feed-issuers have the added incentive of developing a reputation: the more often they maintain their feeds to the satisfaction of those who are betting, the more trusted they will become, and hence the more often betters will bet on feeds issued by them.
This betting system seem to rely on trust... That's a problem to me.
We can bet anything, but the feed output is under control of a single one.
Nothing stop him to feed the opposite of the true result of the bet.
Do I miss something?
'Trust' is a misleading word. Consider the situation from a feed-operator's point of view: For the most part, the actual data that a feed operator is publishing will be extremely easy to find outside of Counterparty, and consequently, whether a feed-operator has tampered with the data will be readily verifiable. Users will then stop betting on feeds published by such a feed-operator.
Betting in Counterparty doesn't depend on feed-operators being honest, it depends on them being rational.
Now one could argue that a feed operator can create arbitrarily many addresses, and consequently can continue publishing faulty under new identities. In my opinion, however, this is not a real concern, given how I see Counterparty developing. Namely: one will discover feed-operators outside of the blockchain (either through forum threads or external websites, or whatever), see which public addresses they says are theirs, and bet accordingly. This further increases the utility of reputation within Counterparty, and hence reduces the risk of being scammed.