Red Mosquito

tragabigzanda wrote:

Green Habit wrote:

Green Habit wrote:Nate Silver also said that Hillary was a 77%-23% favorite just two weeks ago. Let's let the DNC finish and soak in and then see what happens.

Didn't we already learn these sorts of polls are arbitrary back when Gore won the popular vote but lost the election? I think it's just gonna come down to Ohio and Penn this year, with every other vote being an exercise in ego-futility...

It's probably worth clarifying that the numbers above are not giving a poll result but are the output of a probability model. The model actually takes into account the detail down to electoral college level and the likelihood that a nominee can get enough EC votes to win. It's not saying that 38% of voters will vote for Trump. It's saying that if you could theoretically reset The Matrix three times (at this point in time, with people's current opinion of the nominee's) as an Election Day then Trump would end up as president in one of those worlds.

If you are saying that the underlying polls that drive the model aren't of much value then Silver partially agrees with you, at least at this point in time. The polls begin to become much more valuable as inputs into the model the closer it is to the election date. The fact that they use a wide variety of poll sources with weightings is what is meant to ensure that any one "arbitrary" poll shouldn't overwhelmingly affect the model's prediction. If Trump starts consistently getting above 40% in the model (assuming you believe in the methodology it uses) then it really is becoming a coin toss.

Trump winning at around a 40% probablility wouldn't show that the model is useles (although it may need more refinement). Trump winning if he sat at a 5 % probability more strongly indicates that the model should be completely scrapped (although maybe it was that 1 in 20 time that he was expected to win)

TL;DR: Neo for President 2016. VP Morpheus would kickass.

B wrote:

tragabigzanda wrote:

Green Habit wrote:

Green Habit wrote:Nate Silver also said that Hillary was a 77%-23% favorite just two weeks ago. Let's let the DNC finish and soak in and then see what happens.

Didn't we already learn these sorts of polls are arbitrary back when Gore won the popular vote but lost the election? I think it's just gonna come down to Ohio and Penn this year, with every other vote being an exercise in ego-futility...

Nate Silver doesn't publish a poll. He publishes a probability of the candidate being elected based on polls that are weighted according to size, past performance, and a bunch of I don't know what. In that tweet, it's not that 68% of people will vote for Hillary, it's that she has a 68% chance of winning the election, if it were held today.

Beaten by B. And way more succinctly too

Ignore my waffle.

Classic B.

tragabigzanda wrote:Hey, I feel way smarter than before you guys posted. Plus I kind of want to eat waffles.

Pancakes currently sit at 80% as the preferred flour based breakfast, according to the polls plus forecast. Better hope today is the 1 in 5 day that waffles come out on top.

B wrote:

tragabigzanda wrote:

Green Habit wrote:

Green Habit wrote:Nate Silver also said that Hillary was a 77%-23% favorite just two weeks ago. Let's let the DNC finish and soak in and then see what happens.

Didn't we already learn these sorts of polls are arbitrary back when Gore won the popular vote but lost the election? I think it's just gonna come down to Ohio and Penn this year, with every other vote being an exercise in ego-futility...

Nate Silver doesn't publish a poll. He publishes a probability of the candidate being elected based on polls that are weighted according to size, past performance, and a bunch of I don't know what. In that tweet, it's not that 68% of people will vote for Hillary, it's that she has a 68% chance of winning the election, if it were held today.

Classic B! In the sense that...B is basically right. The 538 model runs thousands of simulations based on the variables that B mentions (plus others), and then just lets us know how often each candidate wins. It also does take the electoral college into account, so it's not just looking at popular vote (although nationwide polls are fed into the model to look for trendlines and stuff like that). If you dive deeper, it even lets us know the probability of the scenarios trag describes, i.e.: there's a 6.2% chance that Clinton "pulls a Gore" (wins popular vote, loses electoral college; really this should be called "pulling a Hayes") and a 1.2% chance Trump does. Something else they're tracking: they say there's a 2.8% chance that Gary Johnson wins at least one electoral vote.

With nearly every one of Silver's model updates, he also makes clear that polls from June through most of August are not that predictive, historically. The polls-plus and polls-only models always come with a disclaimer that they don't feel super-confident about them until September at the very earliest.

