Princeton Researcher on Prediction Markets

[luckybit]

There's another dimension to this. There is a large corpus or research showing that prediction markets work best in terms of prediction if the predictions themselves made by each agent are hidden from one another until the prediction date. Open predictions tend to invoke second order effects such as herd behavior and self-fulfilling prophecies.

The more projects the merrier.

What if we want to find the truth? We cannot find the truth if the predictions are secret. For example if you're using an election result prediction as an example then its possible the election could be rigged and a select few people know who will win in advance. If you're talking about a boxing fight then the match could secretly be fixed.

If there are secrets then a prediction market is less accurate than if its open because all a prediction market is about is who has access to information and it uses market forces to convince people who have access to try to profit from the information they have which can then reveal that perhaps for example the fight was rigged if for instance the prediction was that Mike Tyson would knock out Buster Douglas in the 6th round with a jab. If it's that accurate then we know the fight was fixed.

[ruletheworld]

There's another dimension to this. There is a large corpus or research showing that prediction markets work best in terms of prediction if the predictions themselves made by each agent are hidden from one another until the prediction date. Open predictions tend to invoke second order effects such as herd behavior and self-fulfilling prophecies.

The more projects the merrier.

Interesting point. A prediction market like BitShares/BitAssets is going to be completely open. Perhaps we can design a system for event-prediction markets and keep the 'bids' secret until the completion of the event. The major challenge as I see it is to create and verify the final result of the premise in a decentralized manner. Even something as simple as "Obama will be reelected president of the United States" can be challenging to verify in a decentralized way. Perhaps one way would be a decentralized voting system where somehow each participant has an incentive to vote for the truth.

[itnom]

There's another dimension to this. There is a large corpus or research showing that prediction markets work best in terms of prediction if the predictions themselves made by each agent are hidden from one another until the prediction date. Open predictions tend to invoke second order effects such as herd behavior and self-fulfilling prophecies.

The more projects the merrier.

[ruletheworld]

I am posting under BitShares section as I think it's most relevant, but if mods disagree, feel free to move it out of here.

Ref: https://freedom-to-tinker.com/blog/felten/bitcoin-research-in-princeton-cs/

It's interesting that several academics have taken an interest in prediction markets. This is the second such article I am reading within a week. Apparently, there is a wealth of scientific studies ready to be unleashed - they are just waiting for the right regulatory environment.

The prediction marketplace of BitShares is a little different from the prediction markets described here. These are similar to what InTrade was. Essentially, you buy shares in an 'Event', so say, "Bitcoin will trade above $2000 on MtGox at any moment on Dec 31st 2013". This is a well defined event which may or may not happen. People will 'bet' accordingly. It gets interesting where the predictive power is very good, e.g. Intrade's claim to fame was how it could predict the US election results to within a percentage point.

One main challenge is to this in a decentralized way. There shouldn't be a 'moderator' who will decide whether the said event took place or not. The so called 'Oracles' are a few years away at least, so that doesn't seem like an issue. Not sure how voting would work, i.e. how to incentivize people to vote the right way.

Also, is this something that should be incorporated into BitShares or a new DAC perhaps? Thoughts?