How did bitshares ever let Roger Ver sweep down and poach this project?
The answer to that is easy: Bitcoin maximalism.
The real question is why aren't we able to convince Augur to use our platform rather than the much slower Ethereum (plus we already have BitUSD working while they are still working on eDollar).
Speaking of prediction markets. The use of BitAssets 2.0 for PM is unsatisfactory in many ways (although it is nice that we get it "for free" so to speak).
First, it is limited to binary outcomes rather than generally to N disjoint outcomes. That is an easy fix. Create a PM pool that let's you deposit X BitUSD to get assets X BitPM-1, X BitPM-2, ... , X BitPM-N, and also require you to burn X BitPM-1, X BitPM-2, ... X BitPM-N simultaneously in order to withdraw X BitUSD from the pool.
Second, I like that we can do traditional order book prediction markets (Ethereum will struggle with that), but a market maker is very important. I believe it should be possible to have both (
is that correct?
yes, should be possible ). So I would love to see BitShares have these BitPM-type assets trading against BitUSD but also have a liquidity sensitive LMSR autonomous agent making those markets on behalf of the PM creator (who would fund the initial liquidity, but would also be able to potentially profit from the trading fees acquired by the market maker). There wouldn't only be market orders, there would also be limit orders. A limit order might match against the LS-LMSR market maker, but could also match against other limit orders by real users sitting in the order book.
Third, using a multisig judge to provide outcomes of prediction markets is an okay start, but eventually we would need something more decentralized. We should learn from the REP token approach used in Augur. In my opinion, we can generalize this a little by allowing a special UIA type that can be used as a REP token to judge certain prediction markets. Some additional features for UIAs would be a prerequisite. First, UIAs should be allowed to elected managers the same way BTS can elect delegates. These managers would of course still have limited powers with what they could do with the UIAs. Second, there should be a built-in mechanism in the blockchain to shift the order of magnitude of the balance to the left or right as necessary to allow the max supply with desired precision to fit in a 64-bit number. This means that the token can be endlessly inflated and there won't be any overflow (instead balances would lose the starting precision of their balances over time, and very small balances would eventually completely disappear by just sitting there). This mechanism is important because redistribution from non-participating or dishonest REP holders to participating and honest REP holders can be simulated through distributing automatically inflated supply to the participating and honest REP holders, no matter if the other REP are locked in a contract or not. With these features in place for UIAs, a prediction market could be set up to choose a particular UIA as the oracle providing the outcomes. If the managers of the UIA (that represent the UIA holders) commit the UIA holders to act as an oracle to that PM (an action that the managers get compensated for from funds provided by the PM creator), then the PM opens up for trading. Eventually, once the PM expires, the UIA holders are obligated to report on the outcome of some selection of PMs that the UIA managers committed them to, and if they don't they won't receive the newly inflated UIA (thus diluting their ownership). For actual details of how this reporting process works (including details like how the consensus outcome is determined and how PCA is used to redistribute the inflated UIA) check out Augur.
These are three projects that I would absolutely vote a credible worker proposal to fund after BitShares 2.0.
 Here is a paper I found on integrating a market scoring rule (e.g. LMSR) with conventional limit orders: http://www.seas.upenn.edu/~hoda/HLPV.pdf