News for hackers. Stuff that matters.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by dmode on Tuesday March 27, 2018 @03:40PM (#16692021)
Given the margin of Trump's victory was so small, and he lost the overall popular vote, I would be surprised that anyone can be confident of a victory looking at any data set.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by corpMaverick on Tuesday March 27, 2018 @04:26PM (#16692420)
If every one was betting on Clinton 10 to 1. But you know it closer to 2 to 1. That is a good bet.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by creaghpatr on Tuesday March 27, 2018 @05:30PM (#16692910)
On election night, the odds were roughly 5 to 1 on PredictIt
Re:Palantir worked with Cambridge Analytica on the Facebook ... by Tempest1981 on Wednesday March 28, 2018 @04:54AM (#16696384)
On Huffington Post, the odds were 98% Clinton, iirc. I suspect this led some people to not vote.

Here it is: https://www.huffingtonpost.com/entry/high-probability-clinto...

Re:Palantir worked with Cambridge Analytica on the Facebook ... by corpMaverick on Wednesday March 28, 2018 @10:34AM (#16697954)
In 538 the odds were 70% for Clinton. I think the result was spot on.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by sjwright on Wednesday March 28, 2018 @04:08AM (#16696211)
There's no logical conflict between saying Clinton had a 10 to 1 chance of winning and Trump actually winning. 10 to 1 still means Trump wins 10% of the time.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by corpMaverick on Wednesday March 28, 2018 @10:41AM (#16698010)
I don't know why you were down voted. It should be obvious, but people still interpret odds like that as it was a sure thing.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by sjwright on Wednesday March 28, 2018 @08:56PM (#16703628)
Part of the problem is a confusion between raw poll results—where 59-41 means almost certain victory to one side—and statistical likelihood.

On election night, Fivethirtyeight gave Trump a 29 percent chance of winning the electoral college. Not a 71-29 poll result, a nearly three-in-ten chance of winning. Let that sink in for a second, and reevaluate the media narrative that the pollsters got it wrong.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by wpasc on Tuesday March 27, 2018 @04:30PM (#16692460)
Odds like that are meaningless (imo) for singular events. It's a good bet in repeatable scenarios, far less so in elections.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by eloff on Tuesday March 27, 2018 @04:49PM (#16692602)
When you take a bet, you play the odds. You evaluate your expected return should you take many such bets, and if it's positive and you can afford to lose, the smart money is in taking the bet.

In fact this is true of most anything you do in life - when you drove to work this morning you played the odds and the wager was your life against your salary. I'm sure you didn't think about it that way, but you took the bet

Re:Palantir worked with Cambridge Analytica on the Facebook ... by saalweachter on Tuesday March 27, 2018 @05:19PM (#16692816)
Both insurance companies and bookmakers are comfortable aggregating over many unique events.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by the_trapper on Tuesday March 27, 2018 @05:24PM (#16692863)
Fun fact, Trump carried more states in 2016 (30) than Obama carried in 2008 (28). Our country is so polarized that pretty much all elections are close.

https://en.wikipedia.org/wiki/United_States_presidential_ele... https://en.wikipedia.org/wiki/United_States_presidential_ele...

Re:Palantir worked with Cambridge Analytica on the Facebook ... by elgenie on Tuesday March 27, 2018 @05:41PM (#16692990)
States aren't a particularly meaningful counting unit: California encompasses more than 68 Wyomings by population. And the meaning of "carrying" a state by a 44.1% to 44.0% plurality is questionable.

The current division in the US is urban vs rural, with turnout differences and suburbs deciding elections.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by paulddraper on Tuesday March 27, 2018 @05:51PM (#16693073)
Number of states is indeed misleading. If each party gets half the states, it's a Democrat victory.

E.g. Carter carries 24 states and Ford carries 27; Carter wins 297-240.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by paulddraper on Tuesday March 27, 2018 @06:54PM (#16693578)
I'm really confused by the downvotes. This is the material evidence of

> The current division in the US is urban vs rural

I.e. you have to look at populations of states, not just state counts themselves.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by brianzelip on Wednesday March 28, 2018 @01:36AM (#16695611)
Maybe you got down voted because 24 + 27 != 50?
Re:Palantir worked with Cambridge Analytica on the Facebook ... by dragonwriter on Wednesday March 28, 2018 @01:51AM (#16695670)
D.C. is usually counted as a state in discussion of Presidential elections since it was alotted electors under the 23rd Amendment. (This actually comports with a common practice in federal statute law where any of the federal territories that are treated equal to a state in the particular law are simply lumped into the law’s particular definition of “state”, which is then used to refer to actual stated plus the equivalent-to-state territories in the context of that law.)
Re:Palantir worked with Cambridge Analytica on the Facebook ... by grzm on Wednesday March 28, 2018 @01:41AM (#16695632)
Likely 'paulddraper is using 24 to mean 23 states + DC:

https://en.wikipedia.org/wiki/United_States_presidential_ele...

(The election even included a faithless elector, something you don't see every day.)

Re:Palantir worked with Cambridge Analytica on the Facebook ... by paulddraper on Wednesday March 28, 2018 @04:12PM (#16701474)
Indeed. It was a vote by a Republican for the Libertarian nomination.

