I think this is pretty fascinating. I follow a couple of guys who are in the business, and able to analyze polls by looking at their cross tabs. Apparently, a poll of young people just released shows Biden only leading Trump by 4 points. But the raw results weren’t that, they collect raw results, evaluate their sample, and then “weight” it to correct for sampling errors. (I think I have this more or less correct - happy to be corrected if someone knows better). Evidently this poll released the unweighted results accidentally and then released the weighted results later. Pollers won’t always provide that information, so it is nice to get that.