ZyyzYzzy
RIP USA
- 25,295
- 48,789
WE VOTED TRUMP!
- Thread starter Araxen
- Start date
iannis
Musty Nester
- 31,351
- 17,656
Saying that the polls were flawed and his model is just fine is a little bit of a can kick. He was influencing those polls with the predictions he was throwing out based on that model. And he was doing it intentionally. He even apologized and then kept doing it. It's obviously not fine if what is being presented as an objective tool can be manipulated so easily.
If he was just making dry mathematical claims that's one thing. Garbage in, garbage out -- you can blame the pollsters. But he wasn't. It was all one giant appeal to authority. I'm not trying to be snide when I call him a modern astrologer. Instead of reading the bones, he's reading the polls. Statistics is a real thing. But what he's trying to do with it is an entirely political thing.
Really what's he's doing is taking a large set of often conflicting data and trying to logically reconcile it. He completely failed to reconcile it. The editorializing is just icing on his humble pie, but the real flaw is that his model is not predictive even in the abstract. I suppose you can give him some points for being the LEAST wrong. I mean it, I guess that's fair.
If he was just making dry mathematical claims that's one thing. Garbage in, garbage out -- you can blame the pollsters. But he wasn't. It was all one giant appeal to authority. I'm not trying to be snide when I call him a modern astrologer. Instead of reading the bones, he's reading the polls. Statistics is a real thing. But what he's trying to do with it is an entirely political thing.
Really what's he's doing is taking a large set of often conflicting data and trying to logically reconcile it. He completely failed to reconcile it. The editorializing is just icing on his humble pie, but the real flaw is that his model is not predictive even in the abstract. I suppose you can give him some points for being the LEAST wrong. I mean it, I guess that's fair.
- 1
- 3,223
- 3,493
CEO Matt Harrigan Who Plotted “Sniper Attack” Against Trump Has Been Placed on “Administrative Leave”
CEO Matt Harrigan Who Plotted “Sniper Attack” Against Trump Has Been Placed on “Administrative Leave”
How do people that are CEO's this day and age think they can say shit like this and it won't impact their company? Jesus.
CEO Matt Harrigan Who Plotted “Sniper Attack” Against Trump Has Been Placed on “Administrative Leave”
How do people that are CEO's this day and age think they can say shit like this and it won't impact their company? Jesus.
- 1
Swagdaddy
There is a war going on over control of your mind
- 1,960
- 1,870
Mist
Eeyore Enthusiast
- 30,413
- 22,201
Learn2math. If I say you have a 1/6 chance of rolling a 5 on a six sided die, and then you pick up a die, roll it, and get a 5, that doesn't mean my model was wrong.
If I say you have a 3/10 chance of getting a 3 or under on a 10 sided die, you roll, and get a 2, my model was still correct.
This is exactly what happened here. His model is a probabilistic model, that isn't predicting a specific outcome, only the chance of a specific outcome. Further, his model is actually the sum of a series of smaller probabilistic models that function in the same way.
ZyyzYzzy
RIP USA
- 25,295
- 48,789
His model was flawed cunt. Just because it was probalistic does not mean it was not based on flawed assumptions. It was
- 1
kegkilla
The Big Mod
- 11,320
- 14,738
Assigning probabilities to election outcomes is kind of dumb to begin since there's not a hell of a lot left up to chance or random events.
- 1
Mist
Eeyore Enthusiast
- 30,413
- 22,201
Or you could view it as the sum of an extremely large number of weighted random events.
Mist
Eeyore Enthusiast
- 30,413
- 22,201
ZyyzYzzy
RIP USA
- 25,295
- 48,789
A model has to make assumptions of some sort. If those assumptions are incorrect your midel is shit.
I have a model to predict multiple types WMD effects. If my assumptions, (decay rate, source term, etc...) are just flat not accurate or representative of actual things then guess what my model, which is probabilistic is returning shit.
- 1
Woefully Inept
Ssraeszha Raider
- 8,796
- 33,949
That shit at the base of the wall bones and ashes of the idiots that tried to scale it?
- 1
Mist
Eeyore Enthusiast
- 30,413
- 22,201
Your model of radioactive decay rate is based on the sum of probabilistic outcomes of billions (trillions?) of individual atomic-level reactions that produce an average decay rate. The issue with an election is that there's only one event to measure, we don't get to re-run the election hundreds of times under the same conditions to see if the model holds up.
On that note, you also have to take into account that every poll-based model lags behind the news cycle. Given how close the race was in at least 7 states, and the fact that the polls shifted 3-5% depending on the news cycle for or against a given candidate, if the election was held a week later you might have a completely different outcome.
Nate Silver's model was also the first to call Ohio an almost sure loss for Hillary and he had that called weeks ahead of everyone even considering how truly lost that race was (8 points is a lot of fucking points.) The issue was more with the fact that people were reading things into his model that weren't actually there, including Nate himself.
ZyyzYzzy
RIP USA
- 25,295
- 48,789
Okay. His model was still flawed. I don't know why you are defending him.
- 1
Mist
Eeyore Enthusiast
- 30,413
- 22,201
Your only evidence for the fact that it was flawed was that the outcome didn't line up with the favored candidate, when the favored candidate's chance of winning was only 2/3.
ZyyzYzzy
RIP USA
- 25,295
- 48,789
It inaccurately modeled the results on multiple levels in multiple locations. Maybe you and me differ in our definition of a useful model and one that is completely fucking worthless.
- 1
Mist
Eeyore Enthusiast
- 30,413
- 22,201
It wasn't supposed to model the results. It was supposed to model the probability distribution of potential outcomes.
ZyyzYzzy
RIP USA
- 25,295
- 48,789
Mist
Eeyore Enthusiast
- 30,413
- 22,201
The model is based on models similarly developed for sports betting. When a team wins against the bookmakers odds, you don't say that they don't know anything about football. You just realize that on any given Sunday, shit happens.