Jump to content
270soft Forum

Poll of Polls Forecasting Model Useful for PI/PMI


Recommended Posts

Hello,
This is my UK prediction with today projection and near future (1 month) projection if the actual trends detected by model keep happening.

Post image

I make poll of polls as the title says, using reference data from last 32 official polls of every aproached nation, as seen in websites like Wikipedia, after I use a mathematical model to determine the predicted results. I made retro tests with past elections, and I improved the prediction model very much. Now it's very efficient, I can give some examples:

My model vs. Real Clear Politics in past USA Elections since 2004:

Bush 2004: 49.0% vs. 48.9%
Kerry 2004: 47.8% vs. 47.4%

Obama 2008: 51.9% vs. 52.1%
McCain 2008: 45.0% vs. 44.5%

Obama 2012: 48.5% vs. 48.8%
Romney 2012: 47.7% vs. 48.1%

Trump 2016: 44.1% vs. 43.6%
Clinton 2016: 46.5% vs. 46.8%
My Total Error (REPS+DEMS) vs. RCP aggregate polls error:

Post image

I did better than RCP in every election.
Portuguese retro predictions for 2019 with the actual model, 0.28% mean error in every party:

Post image

I can post here United Kingdom and USA predictions every month that can be useful for scenarios. 
My Political personal Twitter is also in Portuguese, but I can do another Twitter page in english and share here if you want and show interest, anyway it's easy to understand the results by the party names and values in %.

Link to post
Share on other sites
1 hour ago, vcczar said:

@Entrecampos I'm curious what your model things the popular vote is for Biden/Trump in 2020. 

Image
I have extrapolation model for each state, but i'm doing improvements, so I did only put the sure states. Next week I will put again with all states colorful.

Link to post
Share on other sites
54 minutes ago, Entrecampos said:

Image
I have extrapolation model for each state, but i'm doing improvements, so I did only put the sure states. Next week I will put again with all states colorful.

Interesting. The PV gap is almost double that of Clinton vs. Trump. 

Link to post
Share on other sites
3 hours ago, Entrecampos said:

Hello,
This is my UK prediction with today projection and near future (1 month) projection if the actual trends detected by model keep happening.

Post image

I make poll of polls as the title says, using reference data from last 32 official polls of every aproached nation, as seen in websites like Wikipedia, after I use a mathematical model to determine the predicted results. I made retro tests with past elections, and I improved the prediction model very much. Now it's very efficient, I can give some examples:

My model vs. Real Clear Politics in past USA Elections since 2004:

Bush 2004: 49.0% vs. 48.9%
Kerry 2004: 47.8% vs. 47.4%

Obama 2008: 51.9% vs. 52.1%
McCain 2008: 45.0% vs. 44.5%

Obama 2012: 48.5% vs. 48.8%
Romney 2012: 47.7% vs. 48.1%

Trump 2016: 44.1% vs. 43.6%
Clinton 2016: 46.5% vs. 46.8%
My Total Error (REPS+DEMS) vs. RCP aggregate polls error:

Post image

I did better than RCP in every election.
Portuguese retro predictions for 2019 with the actual model, 0.28% mean error in every party:

Post image

I can post here United Kingdom and USA predictions every month that can be useful for scenarios. 
My Political personal Twitter is also in Portuguese, but I can do another Twitter page in english and share here if you want and show interest, anyway it's easy to understand the results by the party names and values in %.

You said "every approached nation," (assumably in reference to the PI and PMI elections stated in your title scheme, and then in comparison to your own Portugal project), but there's no Canadian "poll-of-polls," listed, despite Canada being a nation fully approached in a PMI game.

Link to post
Share on other sites
  • 4 months later...
My actual prediction, believing that Democrats will have 49.4% and Republicans 43.2%, if the national polls are coherent, Biden has 72.6% of being elected and Trump 27.4% of being reelected.  I know that the candidates difference is huge, but my model gives lower percentages than what analysts expect for Trump in states like California (28%) or New York (32%). By other way, seems that polls are overpowering Biden in Rust Belt. Trump only needs to conquer Minnesota and North Carolina to be reelected. About Florida, colors mean possibility of victory not advantage, Florida will be close, but model says that is very probable to be democrat.
A imagem pode conter: texto que diz "Democrats 288 209 43 52 250 10 62 WA 12 Republicans 126 OR MN 10 WY SD NV 6 55 UT NY Co IL 20 IN MO 10 AZ 11 WV NM MA OK TN AR RI CT 14 NJ DE TX 38 GA 16 MD DC 270 EWIN"
 
 
 
 
 
Link to post
Share on other sites
18 minutes ago, Entrecampos said:
My actual prediction, believing that Democrats will have 49.4% and Republicans 43.2%, if the national polls are coherent, Biden has 72.6% of being elected and Trump 27.4% of being reelected.  I know that the candidates difference is huge, but my model gives lower percentages than what analysts expect for Trump in states like California (28%) or New York (32%). By other way, seems that polls are overpowering Biden in Rust Belt. Trump only needs to conquer Minnesota and North Carolina to be reelected. About Florida, colors mean possibility of victory not advantage, Florida will be close, but model says that is very probable to be democrat.
A imagem pode conter: texto que diz "Democrats 288 209 43 52 250 10 62 WA 12 Republicans 126 OR MN 10 WY SD NV 6 55 UT NY Co IL 20 IN MO 10 AZ 11 WV NM MA OK TN AR RI CT 14 NJ DE TX 38 GA 16 MD DC 270 EWIN"
 
 
 
 
 

This would be quite an odd map. FL and NC have consistently polls about 1 to 2 pts for Biden, which is close enough that I could see it going Red. Biden is polling landslides in MN, WI, MI, and has a decisive lead in PA, a state where Biden was born. While, this map could happen, it makes me doubt the model you've made/use.

