@DevinY1 I'm a statistician with 30+ yrs experience. Poll result vary widely by how well the sample reflect the target behavior (eg voting) of the population (eg likely voters). Typical demographic, socio-economic groups don't explain well in atypical elections. Pollsters determine the sampling groups. They can simply be wrong with best effort, or worse.
@DevinY1 Pollsters have to first figure out which groups best reflect the voting variation in people who actually come out. There are a lot of groups. For example, proportions of millennials voting blue are different by other factors (gender, social, economical, geo ...). The # groupings easily get up to the 1000's. Even if they get the grouping right, polling is expensive. They can only polled very few people in each group. Some pollster use a static panel of people to answer all surveys.
@LadyIDO I wonder if the samples are bad then. Cause I've never been polled and I doubt other millennials get polled cause we don't answer our phones from unknown numbers lol.