Also, writers at 538 often make this point: the model has Clinton up 67-32, and many people read that as, "Hillary's a huge favorite!" However, things that only have a one-in-three chance of happening do happen (about 1/3 of the time). Consider: if someone was going to roll a three-sided dice (bear with me) 100,000 times, and one side had Trump's face on it and two sides had Hillary's face on it, and they said, "Whoever's face comes up the most gets to be president," a Hillary supporter would probably feel comfortable agreeing to that system. But if the dice was just rolled one time, and the face that comes up gets to be president, I think the Hillary supporter would be sweating a great deal.

What I'm trying to say is: imagine you're on a game show and there are three doors...

Hill really packs 'em in.

Simple Torture wrote:

B wrote:

tragabigzanda wrote:

Green Habit wrote:

Green Habit wrote:Nate Silver also said that Hillary was a 77%-23% favorite just two weeks ago. Let's let the DNC finish and soak in and then see what happens.

Didn't we already learn these sorts of polls are arbitrary back when Gore won the popular vote but lost the election? I think it's just gonna come down to Ohio and Penn this year, with every other vote being an exercise in ego-futility...

Nate Silver doesn't publish a poll. He publishes a probability of the candidate being elected based on polls that are weighted according to size, past performance, and a bunch of I don't know what. In that tweet, it's not that 68% of people will vote for Hillary, it's that she has a 68% chance of winning the election, if it were held today.

Classic B! In the sense that...B is basically right. The 538 model runs thousands of simulations based on the variables that B mentions (plus others), and then just lets us know how often each candidate wins. It also does take the electoral college into account, so it's not just looking at popular vote (although nationwide polls are fed into the model to look for trendlines and stuff like that). If you dive deeper, it even lets us know the probability of the scenarios trag describes, i.e.: there's a 6.2% chance that Clinton "pulls a Gore" (wins popular vote, loses electoral college; really this should be called "pulling a Hayes") and a 1.2% chance Trump does. Something else they're tracking: they say there's a 2.8% chance that Gary Johnson wins at least one electoral vote.

With nearly every one of Silver's model updates, he also makes clear that polls from June through most of August are not that predictive, historically. The polls-plus and polls-only models always come with a disclaimer that they don't feel super-confident about them until September at the very earliest.

Also, writers at 538 often make this point: the model has Clinton up 67-32, and many people read that as, "Hillary's a huge favorite!" However, things that only have a one-in-three chance of happening do happen (about 1/3 of the time). Consider: if someone was going to roll a three-sided dice (bear with me) 100,000 times, and one side had Trump's face on it and two sides had Hillary's face on it, and they said, "Whoever's face comes up the most gets to be president," a Hillary supporter would probably feel comfortable agreeing to that system. But if the dice was just rolled one time, and the face that comes up gets to be president, I think the Hillary supporter would be sweating a great deal.

What I'm trying to say is: imagine you're on a game show and there are three doors...

Wait! Did I just get out waffled AND out B'ed???!!!

what a world we would live in if every poll clearly showed the methodology

this isnt a reference to the nate silver thing - they obviously provide their method

but there seems to be a new poll that dominates the mainstream newswires every 3 days: "a cnn poll".... a "USA today poll"....an "X University poll"

some of these end up being like 380 people, they don't tell you how many people were polled from what states, whether it was landlines only, what time of day (unless it is a family member, I do not answer the telephone during work hours)

i want to know how meaningless each of these polls are

there should be a law

we need poll reform!

I wrote that whole post (and checked on the baby in the middle of it, and checked my stats, and took a dump, and then finally hit submit) and I knew that there were other posts explaining the whole thing when they came up on the next screen. But I wasn't going to let it go to waste.

you did all the work, so might as well

Citizen Dick wrote:

Simple Torture wrote:

B wrote:

tragabigzanda wrote:

Green Habit wrote:

Green Habit wrote:Nate Silver also said that Hillary was a 77%-23% favorite just two weeks ago. Let's let the DNC finish and soak in and then see what happens.

Didn't we already learn these sorts of polls are arbitrary back when Gore won the popular vote but lost the election? I think it's just gonna come down to Ohio and Penn this year, with every other vote being an exercise in ego-futility...

Nate Silver doesn't publish a poll. He publishes a probability of the candidate being elected based on polls that are weighted according to size, past performance, and a bunch of I don't know what. In that tweet, it's not that 68% of people will vote for Hillary, it's that she has a 68% chance of winning the election, if it were held today.