I'm just having a hard time picturing the scenario that would create Trump the victor in WI, MI, and PA, but not in FL, AZ, and NC. It seems from all evidence that for Biden to get FL, AZ, and NC, he'd have to be so strongly supported that he also gets WI, MI, and PA, states much more friendly to Democrats. 

Link to post
Share on other sites
26 minutes ago, vcczar said:

This would be quite an odd map. FL and NC have consistently polls about 1 to 2 pts for Biden, which is close enough that I could see it going Red. Biden is polling landslides in MN, WI, MI, and has a decisive lead in PA, a state where Biden was born. While, this map could happen, it makes me doubt the model you've made/use.

I'm just having a hard time picturing the scenario that would create Trump the victor in WI, MI, and PA, but not in FL, AZ, and NC. It seems from all evidence that for Biden to get FL, AZ, and NC, he'd have to be so strongly supported that he also gets WI, MI, and PA, states much more friendly to Democrats. 

2016 was an odd map, do you remember? Who would expect Trump winning in Rust Belt? Why cant be odd again?
My model can be used in retrospective, with national polls, without knowing state results, in 2016 this model falls only 3 states, in other elections from before is equally effective. Florida is a state with close results, my model says that dems have 60% to 75% of winning (50 to 55, 55 to 60, 60 to 75, 75 to 100 . This doesnt mean that they have more than 2% advantage in the model. 
I can send retro tests maps. The prediction follows a very simple theory, I start with the anterior election results and model with analysis of 10 past elections to detect tendencies and adapting national polls vs past results.  This is the map of 2016.

unknown.png

I have all results from every state in an Excel Spreadsheet that i use to calculate.

image.thumb.png.22dcdcdbf20d2acf461460d556d65c89.png

Link to post
Share on other sites
10 minutes ago, Entrecampos said:

2016 was an odd map, do you remember? Who would expect Trump winning in Rust Belt? Why cant be odd again?
My model can be used in retrospective, with national polls, without knowing state results, in 2016 this model falls only 3 states, in other elections from before is equally effective. Florida is a state with close results, my model says that dems have 60% to 75% of winning (50 to 55, 55 to 60, 60 to 75, 75 to 100 . This doesnt mean that they have more than 2% advantage in the model. 
I can send retro tests maps. The prediction follows a very simple theory, I start with the anterior election results and model with analysis of 10 past elections to detect tendencies and adapting national polls vs past results.  This is the map of 2016.

unknown.png

I have all results from every state in an Excel Spreadsheet that i use to calculate.

image.thumb.png.22dcdcdbf20d2acf461460d556d65c89.png

Where are the 2020 polls in the excel sheet. It looks like you’re using just historical results which won’t really tell you enough in understand 2020. In regards to the odd 2016 map, it was in a much higher realm of possibility. Trump won by larger margins in NC, AZ, and FL to win narrowly in MI, WI, and PA. That makes sense. Your map, and I don’t doubt your intelligence and your passion, doesn’t make as much sense. 
 

 

Link to post
Share on other sites
29 minutes ago, vcczar said:

Where are the 2020 polls in the excel sheet. It looks like you’re using just historical results which won’t really tell you enough in understand 2020. In regards to the odd 2016 map, it was in a much higher realm of possibility. Trump won by larger margins in NC, AZ, and FL to win narrowly in MI, WI, and PA. That makes sense. Your map, and I don’t doubt your intelligence and your passion, doesn’t make as much sense. 
 

 

I use 2020 polls as observation tool, Im already developing a model that compares polls with the historical model but it isnt ready for now. Also an alternative model with counties, that try to predict the counties of a swing state and try to adapt with demographic data like population growth (its only an idea, i didnt the alternative for now).

 

29 minutes ago, vcczar said:

Where are the 2020 polls in the excel sheet. It looks like you’re using just historical results which won’t really tell you enough in understand 2020. In regards to the odd 2016 map, it was in a much higher realm of possibility. Trump won by larger margins in NC, AZ, and FL to win narrowly in MI, WI, and PA. That makes sense. Your map, and I don’t doubt your intelligence and your passion, doesn’t make as much sense. 
 

 

What you said is true, but nobody was expecting a narrow victory of Trump in MI, WI and PA. Wisconsin had 0 polls with GOP ahead!!! Virginia was much more GOP or Nevada or Colorado, than Wisconsin in polls, I could see many people saying the same adapted phrase in 2016...

Obama won by larger margins in MI, WI, and PA to win narrowly in VI, NV, and CO.

General election polls 2016 Clinton v Trump.svg

Margins of error in 2016 polls.

Link to post
Share on other sites
14 minutes ago, Entrecampos said:

I use 2020 polls as observation tool, Im already developing a model that compares polls with the historical model but it isnt ready for now. Also an alternative model with counties, that try to predict the counties of a swing state and try to adapt with demographic data like population growth (its only an idea, i didnt the alternative for now).

 

What you said is true, but nobody was expecting a narrow victory of Trump in MI, WI and PA. Wisconsin had 0 polls with GOP ahead!!! Virginia was much more GOP or Nevada or Colorado, than Wisconsin in polls, I could see many people saying the same adapted phrase in 2016...

Obama won by larger margins in MI, WI, and PA to win narrowly in VI, NV, and CO.

General election polls 2016 Clinton v Trump.svg

Margins of error in 2016 polls.

Are you talking about 2004 to 2008? That would make since regarding those narrow margins etc. 

The 2016-2020 polling margin sustains the view that Biden would win WI, MI, and PA before winning FL, AZ, and NC. 

i have my own map in my “State of the Race” threads. See day 33. I’ve been keeping track of the polling since the 100 day period. 

  • Upvote 1
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...