Looking at current 2020 presidential polling it's more and more obvious to me that I'll probably have to change my methods for forecasting the state-by-state vote percentages and electoral college votes.

Some of this is based on what @Peoples_Pundit and @Barnes_Law have said...

...about poll construction and polls sometimes being held back if not favorable to a certain candidate.

Since I don't have my own polling operation or the focus/patience to dig deep into demographics my method has been to evaluate/correct poll bias.

This was my 2016 Electoral College Final Projection.

Here is a blog post (blog no longer exists) outlining my methods for projecting the popular and electoral college in 2016:

As noted in the above blog post for my 2016 presidential forecast:

"State-by-State Vote Percentage Projections: Correct 42 (82.35%), Incorrect 5 (9.8%), Toss-Up 4 (7.85%);
Electoral College Vote Projections: Correct 45 (88.24%), Incorrect 4 (7.85%), Toss-Up 2 (3.91%)"

Any my Electoral College Map Note:

Turning back to the 2020 Presidential Campaign:

*The polling is very thin in some states
* The Margin of Error (MOE) in too many polls is ridiculously wide
*The inherent bias in polls looks as bad or worse in 2020 as it was in 2016

My method of forecasting the polls in 2016 relied on the overarching idea that the sum of the recent polls could be adjusted to accurately reflect the state of the race.

My feeling this time around is that there aren't enough polls of sufficient quality to be able to repeat closely what I did in 2016.

@Barnes_Law has pointed to watching trends over time and I have some good ideas on how to couple this with my other techniques.

I think both @Peoples_Pundit and @Barnes_Law have pointed to the importance of 3rd party candidates in 2020. You can't really come up with a good forecast without considering 3rd party/undecideds in every state.

This is...

...explained earlier in the blog post I attached titled "Trumping the Pollsters..."

There is a herding effect near the end of campaigns, but you have to know the size of that herd (100% - Trump% - Biden% - 3rd party% = Eligible Herd)

Knowing who is left (the herd) and which way most of that herd will go takes away some of the "too close to call" states. Undecideds disproportionately swing one direction in a known (to me) ratio.

There are some other "tricks" in my bag that I believe will help me fill in some of the blanks left by the sparse and suspect poll quality in some states.

If I had access to good polling data I would be exceedingly dangerous and accurate.