If yes, why? If no, why not?
Alex Forshaw commenting here at Midas Oracle:
And notwithstanding my presumed unfamiliarity with prediction markets literature, if Mr Servan-Schreiber were as confident about his exchange’s information integrity as his marketing suggests, he wouldn’t have passed up the obvious arbitrage potential shown by that graph. He could have made hundreds of thousands, potentially millions of dollars in very little time, just on Tradesports. (Not to mention options and derivatives of companies and sectors that profit from Republican subsidies, vs. those that profit from Dem subsidies.)
Addendum: Professor Justin Wolfers (Wharton) has posted a comment…
I actually offered Emile Servan-Schreiber precisely the bet that Alex Forshaw suggests: I bet him a bottle of Aussie wine against his bottle of French wine that Tradesports would do better than NewsFutures in predicting NFL outcomes. (This was the only reasonably large sample of comparable events that we could find.) And to his credit, Emile believed enough in his mechanism to take the bet.
The data we collected to settle this bet ended up as a paper (â€Does money matter?…â€). [CFM: paper PDF]
And the moral of the story? I lost a bottle of good Aussie red to Emile as NewsFutures outpredicted Tradesports. While this isn’t necessarily persuasive that play-money markets do better than real-money markets (samples are still small), it does at least suggest that the difference in forecasting ability is small.
Addendum 2: Jason Ruspini has posted a comment…
Of course arbitrage per se isn’t possible, but I assume that was part of the point.
Emile wrote that play-money PMs can be as accurate as real-money ones “everything else being equalâ€.. meaning that traders will produce similar results if they have similar information and motivation. The motivation is the trick. There are other ways to motivate traders aside from money, though money is probably the most reliable and universal in this context. What motivates people to play games in general (much less wait in line for 2 days to buy an incrementally better game system)? What motivates people to blog? What motivates people to make money once they have more than say $100 million? Reputation. Fun.
In some cases, real-money markets can also suffer from distortions that may not affect play-money markets. More on this soon.. not enough time in the day.
I actually offered Emile Servan-Schreiber precisely the bet that Alex Forshaw suggests: I bet him a bottle of Aussie wine against his bottle of French wine that Tradesports would do better than NewsFutures in predicting NFL outcomes. (This was the only reasonably large sample of comparable events that we could find.) And to his credit, Emile believed enough in his mechanism to take the bet.
The data we collected to settle this bet ended up as a paper (“Does money matter?…”).
And the moral of the story? I lost a bottle of good Aussie red to Emile as NewsFutures outpredicted Tradesports. While this isn’t necessarily persuasive that play-money markets do better than real-money markets (samples are still small), it does at least suggest that the difference in forecasting ability is small.
Of course arbitrage per se isn’t possible, but I assume that was part of the point.
Emile wrote that play-money PMs can be as accurate as real-money ones “everything else being equal”.. meaning that traders will produce similar results if they have similar information and motivation. The motivation is the trick. There are other ways to motivate traders aside from money, though money is probably the most reliable and universal in this context. What motivates people to play games in general (much less wait in line for 2 days to buy an incrementally better game system)? What motivates people to blog? What motivates people to make money once they have more than say $100 million? Reputation. Fun.
In some cases, real-money markets can also suffer from distortions that may not affect play-money markets. More on this soon.. not enough time in the day.
Prof Wolfers–that’s fascinating…couldn’t access the paper, but I imagine it would have to do with, people need proportionally more information to admit that they’re wrong (bail on a position) the more heavily they are staked in a given outcome. (Props to Mr Servan-Schreiber, gotta give him his due)
I would have argued that, judging from the TS forum, the TS trading population is significantly more libertarian-Republican than the population at large, so their perceptions would be skewed in favor of the GOP. (I have also believed, judging from exit poll skewing, that Republicans are significantly more introverted than Democrats, so there are more “quiet Republicans” out there who can’t easily be sampled than there are quiet non-participatory (in the poll) Democrats who can’t easily be sampled, and I would think that would apply to a population of traders as much as it would to any other population.) Did that make any sense? If it does, you’ll understand that given the apparent domination of the forum by Republicans, the actual sample of Republican v. Democrat traders was even more lopsided than that.
Personally, I thought it was screamingly obvious for as long as I was blogging (and you can check the archives) that the Republicans were certain to lose the House, and it seemed truly irrational to me that TS was hovering at 35% a week out from the election (going from memory there).
I would have thought that the biased TS trader population would have dealt conclusively with that problem…but the NFL thing really throws a monkey wrench into that hypothesis. Hmm…..what is Mr Servan-Schreiber’s trader sample size compared to TS’? Anyone know?
Perhaps groups that have specific brand attachment, i.e. the overwhelming majority of people trading the security side with one of the brands being traded, is somehow amplified on real-money exchanges…I dunno.
PDF of Wolfers / Pennock / EJSS paper:
http://www.newsfutures.com/pdf/Does_money_matter.pdf
I scanned through it. 135/208 vs 139/208.
Interesting. But the NFL data isn’t at all diffuse…pretty much all the data is readily available, no? And the actual playing out of the event is also extremely clear. The same definitely doesn’t go for an election, where costs of aggregating much information pre-event and in-process are both much higher. Many NFL traders “aggregate,” to the extent that there’s any aggregating left to do, in their free time. I am not sure to what extent aggregating took place for the elections trading; the founding purpose of my blog was to preach that TS traders’ elections futures aggregation left a lot to be desired, and was significantly inaccurate. And so it was.
Also, I wonder how other NewsFutures elections contracts performed. The NF House contract was extremely volatile–its “prediction” could be completely different the next day, in a way that TS’ wasn’t. While the mean NF price was certainly closer to the actual outcome, I think a lot of NF’s earlier accuracy (with *way* more volatility the whole way through) is much more attributable to evident biases in the TS trader population.
Color me skeptical in the extreme, in spite of the historical record.