This seems somewhat obvious, but it’s worth restating. Polling is not always perfect. It’s right more than it’s wrong, and its notable failures are what makes the news. Polling accurately predicting something is boring and, thus, you don’t hear about it often. It’s not a perfect science, but it’s better than taking a blind shot in the dark.
In 2016, the polls failed to predict the outcome of the presidential race. Or, at least that’s the prevailing narrative. As Nate Silver pointed out (I think this is the link), Trump’s win was actually within the margin of error of most national polls. What failed us were the models that he and most other prominent statisticians put forth.
As the 2018 midterms heat up, a lot of people will dismiss the polls. And perhaps it’s important to take them with a grain of salt! But, it’s worth remembering that they do have some value. To not use polling (even sparingly!) is to make a blind guess. At that point, it’s not strategy; it’s gambling.