I briefly skimmed the whitepaper, but it doesn't really have much details on how RNG / shuffling works, which is the most interesting problem here. In base Graphene / Bitshares / Steem there is some RNG used to determine the witness ordering, but some daunting implementation issues [1] prevented us from building a bullet-proof cryptographically strong RNG. Which is a requirement for gambling. This wikipedia page has some information on the theoretical state of decentralized strong RNG.
If peerplays has solved these problems to the level that a production blockchain for online poker -- or even online High Card [2] -- is possible, it's an impressive technical achievement.
[1] Mostly the "last participant being evil" problem -- if the last person to go doesn't like the RNG result, he suddenly develops mysterious network issues and drops out, meaning the protocol has to either hang until he comes back (which may never happen) or somehow recover without him (which effectively causes a re-roll because the private information of the guy who left was necessary to determine what the outcome was).
[2] A simple example game discussed in the whitepaper, where two players are each randomly assigned distinct integers 0-51, and whoever has the higher integer wins.