Classic B! In the sense that...B is basically right. The 538 model runs thousands of simulations based on the variables that B mentions (plus others), and then just lets us know how often each candidate wins. It also does take the electoral college into account, so it's not just looking at popular vote (although nationwide polls are fed into the model to look for trendlines and stuff like that). If you dive deeper, it even lets us know the probability of the scenarios trag describes, i.e.: there's a 6.2% chance that Clinton "pulls a Gore" (wins popular vote, loses electoral college; really this should be called "pulling a Hayes") and a 1.2% chance Trump does. Something else they're tracking: they say there's a 2.8% chance that Gary Johnson wins at least one electoral vote.

With nearly every one of Silver's model updates, he also makes clear that polls from June through most of August are not that predictive, historically. The polls-plus and polls-only models always come with a disclaimer that they don't feel super-confident about them until September at the very earliest.

Also, writers at 538 often make this point: the model has Clinton up 67-32, and many people read that as, "Hillary's a huge favorite!" However, things that only have a one-in-three chance of happening do happen (about 1/3 of the time). Consider: if someone was going to roll a three-sided dice (bear with me) 100,000 times, and one side had Trump's face on it and two sides had Hillary's face on it, and they said, "Whoever's face comes up the most gets to be president," a Hillary supporter would probably feel comfortable agreeing to that system. But if the dice was just rolled one time, and the face that comes up gets to be president, I think the Hillary supporter would be sweating a great deal.

What I'm trying to say is: imagine you're on a game show and there are three doors...

Wait! Did I just get out waffled AND out B'ed???!!!

And door question'd.

i liked ST's version

the dice thing really resonated with me

they should only run these polls in swing states.

i dont care what 900 people from across the country think

i care what the 23 people who vote in Okocheebee FL think

Simple Torture wrote:I wrote that whole post (and checked on the baby in the middle of it, and checked my stats, and took a dump, and then finally hit submit) and I knew that there were other posts explaining the whole thing when they came up on the next screen. But I wasn't going to let it go to waste.

Totally get it. It was a good post.

Hope the dump was invigorating.

96583UP wrote:what a world we would live in if every poll clearly showed the methodology

this isnt a reference to the nate silver thing - they obviously provide their method

but there seems to be a new poll that dominates the mainstream newswires every 3 days: "a cnn poll".... a "USA today poll"....an "X University poll"

some of these end up being like 380 people, they don't tell you how many people were polled from what states, whether it was landlines only, what time of day (unless it is a family member, I do not answer the telephone during work hours)

i want to know how meaningless each of these polls are

there should be a law

we need poll reform!

You'll have to, of course, look at this with the full knowledge that it's just one website's point of view (as unbiased as they might try to be), but 538 has an extensive rating system for the polls they use: http://projects.fivethirtyeight.com/pollster-ratings/

They look at the number of times the poll appears in their database, how many races they've called correctly, and how off they are even when they do call a race correctly. They even try to course correct for polling organizations by seeing if they constantly over-rate Democrats or Republicans.

Simple Torture wrote:

96583UP wrote:what a world we would live in if every poll clearly showed the methodology

this isnt a reference to the nate silver thing - they obviously provide their method

but there seems to be a new poll that dominates the mainstream newswires every 3 days: "a cnn poll".... a "USA today poll"....an "X University poll"

some of these end up being like 380 people, they don't tell you how many people were polled from what states, whether it was landlines only, what time of day (unless it is a family member, I do not answer the telephone during work hours)

i want to know how meaningless each of these polls are

there should be a law

we need poll reform!

You'll have to, of course, look at this with the full knowledge that it's just one website's point of view (as unbiased as they might try to be), but 538 has an extensive rating system for the polls they use: http://projects.fivethirtyeight.com/pollster-ratings/

They look at the number of times the poll appears in their database, how many races they've called correctly, and how off they are even when they do call a race correctly. They even try to course correct for polling organizations by seeing if they constantly over-rate Democrats or Republicans.

hm,

less than 100 days, then we never have to hear about stupid elections again.

There's this funny or die trump spoof on Netflix called The Art of the Deal.

dprival78 wrote:


Hill really packs 'em in.

It's those goddamn fire marshals