I found an error on the sheets that actually potentially ruined everything. It was a single letter I mis-typed, but as a result, the NDP's ratio total needed for vote calculation was actually being pulled from the PC Party; which is why the AFF was so high.
When I fix it
https://docs.google.com/spreadsheets/d/1nRdNOdXUtW0F0hKHcNXnNwzveTxeYOxcmgoBo7bG-78/edit?usp=sharing
I end up with more errors however, 20.
The last time a projection went this disastrously for me was also the last time I tinkered with the sheets as much as I did (Quebec in 2012)
Friday, June 8, 2018
The math was smarter than me
Still going through my check of "what went wrong" and one thing I've already discovered is the pure unadulterated math is smarter than me.
The pure math, with 0 adjustments (such as upping ford's vote in his riding, or increasing the liberal vote in the few ridings they could have won) had 16 errors.
https://docs.google.com/spreadsheets/d/1N_L5r7hxrKGT0zPhPT1dieVUbGgf8BsllU2wCYCY8HQ/edit?usp=sharing
whereas the adjusted math, had 19
https://docs.google.com/spreadsheets/d/1w07nUt6m8aw1KL0QrMv72jGHeTn2VLvPOUnUbeZJcjU/edit?usp=sharing
In short, I am going to be quite a bit more conservative with any fancy formulas I develop in the future to make the sheets "better" as clearly the pure math itself did a far better job than I did with this election.
The pure math, with 0 adjustments (such as upping ford's vote in his riding, or increasing the liberal vote in the few ridings they could have won) had 16 errors.
https://docs.google.com/spreadsheets/d/1N_L5r7hxrKGT0zPhPT1dieVUbGgf8BsllU2wCYCY8HQ/edit?usp=sharing
whereas the adjusted math, had 19
https://docs.google.com/spreadsheets/d/1w07nUt6m8aw1KL0QrMv72jGHeTn2VLvPOUnUbeZJcjU/edit?usp=sharing
In short, I am going to be quite a bit more conservative with any fancy formulas I develop in the future to make the sheets "better" as clearly the pure math itself did a far better job than I did with this election.
Post Mortem
As usual lets start with the ridings I was wrong in
Which as you can see is quite a lot of them.
Additionally even with the proper vote shares, the sheets were off by a massive margin, mostly due to that simple yet complex math error I don't know how to correct. I will make a post later explaining the problem and seeking for help to fix it. When you correct for this math error (which I've done here) the sheets are actually fairly accurate.
The biggest problem, however, was not the sheets, but rather, that I went with my gut and not the math.
My gut does not have a good history in recent elections. To correct for this, I will be focusing much more on the math going forward.
Additionally, the riding polls from Mainstreet ended up being rather accurate, and, with the purchase formula outlined here, I will be buying them when possible.
In the end it's been a terrible night for me, both on a "who I wanted to win" level and a "made bad projections" level; but I will continue on. There are always more elections to look at around the world, and I'll be starting that within a few days by going over recent results in Slovenia.
Which as you can see is quite a lot of them.
Additionally even with the proper vote shares, the sheets were off by a massive margin, mostly due to that simple yet complex math error I don't know how to correct. I will make a post later explaining the problem and seeking for help to fix it. When you correct for this math error (which I've done here) the sheets are actually fairly accurate.
The biggest problem, however, was not the sheets, but rather, that I went with my gut and not the math.
My gut does not have a good history in recent elections. To correct for this, I will be focusing much more on the math going forward.
Additionally, the riding polls from Mainstreet ended up being rather accurate, and, with the purchase formula outlined here, I will be buying them when possible.
In the end it's been a terrible night for me, both on a "who I wanted to win" level and a "made bad projections" level; but I will continue on. There are always more elections to look at around the world, and I'll be starting that within a few days by going over recent results in Slovenia.
Thursday, June 7, 2018
Ontario; counting continues
The maps show where things stand as counting continues.
So what went wrong?
In short; my gut was wrong and the math was right.
Post-mortem tomorrow morning.
Ontario E-Day commentary; ridings for an NDP majority
If the NDP wins these ridings:
Whitby
Mississauga East--Cooksville
Mississauga--Erin Mills
Mississauga--Centre
Flamborough--Glanbrook
Peterborough--Kawartha
Hastings--Kawartha Lakes--Brock
Huron--Bruce
Whitby
Mississauga East--Cooksville
Mississauga--Erin Mills
Mississauga--Centre
Flamborough--Glanbrook
Peterborough--Kawartha
Hastings--Kawartha Lakes--Brock
Huron--Bruce
They may win a majority.
Ontario E-Day commentary; PEI Projection
For fun, to pass the time as we all wait, I thought I would do a little PEI Projection:
Not much commentary; just that this is where things probably stand given the polls.
Subscribe to:
Posts (Atom)