4 ms·

Any global predictor that claims to have accurately determined Trump's victory should keep in mind that he lost the popular vote. Only a state-based predictor

Any global predictor that claims to have accurately determined Trump's victory should keep in mind that he lost the popular vote. Only a state-based predictor can be plausibly said to have been able to make such a prediction.

anigbrowl 10y ago
Yeah it really comes down to what you base your hunch on. when I saw the first exit poll saying that many voters had opted for a 'strong leader' I knew it was over even thought he polls hadn't closed anywhere. I have a great many hunches and some facility with visualization of directed graphs and other analytics, but am not clear on how to scrape and normalize the datasets I need.
th0raway 10y ago
I've seen one of those predictors. It looked at an honest signal, captured in a way that would be fairly representative of the population at large. Since the data had some location information attached, you could attach every point of signal to a congressional district. The model was simple, comparing the signal per district, and then building an electoral college map. It didn't nail every state, but it wasn't off by much, and gave Trump a slightly smaller victory than he actually had. I won't mention the signal because I want to keep my job. However I'd be surprised if there weren't at least a dozen companies that have access to some honest signal with such a good sample of the population at large as to be better at predicting the election than any pollster.
- lfam 10y ago
  It's interesting to think about where to find this information. Big ISPs have access to these "honest" signals on a large scale.
SomeStupidPoint 10y ago
Could you elaborate on your point? Im not sure I understand.
- Jweb_Guru 10y ago
  An accurate global predictor with no regional information should have predicted a Hillary win (since Hillary won the popular vote by a fair margin). So if you had a global predictor predicting a Trump win, either (1) it must have taken regional information into account (in order to understand that the distribution of votes was concentrated in such a way that the electoral college would not swing Hillary's way), or (2) it must not have been an accurate predictor. Most of the polls or methods that predicted a Trump win did not take regional information into account, which means they were just inaccurate predictors.
  - SomeStupidPoint 10y ago
    Ah, I don't disagree. But there might be "hidden" location information in datasets that dont appear to have it, such as Twitter. Word choice carries regional and social information. So it's conceivable Twitter data could correctly predict the election with ML techniques that essentially are estimating the electoral college without being trained on an explicit model of it.

anigbrowl 10y ago

Yeah it really comes down to what you base your hunch on. when I saw the first exit poll saying that many voters had opted for a 'strong leader' I knew it was over even thought he polls hadn't closed anywhere. I have a great many hunches and some facility with visualization of directed graphs and other analytics, but am not clear on how to scrape and normalize the datasets I need.

th0raway 10y ago

I've seen one of those predictors. It looked at an honest signal, captured in a way that would be fairly representative of the population at large. Since the data had some location information attached, you could attach every point of signal to a congressional district. The model was simple, comparing the signal per district, and then building an electoral college map. It didn't nail every state, but it wasn't off by much, and gave Trump a slightly smaller victory than he actually had. I won't mention the signal because I want to keep my job. However I'd be surprised if there weren't at least a dozen companies that have access to some honest signal with such a good sample of the population at large as to be better at predicting the election than any pollster.

lfam 10y ago
It's interesting to think about where to find this information. Big ISPs have access to these "honest" signals on a large scale.

SomeStupidPoint 10y ago

Could you elaborate on your point? Im not sure I understand.

Jweb_Guru 10y ago
An accurate global predictor with no regional information should have predicted a Hillary win (since Hillary won the popular vote by a fair margin). So if you had a global predictor predicting a Trump win, either (1) it must have taken regional information into account (in order to understand that the distribution of votes was concentrated in such a way that the electoral college would not swing Hillary's way), or (2) it must not have been an accurate predictor. Most of the polls or methods that predicted a Trump win did not take regional information into account, which means they were just inaccurate predictors.
- SomeStupidPoint 10y ago
  Ah, I don't disagree. But there might be "hidden" location information in datasets that dont appear to have it, such as Twitter. Word choice carries regional and social information. So it's conceivable Twitter data could correctly predict the election with ML techniques that essentially are estimating the electoral college without being trained on an explicit model of it.