In today's episode of why polls are generally bullshit - a new survey of voters from Harvard-Harris has Donald Trump beating Kamala Harris 48-45, despite yet another egregious oversampling of roughly 25% Democrats which was then 'weighted to the US general adult population' - that still resulted in a Democrat oversample.

This, despite the fact that Gallup has national party identification at +6 Republicans/Republican leaning independents.

Oddly, we couldn't find any polls which oversampled Republicans.

When we pointed out the oversample on X, the 'weighted' crowd suggested that 'this is how any survey works.'

Except, again, it isn't an accurate reflection of US voters whatsoever, and their 'weighting' of such a vast disparity between those polled and reality means we're trusting them to be honest with their black box.

Now, imagine if the same poll was accurately sampled at +6R.

Meanwhile, here's an overview of major polls of Trump vs. Harris via Statista.