This is just copy/pasted from my notes, so I apologize if it's hard to read. Some background reading is here:
https://bitsharestalk.org/index.php?topic=2690.0** perfect information games: the best kind
You can construct a payoff matrix in a way where players will tend agree to admit losses and when they don't the network makes a profit. Massively reduce storage costs.
** perfect information with an RNG: same thing, you need a "trusted seed" which is just H(block||players||timestamp)
** trusted non-collusive hosts, no mucking
The host can reveal what cards other players have but cannot influence the probability of cards you get (infinite deck poker)
4) ** incentivized non-collusive hosts -> "normal" poker possible
hidden RNG state between player actions, host can share information. Might be possible to user clever zero-knowledge proofs to show a host did a fairly generated deck without revealing players' mucked cards. Or mucker could have a chance to rat the host out and players get paid out from hosts' collateral
So the idea is, BTS Lotto takes the most basic "game service" - an RNG - and builds the most basic game you can out of it. Games that we *know* are fun and profitable should follow this model. This will be a "catch-all" DAC that will provide the basic information services you need to host games. So people could figure out a chess engine using this DAC, then spawn a dedicated chess DAC which can embed crazy stuff like ELO and paid tournaments and stuff.
* Brand New Game Types*
If you think about the more general game players are playing when they are connected to a random host (do I trust this host not to cheat, or maybe this host actually deliberately made public), it spurs some thinking about what sort of games could be designed.
I'm actually excited about this because inventing new games is basically incentive structure research and could spawn new DAC ideas.