Are you sure you're not the one seeing the survivorship bias? Something that is 10% likely to happen ends up switching to the opposite outcome at the very end 1/10 times. There are thousands of prediction markets up at any given time, so there are going to be plenty of examples of unlikely events happening.

But there is plenty of research on how well-calibrated they are. For example, https://polymarket.com/accuracy

Prediction markets, like many other micro-financialization trends, is unhealthy for society. I'm not going to trust research from the very company selling the product. History provides ample examples of how that works without the need to gamble on it.

I would invite you to look into the statistics on foreclosures, bankruptcy, and gambling hotline traffic which compare jurisdictions that have allowed this stuff vs not. Those with demographic breakdowns help to show those most at risk.