me: the problem isn't 338, the problem is you. you crave the certainty of a coloured riding map, you crave being told there is a 82% chance of something, when the truth is to live with uncertainty and randomness is what makes us human you: i'm sorry sir, but the question was "do you want fries?" - ThreadSky

jmcelroy.bsky.social • 20 hours ago

me: the problem isn't 338, the problem is you. you crave the certainty of a coloured riding map, you crave being told there is a 82% chance of something, when the truth is to live with uncertainty and randomness is what makes us human

you: i'm sorry sir, but the question was "do you want fries?"

Comments

adrian-dalessandro.bsky.social•16 hours ago

The riding level predictions are a bit of a problem, because of the last 10 years of vote-together-esque ABC strategic voting. I know a lot of people voting liberal purely because 338 projects liberals as having the best chance, and so it becomes a self-fulfilling prophecy. How much of their

adrian-dalessandro.bsky.social•16 hours ago

accuracy in toss-up ridings is due to influencing the ABC vote?

stateofthecity.bsky.social•2 hours ago

Riding level predictions are also a problem because they're tough to deliver in general, given costs & variables involved. If a major Party HQ gave me a tracked local number I'd trust it, even though central Party HQs are often wrong. Regional polls? Maybe. Extrapolations from national polls? LOL.

stateofthecity.bsky.social•2 hours ago

Example: most of my career, my job in an elxn race was GOTV. Good GOTV can change the outcome by at least a few % points. Decisions on where GOTV resources are made only days out by national/provincial (or even mayoral) HQs.

stateofthecity.bsky.social•2 hours ago

So while I will trust a day-out national horse race result on a public poll by a good pollster, how in the hell is an IVR national poll with a 1200 sample supposed to account for that kind of variance *riding by riding* across the country? Or a model built largely on similar polls?

tds.xyz•8 hours ago

probabilistic forecasts should describe, not contest, uncertainty. there's no conflict here. it's bad, when things are bad, though, and 338 is methodologically bad

jensvb.bsky.social•8 hours ago

Every election I tell myself I should build my own model, but building a good model is hard and quickly becomes a full-time job. And then I remember that I already have a job and some deadline is breathing down my neck and I go back to just complaining about other people’s models. 🥲

sv-rub-a-dub.bsky.social•20 hours ago

The problem is 338 for a whole other reason. Because it skews peoples perceptions and expectations.

deirdriu.bsky.social•20 hours ago

I feel that all 338 is doing is muddying the water. I have to wonder about certain organizations and their motives.
I voted according to historical patterns in my riding, with a single goal: shut out the CPC.

cognoscento.bsky.social•19 hours ago

There are lots of ridings, like mine, that don’t have clear historical patterns because of realignment of boundaries. We’re now made of an NDP stronghold and a CPC one, and the models haven’t kept up. The LPC has a strong candidate here that *could* come up the middle, but the models don’t see it.

deirdriu.bsky.social•18 hours ago

That's where people need to just pay attention. This may be happening in my riding, too - but one of the candidates keeps commissioning polls and adding extra mud to the image. It's been a difficult election.