Tuesday, June 03, 2008


John Beamer was good enough to reply to my commentary. He makes some good points, but repeats the mistake of the original article when he says:

"If you take a look at the price of the NL east since Saturday the probabilities have changed more or less as you'd expect."

Yes, they have. However, the prices did not change because people were trading futures; they moved because TradeSports deliberately altered the prices at which they would accept a purchase or sale.

To wit, here's a chart tracking Atlanta's share price to win the NL East, over the past week:

Braves chart

Notice that no one traded any Braves contracts from May 31-June 2. Even though the Braves lost three straight games in this span, their share price on the graph remained constant, because no one was buying or selling. (In keeping with the focus of my last post, notice also that their share price moved in response to very little activity early in the week--no 'wisdom of crowds' here, just the wisdom of one individual trader.)

Now, during this losing streak, the price at which you could buy or sell stock in the Braves did change each day in response to the losses--but this is only because TradeSports changed those prices themselves without any input from bettors.

Let's compare two ways of measuring how much people will pay for an ear of corn in a small town. At one end of the spectrum, we have a farmer's market where competition is fierce and customers are willing to haggle to get the best deal. At the polar opposite, we have a local supermarket, where there's no competition and prices are fixed but subject to change. If the supermarket puts corn on sale for half price, does this mean that the value of the corn itself has changed? My guess is that the prices at the farmer's market will provide a much better estimate of how much the locals really value their corn.

TradeSports isn't as one-dimensional as the supermarket in this example--the prices do change in response to trading activity as well as TradeSports's own manipulation--but if we're going to use this data scientifically, we must differentiate between those price changes that result from traders actively buying and selling, and those that simply reflect the opinion of a TradeSports employee. If not, we're giving that employee too much credit.

No comments: