David Ernst Chats with the World

Well, I finally found an online computer game that suits my personality: NewsFutures.

NewsFutures is just one of many Prediction Markets available online. They come in a variety of flavors, focusing more on sports, Hollywood gossip, financial news, etc. I’m curious about The Popular Science Predictions Exchange, where you can examine the likelihood of questions such as “Will a blood test for breast cancer be released in the U.S. by the end of March?” (presently 50.75%), “Will the Western U.S. suffer a catastrophic water crisis by 2010?” (66.25%), and that question haunting so many of us: “Will a team of androids beat the human World Cup champs at a game of soccer by 2050?” (48% likely).

Whatever their subject matter, these things all have about the same format. You “buy” (usually using a pretend currency) shares that will pay (usually 100 units) if some future event happens, and will be worthless if the event doesn’t happen as specified. In some elections on some markets, shares will pay out the amount of the percentage popular vote of a particular candidate or party… so, if you “buy” a share in candidate X for 50 units, and that candidate earns 52% of the popular vote, you net 2 units.

Exciting, huh?

Well, yes, to some people, including me. The thing is that from that very simple setup, you see extremely accurate predictions. There’s a big debate amongst academics over whether these markets are actually better at predicting elections than polls are. I’m not in a position to even have an opinion on that question, but considering that polling is extremely complex and expensive, it’s remarkable to me that a simple web site and a bunch of self-selecting participants from around the world can even be close in accuracy, let alone possibly better.

Who doesn’t want to know the future? There are many events on which I certainly do want to know, and the upcoming US presidential election is one. Although, I guess it’s more fair to say I want to have a sense of the relative likelihood. So, this is a big draw to me. It also puts an element of fun into listening to news reports and polling data. So, contrary to so many computer games, this one actually encourages me to pay more attention to the news.

I picked NewsFutures mostly because I liked their mix of solid rational approach and usability for average users. Some sites I found I couldn’t even figure out how to sign up, which I take as some kind of indication that they are there only for some “in crowd”. Also, I didn’t want to bet real money, at least not at first (the Iowa Electronic Markets is actually a legal place to play with a small amount of real money). Finally, NewsFutures focuses on news, broadly defined. I’m much more interested in political news than I am in things like sports scores and Hollywood gossip, but you can buy shares in those things too on NewsFutures.

But, who would spend any amount of time playing such a game for fake money? Well, those of us who think it’s fun, of course. And those of us who want to feel like they have a sense of how viable (or not) certain political candidates (or whatever) are… or at least find out if we really do have a good sense of such things or not.

NewsFutures presently claims 19,323 players. I’m presently ranked number 3,340. This is pretty funny, actually, because I’m definitely not in a position where I’d think I’d be above the 80th percentile of players. Almost all of my 12,419 “exchange dollars” (abbreviated “X$”) were given to me through the sites’ program by which people with net worth less than X$20,000 can just get “free X$”, X$1000-X$1500 per day. So, in other words, over 80% of those 19,000 players don’t even bother to play often enough to get all the free fake money they are entitled to. On the other hand, everyone on their “top 1000” board has over X$30,000, so more than 5% of players have played enough (and well enough) to earn a 50% or greater return on their free money.

By the way, the real leaders have WAY more than the rest of us (the present leader presently has a net worth of X$55,697,715). I’m almost positive that most of these earning are coming from “faster moving” markets than these elections. Sporting events are very popular, including trading on outcomes during games! Imagine, there are probably as many NBA games to bet on this week as there will be elections to bet on throughout all of next year! So, I doubt I’ll be among the real leaders at any point.

But I’m fine with that. I like the game I’m playing. I like playing by a very precise set of rules on a bunch of precisely defined outcomes of interest to me. I’ll be curious to see how my status as an NPR junkie stands up to a test.

My biggest complaint is that I’d like there to be way more things to bet on. We’ve got a lot of very important decisions to make regarding the future. To take one example, environmentalism has largely shifted not to environmental problems of the present but to potential problems of the future. I hear countless people making and denying predictions of future events, but I hear very little accounting for how well people have predicted things. I think it’s a good challenge for the information age to keep track of such things, and, with a bit of clever work, fortune telling might start to become more science than pseudo-science.

I got curious about how the whole Radiohead thing had played out when I heard mention of it on NPR’s Morning Edition this morning. Turns out the band is not presently releasing statistics about it. An independent market research company called comScore, though, released this report on how the downloads have played out. They say that 1.2 million people visited the site, a “significant percentage of visitors ultimately downloading the album” (how’s that for market data?). They were willing to be much more precise with the percentages than the totals, though, saying that about 38% of the people who downloaded the album paid something (40% of US downloaders), and the average voluntary payment was $6 ($8.05 for US downloaders).

Anyway, Radiohead released this statement in response:

“In response to purely speculative figures announced in the press regarding the number of downloads and the price paid for the album, the group’s representatives would like to remind people that… it is impossible for outside organisations to have accurate figures on sales.

However, they can confirm that the figures quoted by the company comScore Inc are wholly inaccurate and in no way reflect definitive market intelligence or, indeed, the true success of the project.”

If by “accurate” they mean “exactly correct”, then I agree. However, comScore is being quite open about their methodology and their approach seems sound to me. Good statisticians can be surprisingly accurate, and the band’s unwillingness to publish the statistics themselves don’t inspire me with confidence that comScore is “wholly inaccurate”. Just how far off does one need to be to count as “wholly inaccurate”, anyway? So, needless to say, I don’t know, but if I had to bet, I’d bet that comScore’s data are not far off the mark (I’ll say +- 5%).

Wired’s blog post on the matter seems to miss the “significant percentage of visitors” part of comScore’s report and thus estimates with faulty reasoning that the band netted $2.7 million from the stunt. Without knowing just what this “significant percentage” is, even approximately, then I do think it’s fair to say that estimates of the total are “purely speculative”. However, I’ll go out on a limb and bet that the band netted over $1 million in the first month of the ordeal. I’ll also stand by my previous prediction that over the course of their experiment they will serve up over a million downloads and gross 2-5 million dollars. That, of course, was and is wild speculation. But comScores numbers are similar. A small but very well-known group of people have access to the real answers, and I challenge them to go public with them. But, they know everyone wants to know, and I suspect they like the attention of keeping the secret. It’s certainly their right to do so, but I think it’s kind of lame.