alt Hacker NewsThe key is avoiding the bets with controlling insiders, i.e. those that could have a potential conflict of interest. Even something as banal as weather data has some insider knowledge, but an insider has no practical control over it, i.e. the insider is non-controlling, with no conflict-of-interest.
Weather data in prediction markets can definitely be gamed. One example that exists in real prediction markets is that the contract specifies a single source as the source of truth. But that source rounds data during unit conversion twice (F -> C -> F), meaning there’s an unequal probability distribution, and some numbers have a 0% chance of winning.