At the present time (Dec 2014) the BitShares/BitUSD market peg is not being enforced by the prediction market described in the link above. Price feeds are being supplied by delegates. I believe they intend to transition to such a scheme when their system is more mature.
Although the effectiveness of prediction markets is well established, the implementation being proposed by Bitshares has never been tried before. Unlike traditional markets with fixed settlement dates and deliverables, the Bitshares prediction market will run in perpetuity and deliver nothing. This is not without controversy, a sampling of which follows:
I think you are mischaracterizing things. Prediction markets rarely deliver anything.
The novel part was the lack of a price feed. That failed. Now there’s a price feed.
The remaining problems are:
- BitUSD trade at a discount, not at a 1 to 1 peg
- There is no system in place to manage the volume of bitUSD in circulation. Currently volumes are quite low, indicating inadequate demand.
The lack of demand reflects a failure to achieve parity.
On the bitshares forum, there is a discussion of an ad-hoc subsidy to bitUSD yields to address this. But that of course solves nothing. The yield would need to be frequently adjusted to maintain parity.
The bitshares system doesn’t allow for this.
When I was attempting to understand the Bitshares prediction market I was told to think of it as an extrapolation of a traditional futures exchange. The key differences are that there is no threat of commodity delivery (i.e. a real asset changing hands) and no contract settlement date. Yes, I understand that prediction markets aren’t supposed to deliver anything, but a commodity exchange sure does and it does this on a set date.
The self-referencing, circular logic of the Bitshares prediction market proposal is reflected in the comments linked above.
It could be a good research project to do a simulation and examine the behavior of the system under perturbation. The nxt community has conducted simulation of multibranch forging with coq/haskell.