Prediction Trading
TradeSports is a brilliant idea. I had thought about making a site of this nature as a startup idea and even tried a very naive implementation of it for my cis422 project (which failed miserably, btw). This was right after I read about Nassim Taleb (thanks to Prof. Hsu's blog). Taleb proposes that being an 'expert' isn't (perhaps shouldn't be) worth that much because even experts get dominated by statistics when it comes to making predictions and do no better and even worse than non-experts. (this is a good talk on the subject). After having listened/read what Taleb has to say, it seems that he is very right. So, what better than a site which allows everyone to trade on predictions? Not only does it allow to test what Taleb says, but it has other possible implications.
First, if the site becomes popular the people running it get a large database of 'experts' (common people) predicting. They can then use Bayesian techniques to make their own predictions and do better than the best of their 'experts'. See this recent entry on Prof. Dave Bacon's blog, for example.
Second, imagine having a track record of all the times you've been right about something. If you are usually right this is the kind of thing you'd like to show to a potential employer or admissions committee. If nothing else it may give you some bragging rights. Such a record would also be desirable from an employer's perspective: in addition to comparing people based on how perfect they say they are compare them on how often they are right. Selling prediction reports could be a viable business plan if the site was popular. Of course, they already make money on each transaction so perhaps this would be unnecessary, but still.
For now, though, TradeSports seems to be a little too focused of sports betting. This probably because that is the easiest to implement/automate and has a large potential user-base already. Also, I am yet to make sure I understand exactly how TradeSports decides the outcomes of events. For things like sports or stock market predictions this is straight forward. For things like current events, not so much. Probably more on this topic as I look into it. It may be a TradeSports' trade secret, though. :-)
First, if the site becomes popular the people running it get a large database of 'experts' (common people) predicting. They can then use Bayesian techniques to make their own predictions and do better than the best of their 'experts'. See this recent entry on Prof. Dave Bacon's blog, for example.
Second, imagine having a track record of all the times you've been right about something. If you are usually right this is the kind of thing you'd like to show to a potential employer or admissions committee. If nothing else it may give you some bragging rights. Such a record would also be desirable from an employer's perspective: in addition to comparing people based on how perfect they say they are compare them on how often they are right. Selling prediction reports could be a viable business plan if the site was popular. Of course, they already make money on each transaction so perhaps this would be unnecessary, but still.
For now, though, TradeSports seems to be a little too focused of sports betting. This probably because that is the easiest to implement/automate and has a large potential user-base already. Also, I am yet to make sure I understand exactly how TradeSports decides the outcomes of events. For things like sports or stock market predictions this is straight forward. For things like current events, not so much. Probably more on this topic as I look into it. It may be a TradeSports' trade secret, though. :-)
posted by rz at 2:36 AM
3 Comments:
TradeSports has a non-sports website called www.intrade.com
"It may be a TradeSports' trade secret, though. :-)"
No secret at all.
Every contract or prediction can only have one clear and public redult.
E.g. will X happen Yes or No.
I did look at intrade.com very briefly. I wasn't sure what the site did at first glance. I'll have another look at some point.
I understand that the predictions are yes/no. What I meant is that having a computer decide without human intervention on what actually happens in the world seems like a slightly non-trivial problem. For things like a sports team winning, or a stock going up or down it seems straight forward enough. For things like current events or politics, less so. How does the computer decide whether Karl Rove was has been asked to resign or whether diebold machines have been found to have security flaws? ie I'm curious about how tradeSports decides whether someone is right or not.
Perhaps I just need to play around with the site for a little while...
Post a Comment
<< Home