Around what month would you find polls to be more accurate?

[Mr.Phips]

After the conventions, the type of race that we're dealing with will be obvious. The trends within the electorate should be decently clear too.

At the moment, we might have a Lean Trump race but this early is pretty mixed at being predictive-the famous example is how Dukakis was winning comfortably early in the campaign and then the race flipped as voters switched on.

The final margin can still change due to undecideds flipping and/or a polling error.

Up to what month in 1988 was Dukakis winning?

I would hope polling has grown a little more accurate since then...

It wasn't anything to do with the accuracy. Him being up by 17 points and not even coming close to winning has been memorable. This kind of effect has been seen in more modern elections to a lesser degree, McCain 2008 was similar but not as extreme.

Dukakis being up 17 points was an outlier.  Most other polls showed the race much closer at the time.

[cherry mandarin]

If we're still seeing national and/or state polls that range from Biden +5 to Trump +5

I hate to break it to you, but this happens every cycle. When you get as many polls as we're getting, there's bound to be a few outliers.

A major gap between national and state polls

The "gap" can be explained by the fact that there's been a re-alignment of sorts since 2020, which is apparent from looking at the demographic cross-tabs.

No candidate passing 44%

Which is caused by an unusually large share of respondents intending to vote third-party, not an usually high number of undecided voters.