It was also notable for being the first electoral vote for a ticket with a woman.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by noobermin on Tuesday March 27, 2018 @07:06PM (#16693649)
It doesn't matter, the US presidential election isn't very representative, it values some groups' votes more than others.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by raarts on Tuesday March 27, 2018 @05:45PM (#16693028)
> Our country is so polarized that pretty much all elections are close

Which is to be expected in a two party system.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by jdavis703 on Tuesday March 27, 2018 @06:38PM (#16693446)
The national party apparatuses should be able to find a way to have moderating voices. Most large cities in the U.S. are a one party system dominated by Democrats. Yet if you look at the politicians who are mayors and council members, there's a lot of diversity in ideas and bitter policy disputes.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by mtgx on Wednesday March 28, 2018 @09:32AM (#16697481)
Single-winner voting systems such as FPTP are what causes a party to dominate at a local level, and also what creates a century-old two party system.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by na85 on Tuesday March 27, 2018 @06:47PM (#16693527)
Surely that is because cities are diverse and multicultural, something that the GOP has been trying to cast as a negative since at least Reagan.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by jiggliemon on Tuesday March 27, 2018 @08:35PM (#16694124)
Citation needed.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by jdavis703 on Tuesday March 27, 2018 @08:46PM (#16694172)
Let's not turn this in to a pointless political debate. But stories such as the Republican appointee to the DoI saying "diversity isn't important" [0] lead some people to conclude that there's effort to cast other goals as more important than diversity. There are enough threads on HN debating the relative merits of diversity and meritocracy that there's no need to add them here, but this is just an answer to the citation request.

0: https://www.cnn.com/2018/03/26/politics/ryan-zinke-diversity...

Re:Palantir worked with Cambridge Analytica on the Facebook ... by the_trapper on Tuesday March 27, 2018 @09:43PM (#16694489)
Is he really wrong though? What purpose does artificially created diversity serve? Shouldn't we always be looking for the best qualified individuals?

The goal should always be equal opportunity, not equal outcomes.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by fredophile on Tuesday March 27, 2018 @10:04PM (#16694616)
We should be looking for the individuals that will most improve the organization. That may mean hiring someone adequate today instead of someone spectacular who can start in 3 months. That could also mean choosing someone with a different background to most people in the organization. People with different backgrounds will have different biases and think about things differently. Having multiple perspectives is valuable in many organizations.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by spacehome on Wednesday March 28, 2018 @03:01AM (#16695973)
This is the bailey in the motte-and-bailey defense of diversity. Of course people from different backgrounds will contribute differing perspectives. But in practice, it always ends up being race, gender, and sexual orientation. No talk about getting representatives from groups that actually think in ways that are different. Where are the quotas for libertarians, green party members, immigrants, aspies, former attorneys, artists, competitive athletes, scoutmasters, stamp collectors, gearheads, all shades of personality types?
Re:Palantir worked with Cambridge Analytica on the Facebook ... by mbaha on Wednesday March 28, 2018 @03:45AM (#16696136)
Spot on. While the idea behind can be interesting, it is implemented by accepting people who have the same "social status" background (same schools, same opinions), but are of different gender/race.

I think that a white factory worker could add way more insight into how other communities views things, than hiring a black/female attorney does...

Having said that, I'm really afraid what the backlash against this type of diversity-thinking will be. The system will surely corrects itself in the future (it's already starting), but I'm not certain going back to another extreme would be very wise...

Re:Palantir worked with Cambridge Analytica on the Facebook ... by paulddraper on Wednesday March 28, 2018 @04:19PM (#16701529)
> diverse and multicultural, something that the GOP has been trying to cast as a negative since at least Reagan. reply

> stories such as the Republican appointee to the DoI saying "diversity isn't important" [0] lead some people to conclude that there's effort to cast other goals as more important than diversity.

Diversity being negative, and diversity being unimportant are two different statements.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by dfee on Tuesday March 27, 2018 @05:40PM (#16692980)
Polarization and margin of victory are orthogonal concepts, and certainly not linked by causation.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by dragonwriter on Tuesday March 27, 2018 @05:49PM (#16693054)
> Our country is so polarized that pretty much all elections are close.

I assume you mean pretty much all Presidential elections (because otherwise it's not true; lots of elections aren't close, and some electoral districts are so structurally guaranteed to be not close that elections are usually uncontested), but even then polarization does nothing to explain that, but the incentive to seek a winning coalition with minimal compromise on issues in a two-party system does.

Re:Palantir worked with Cambridge Analytica on the Facebook ... by specialist on Tuesday March 27, 2018 @11:49PM (#16695110)
Duverger's Law is why FPTP elections are close.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by grzm on Tuesday March 27, 2018 @05:43PM (#16693004)
Just as relevant as the popular vote in determining the outcome of an election based on the electoral college.
Re:Palantir worked with Cambridge Analytica on the Facebook ... by mhue on Wednesday March 28, 2018 @04:13AM (#16696233)
Imagine you train an algorithm to influence the voters to win the election. What would be the objective ? Maximize the probability that your candidate wins. This does not imply winning the popular vote, nor maximizing the margin of the electoral vote.