4 ms·

This is the main idea behind prediction markets. http://en.wikipedia.org/wiki/Prediction_market http://en.wikipedia.org/wiki/Prediction_market Prediction mark

This is the main idea behind prediction markets. http://en.wikipedia.org/wiki/Prediction_market http://en.wikipedia.org/wiki/Prediction_market Prediction markets aren't effective because they average over a big crowd; prediction markets are effective because they provide a financial incentive for people who know nothing to bow out of the discussion, leaving only the confident people to form a market consensus.

SilasX 12y ago
I'm not sure it's the same here -- the write-up doesn't mention providing a reward based on correctness. It just says they're weighted based on how little their opinion moves when learning the group's beliefs. EDIT: I stand corrected. As klochner points out, there was a monetary incentive to get it right, so this is like prediction markets. Not sure why the article doesn't mention this critical part. https://news.ycombinator.com/item?id=8032493 https://news.ycombinator.com/item?id=8032493
- x1798DE 12y ago
  I believe the parent is actually saying that the financial incentive is used to select people who know what they are talking about. It's not necessarily a 1:1 match, but basically that selects for people who believe that they know what they are talking about (and thus are less likely to be swayed by the group opinion). To the extent that the current price of a proposition in prediction markets represents the group opinion, you only expect to win the bet if you think you know better than the group opinion.
- john_b 12y ago
  They may not be exactly the same, but a prediction market provides a quantitative way to measure confidence at both an individual and group level without requiring people to revise their estimates at a later time. This is arguably superior to a weighting based on differences in opinion prior to exposure to the consensus of a different group, and after exposure to that consensus.
- codyb 12y ago
  See Klochner's comment. They actually were given incentives it just wasn't written about in the article. https://news.ycombinator.com/item?id=8032493 https://news.ycombinator.com/item?id=8032493
nfc 12y ago
Seems like in this study though there is no incentive for those "who know nothing" to bow out since there is no penalty in guessing wrong. I'd say that the population that constitutes the prediction market should be "confident", while the population that guesses right in this study would be described as having opinions not influenced by the opinion of others (or at least its mean). I guess that the two are very similar but you can be also very confident in an opinion that you have changed based on the opinions of others ;)
Dn_Ab 12y ago
While you're correct that real money markets will do a better job of filtering the less knowledgeable, play money markets are still able to hold their own against real money PMs. There have not been many studies comparing the two, but what few there are find play and real money markets do nearly as well as each other - with caveats. Both fake and real money markets offer better than baseline predictions but real money prediction markets do better when signals are less clear, i.e. no odds markets to serve as a basis. And according to the markets studied in "PM: Does Money Matter", the discrepancy is eliminated when play money markets weight based on confidence (amounts bet this takes past success into account by noting larger bets also reflect wealth which can only be accumulated by past success in play money settings). Even more than money, the key factors in Prediction Market Design are the availability of tools which enable the easier aggregation of information and strong incentives (monetary or not) for being correct. http://aws.iwi.uni-leipzig.de/em/fileadmin/user_upload/doc/Issues/Volume_16/Issue_01/V16I1_Statistical_Tests_of_Real-Money_versus_Play-Money_Prediction_Markets.pdf http://aws.iwi.uni-leipzig.de/em/fileadmin/user_upload/doc/I...

SilasX 12y ago

I'm not sure it's the same here -- the write-up doesn't mention providing a reward based on correctness. It just says they're weighted based on how little their opinion moves when learning the group's beliefs. EDIT: I stand corrected. As klochner points out, there was a monetary incentive to get it right, so this is like prediction markets. Not sure why the article doesn't mention this critical part. https://news.ycombinator.com/item?id=8032493 https://news.ycombinator.com/item?id=8032493

x1798DE 12y ago
I believe the parent is actually saying that the financial incentive is used to select people who know what they are talking about. It's not necessarily a 1:1 match, but basically that selects for people who believe that they know what they are talking about (and thus are less likely to be swayed by the group opinion). To the extent that the current price of a proposition in prediction markets represents the group opinion, you only expect to win the bet if you think you know better than the group opinion.
john_b 12y ago
They may not be exactly the same, but a prediction market provides a quantitative way to measure confidence at both an individual and group level without requiring people to revise their estimates at a later time. This is arguably superior to a weighting based on differences in opinion prior to exposure to the consensus of a different group, and after exposure to that consensus.
codyb 12y ago
See Klochner's comment. They actually were given incentives it just wasn't written about in the article. https://news.ycombinator.com/item?id=8032493 https://news.ycombinator.com/item?id=8032493

nfc 12y ago

Seems like in this study though there is no incentive for those "who know nothing" to bow out since there is no penalty in guessing wrong. I'd say that the population that constitutes the prediction market should be "confident", while the population that guesses right in this study would be described as having opinions not influenced by the opinion of others (or at least its mean). I guess that the two are very similar but you can be also very confident in an opinion that you have changed based on the opinions of others ;)

Dn_Ab 12y ago

While you're correct that real money markets will do a better job of filtering the less knowledgeable, play money markets are still able to hold their own against real money PMs. There have not been many studies comparing the two, but what few there are find play and real money markets do nearly as well as each other - with caveats. Both fake and real money markets offer better than baseline predictions but real money prediction markets do better when signals are less clear, i.e. no odds markets to serve as a basis. And according to the markets studied in "PM: Does Money Matter", the discrepancy is eliminated when play money markets weight based on confidence (amounts bet this takes past success into account by noting larger bets also reflect wealth which can only be accumulated by past success in play money settings). Even more than money, the key factors in Prediction Market Design are the availability of tools which enable the easier aggregation of information and strong incentives (monetary or not) for being correct. http://aws.iwi.uni-leipzig.de/em/fileadmin/user_upload/doc/Issues/Volume_16/Issue_01/V16I1_Statistical_Tests_of_Real-Money_versus_Play-Money_Prediction_Markets.pdf http://aws.iwi.uni-leipzig.de/em/fileadmin/user_upload/doc/I...