The key to doing statistics well is to make sure you aren’t changing the results with any bias. This means enough samples, a good selection of samples, and weighing the outcome correctly. Even honest polling in pre-election is hard to get right, and because of that it’s easy to make things lean towards results if you want to get certain results, or or getting paid to get those results.

There’s only one poll that matters, and that poll should include as large of a sample as possible, and be counted correctly. Even though some will try to prevent that from happening.