Mid-terms only have the most “die hards” go vote typically which usually favors Republicans. Presidential elections get a lot of the “softer” voters which typically lean democrats. I believe the average over the last 4-5 elections in a presidential election is +3-4D but I wouldn’t be surprised if it was 0-+2D this time.
Yes, many pollsters do have a credible turnout model where they make phone calls and force the calls to conform to a D/R/I, young/old, black/hispanic/white model. Yes, they do exist.
But there are those who try to geographically weight their area codes by strictly population, ask a self identification of likeliness of voting and report the raw results, with no screen at all. BUT . . . they may ask the D/R/I question in the call and report it in the results. THAT would be bad were it D+10. I believe Gallup uses no partisan screen at all.