It's not only the Presidential race --- let's take a look at the polls in the Tillis/Cunningham race in North Carolina. Tillis beat Cunningham by a good margin --- yet many polls in the recent days had Cunningham up by 5, 6, 8 , or 10 points. Well outside the margin of error. https://projects.fivethirtyeight.com/polls/senate/north-carolina/
I've said before state polls aren't as good as national polls.I have no faith in them now. For presidential elections I'm looking at national polls and approval ratings only.
Many of these polls of the North Carolina Senate race are done by large national polling firms --- in fact they are the ones where the polling is most off. The polls done by actual in-state polling entities located in North Carolina such as East Carolina University seem to be the most accurate (in that they had Cunningham only up by a point or two and the election results were inside the margin of error.
I don't find that to be true. The national pollster firms recent polling of states (the EC is won by state) is once again significantly off from the actual results in the presidential election. Statistically there is no reasonable explanation of why so many of them should be so off -- at this point they really need to straighten out their act to have any credibility.
National approval rating polls told us the last 4 years Trumps approval was below 50% and more Americans disapproved of him than approved,and they were right.That has predicted every presidential election since approval ratings began when there is an incumbent president. Since aggregate polls started national polls have picked the popular vote winner in 5 of the last 5 presidential elections and picked the EC winner in 4 of the last 5 presidential elections. That part of polling is pretty credible imo.