What are the differences?
There are a lot of different manner to gather polls and create averages and as you can see, there are a lot of variances in what different pollsters come up with. At this point, these various polling averages are anywhere from Trump up by a small amount to Harris being up by over three points.
My crosstab model has been explained. It basically uses only the crosstab averages of recent polls (up to 23 polls right now) to create one super poll. It ignores toplines and the ability of pollsters to skewer the results by over sampling or under sampling.
My personal poll average is a weighted average of all pollsters I track (49 polls currently). It does weigh newer polls heavier than older polls but does not ignore an older poll until that pollster replaces that poll with a new one. Because of this I still have some polls in my average from right around the time Biden dropped out.
Real Clear Politics chooses only specific pollsters, ignoring many of the more obscure polls that others include in their average. It is a pure average of the polls on their list without any weighing. Real Clear Politics drops off polls after a period of time and then basically gives no weight to them. They also specifically separate the head-to-head polling from the polls that include other candidates and have two averages.
538 uses a system where they weight pretty much every pollster and even assign them bias that they can "correct for". They generally use most every poll, with the exception of a few whom they consider too biased toward the GOP (e.g. Rasmussen). They include both head-to-head and polling including third party together for their calculations. From what I can tell Nate Silver is using a very similar formula (as he created the original 538 design), but also incorporates state polling into his national polling averages, which I am not sure 538 does.
The Hill seems to be a hybrid. They have some sort of weighing, but nothing like 538 or Silver. They use most every poll, including Rasmussen and others ignored by 538 and Silver. Like RCP, they also supply two separate polling averages for head-to-head and those pollsters who include RFK and others.
Lastly, for reference, I simply use all seven different averages to create a grand total average. This, in theory would even out the different manners of how these polling averages are created.
What does it all mean? Well... if you put a gun to my head, I would suggest that the true race is right around where my crosstab model is showing which is only 0.31 off from the overall average of the seven averages being tracked. In other words, despite all of the recent hype, Harris is probably leading right now by around a point.
unbelievable, but thanks