Why have blockchain-based prediction markets not seen greater adoption?

I’m curious if lack of adoption has to do with any of the following:

• Was there no project mature enough to capitalize on the opportunity (Augur to late, Polymarket too early)? 
• Were there technical limitations on the space of possible bets that could be verified with an oracle? 
…

• Do they require a greater amount of utility (other projects leveraging prediction markets, e.g. a lite version of retroactive public goods funding)?

I’m interested in identifying the failures of prediction markets to
• identify limiting criteria/where development should be focused
• to look for any future apps that solve these problems (identify investment opportunity)
• to update my priors —> maybe not every theoretically sound app is commercially viable

A prediction market was the first real crypto application I came across that was easy to intuit, and had obvious benefits for individual users and society at large (people with the most accurate models —> accumulate the most capital —> are best fit to allocate that capital)