Pollsters spent a lot of time figuring out why Donald Trump's win was such a surprise in 2016 — but the reality is that there isn't going to be a radical change in most election polling for 2020.Why it matters: Everyone should be more cautious in 2020 about what the polls can tell us and what they can't. There will be some improvements in state polls, which is what really mattered in 2016. But polling experts warn that state surveys in general are still a weak spot, and other aspects of election polling are still a challenge."The jury's out for 2020. Everyone's smarter after the fact," said Republican pollster Glen Bolger.Context: Most pollsters agree that the national polls weren't wrong in 2016. They showed Clinton ahead by a few percentage points, and she won the popular vote by about 2 percentage points.But, of course, Trump won in the Electoral College by squeezing out victories in the upper Midwest — which you're not going to see in national polls. You need reliable state polls to tell you that. The three main reasons the Trump win was a surprise, according to a postmortem report on the 2016 election polls by a committee of pollsters:Some state polls weren't weighted to get the right mix of educational levels. (They had too many college graduates, who were more likely to support Hillary Clinton.)There was a late break for Trump among voters in Wisconsin, Florida and Pennsylvania in the last week of the campaign.Some people didn't identify themselves as Trump voters until after the election (which could have included some who decided late).What's changed and what hasn't:State polls are more likely to weight their samples for education — but it won't be all of them.It's still hard to predict who will actually vote, and it may be getting harder. "In pre-Trumpian times, one side would surge and the other side wouldn't. Now both sides surge," said Democratic pollster Celinda Lake.State polls are still more likely to be underfunded than national polls. "I don't see an infusion of high-quality state-level polls that weren't there in 2016," said Courtney Kennedy of the Pew Research Center, one of the members of the committee that wrote the postmortem.Voters can still decide at the last minute — and what we don't know yet is whether 2020 polls in the battleground states will run later than they did in 2016.The good news, pollsters point out, is that the 2018 midterm election polling was largely right — especially on control of Congress. And not all pollsters are convinced that there were major problems in 2016, if you knew what to look for.Their main advice for 2020:Pay attention to who did the poll. If you haven't heard of them before, and you don't know if they have a reputation for reliable polling, watch out.Look at the sample size and margin of error. If it's only a few hundred people, the margin of error will be too big. A thousand or more is better. And if it's a subgroup — like Democrats only — it's a smaller group and the margin of error goes way up.They should be transparent about what they're measuring. If you can't see breakdowns by age, gender, race, education, party identification and ideology, "that should be a red flag," said Republican pollster David Winston.Think about who's paying for the poll. If it's a campaign or a group with an agenda, that's a red flag, too.Read multiple polls, not just one. In the last week of the 2016 election, the trend across multiple state polls in Pennsylvania and Michigan was "clearly moving toward Trump," said Joel Benenson, a former pollster for Clinton and Barack Obama.Don't just follow the horse race — look for the reasons why one candidate is gaining or losing."The dynamics about why it's changing, who it's changing with — those measures are much more valuable," said Benenson.Don't use polls for predictions. They're good at showing trends and snapshots of public opinion, but "what [polling] doesn’t do as well is 'predict' who’s going to win/lose," said Republican pollster Neil Newhouse.The bottom line: Will polling be better in 2020? Some of it will be. Is another surprise possible? Definitely.