Monday, June 02, 2008

Where Prediction Markets Fail

A lot has been written about the usefulness of prediction markets like TradeSports and InTrade. The idea is that the 'wisdom of crowds' provides a fairly accurate estimate of the probability of a given event, whether it's Barack Obama being elected or the Red Sox winning the World Series.

However, some people give prediction markets way too much respect. Specifically, in this Hardball Times article, John Beamer uses current TradeSports prices to calculate teams' chances of winning their divisions. This is a good idea in theory, but badly flawed in practice.

All markets are not created equal. The share price of a major stock will generally be in line with its value, because millions of shares are being traded each day by thousands of people who each think they're getting the better end of the deal. The trade price is set directly by the traders, rather than a third party.

TradeSports, on the other hand, often sets the prices in the futures markets by themselves, and many a day goes by when no one buys or sells any futures for a given team. If TradeSports says the Red Sox will win their division 55% of the time, but no one is trading Red Sox futures, does this price reflect the public's opinion of Boston's chances, or TradeSports's? The 'wisdom of crowds' is often nothing more than the wisdom of the TradeSports employee who sets the prices.

But wait a minute, you're saying. The fact that no one is buying or selling futures means that the TradeSports price must be accurate, right? Otherwise someone would be getting rich off their mistake.

Don't you believe it. TradeSports limits their risk in these markets by only making a few shares available for trading--typically the most you can potentially win by exploiting their errors is under $300. Furthermore, they build in hefty trading fees, and usually there is a large gap between the 'buy' and 'sell' prices, so the market price needs to be well off target before there's any opportunity for a reasonable profit. A professional gambler isn't going to waste his time looking for a tiny edge when the potential reward is so small.

Many millions of dollars are wagered on a typical Yankees game, but this season TradeSports has averaged about $130 per day in trades on the Yankees to win the AL East. This isn't some obscure Division I-AAA college basketball team--it's the NEW YORK YANKEES. $130/day is a drop in the ocean of the sports betting world, and it tells you nothing useful about the Yankees' chances.

So, TradeSports is setting the prices themselves, and the real wise guys aren't interested in betting them because there's not enough money to be made. That's damning enough, but there's another big problem: because the market is so small, each individual trader has a huge impact on the market price.

In a large market like a major stock exchange, it takes millions of trades to move a stock price significantly. Not so on TradeSports. Suppose someone out there is interested in buying lots of Orioles futures. Maybe he's their biggest fan, or perhaps he just wants to blow $500 and mess with some data. Either way, he starts buying up all the Baltimore futures he can find. $500 is a huge amount of money in this market, so the price of Orioles stock is going to skyrocket.

Meanwhile, what has happened to Baltimore's chances of making the playoffs? Have they improved? Of course not. We know this, but the market price still reflects a real change in Baltimore's outlook.

Think twice before assuming that TradeSports can nail the probabilities better than anyone else. Their market prices provide one opinion of the future and nothing more.

No comments: