Tuesday, October 22, 2019

Errors

There were a lot of errors. I want to go through each one and why.

Acadie Bathurst and Halifax
As outlined in my post, I relied on what other projections had, and many had the NDP winning here. My error was not realizing that this, like their egmont projections, was based on last-time math that no longer applied.

Saint Johns South and Avalon
I presumed the Liberals would not do as well in Newfoundland and Labrador; this was an error.

Cardigan, Saint John, Central Nova, and Miramichi
These errors all come from under-estimating Liberal strength and over-estimating Tory strength.

West Nova
I ignored a strong Tory candidate here, and should not have. This riding is traditionally Liberal, I found it odd thus that it could go CPC before any other NS riding; I should have known a strong candidate could make that happen.

Fredericton
I did know the Greens were strong here and thus, unlike the above, I do not count this one as a mistake of judgement, simply 'part of the job' as some ridings will be on the "knife's edge" and its hard to guess which way it will come down.


Le Moyne, La Prairie, Laurier Saint Marie, and Hochelaga
The pattern and play of the Bloc vote was always a bit unclear, this is not exactly the same Bloc that has run in previous elections. As such, like with Fredericton, I count these as simple 'part of the job'

Abitibi Temiscamingue, Berthier Maskinonge, Sherbrooke, Rimouski
I wanted the NDP to win these ridings. As such I let that bias creep into my projections. It is very hard to avoid, but I should have done better.

Saint Maurice
As I mentioned in my (s)election post, I want this riding to stop acting strangely. It did so again regardless.

Quebec, Compton, Chicoutimi, Trois-Rivières
These ridings all fell to that pattern of Bloc vote, and thus are 'part of the job'

Hamilton East, Kitchener Conestoga, Cambridge, Toronto Danforth, Davenport, Parkdale High Park, Whitby, Richmond Hill, King Vaughan, Newmarket Aurora
All of these ridings, all in the GTA-GHA, all went Liberal, and all were projected by me not to; yet all could have been projected to go Liberal if I'd held back my final projection for a few days. I simply guessed these ones prematurely

Kenora, Essex, Rainy River, Nickel Belt, Windsor West, Windsor Tecumseh
All 'part of the job'

Peterborough, Northumberland
I don't like Monsef. I still hold her partly responsible for the defeat of electoral reform. I wanted her to lose, and so, simply swapped a liberal win in the area for a win in another riding. This was foolish.

Elmwood Transcona, St. Boniface, Desenethe-Churchill River, Saskatoon West, Regina Lewvan, Regina Wascana
Here I failed to note how strong the Tories were in SK compared to MB. This is hard to see with the kind of polls we get here in Canada, and as such, I'm counting this as well as 'part of the job'

Nunavut
Similar to above, we just don't have proper polling to do this.

(Note, Alberta was called 100% correctly by me and many others)

Victoria
I wanted the Greens to win here.

Port Moody, Coquitlam, Kootenay, Burnaby North, Langley City
More 'part of the job' ridings



Ridings that I failed: 15 (IE, I made a mistake/bias)

Ridings guessed prematurely: 13 (IE guessed too early)

Ridings "the model" failed: 27 (IE 'part of the job')

As such if I'd done a better job with my guesses, I'd have only 27 errors, or, 311 correct calls. If I'd simply not gone early, I'd have had 40 errors, or 298 correct calls. As such I'm going to stick with the gut system for the time being and try to better my calls.

No comments:

Post a Comment