[1andrew1]

Originally Posted by spiderplant

So how many would you poll?

Spoiler: 

A sample size near 2,000 yields a tiny margin of error of ~2.16%, making it much easier to detect small shifts or close races (e.g., a 51% vs 49% split) that a 1,000-person survey (MOE ±3.1%) might miss. It hits the optimal threshold where adding more respondents costs significantly more time and money without offering much added precision

We've had enough of experts, that nice Michael Gove told me so and he supported Brexit so he must be right.Besides, I've never been polled, nor has anyone in my family including Oscar the cat so all those polls must definitely be made up anyway. Nice work if you can get it!