happening. Similarly even if a stock is worth $95 (95% chance of answer happening) we cannot say that answer will occur for sure because there is still a 5% chance other answers could occur.

If you really want to judge prediction markets on their merits, you have to look at the number of times an answer occurred according to the probabilities assigned by the crowd. For example, for answers that ended with a price of $10, that answer should have only occurred 10% of the time. For answers that ended with a price of $90, that answer should have occurred 90% of the time. Only by looking at many markets in aggregate can you say whether your prediction marketplace is "accurate" or not.

This is also why we advise our clients that prediction market results should serve as inputs to decision making. The results from a market are only providing a range of probabilities of various answers occurring versus serving as a crystal ball to definitively predict the future.

Nate recently presented at a CFO's conference here in Chicago put on by the Corporate Executive Board. In preparing for his presentation he created some nifty graphs showing how Inkling has done across all its markets. He's going to post those in the next few days and you'll be able to see Inkling's comprehensive market performance.

No comments: