|
 Who changed the thread tag to fantasy d20
|
#
?
Mar 12, 2016 10:15
|
|
|
|
| # ? Jun 1, 2024 21:52 |
|
|
Nate Silver's sweaty, pallid face and receding hairline make him a hell of a brand.
|
|
#
?
Mar 13, 2016 23:30
|
|
|
A guy with a BA in Econ and an excel that takes 10+minutes to load has become some kind of renowned voice of very serious people. Why.
|
|
#
?
Mar 14, 2016 02:00
|
|
|
The Revenant posted:A guy with a BA in Econ and an excel that takes 10+minutes to load has become some kind of renowned voice of very serious people. Why. because he was demonstrably better at making predictions than the people they'd been listening to previously and lol just lol if you don't think very serious decision makers aren't making very serious decisions based on excel work from dudes who have a BA in econ- I work in primary market research so I have some familiarity with this LGD has issued a correction as of 02:06 on Mar 14, 2016 |
|
#
?
Mar 14, 2016 02:03
|
|
|
The Revenant posted:A guy with a BA in Econ and an excel that takes 10+minutes to load has become some kind of renowned voice of very serious people. Why. He is the best they've got
|
|
#
?
Mar 14, 2016 02:03
|
|
|
The Revenant posted:A guy with a BA in Econ and an excel that takes 10+minutes to load has become some kind of renowned voice of very serious people. Why. Cuz as it turns out having more advanced degrees don't mean poo poo and often just means you are unemployed instead of making money posting about politics hell we are paying lowtax $10 to post about politics
|
|
#
?
Mar 14, 2016 04:39
|
|
|
|
|
#
?
Mar 14, 2016 04:47
|
|
|
Meta Ridley posted:538's model is stupid though. He doesn't weigh accuracy of polls, he doesn't account for Bernie's surge effect, and polls-plus weighs by loving endorsements in the most anti-establishment election cycle in decades. They have a page on historical poll bias and accuracy which is weighted in ~somehow~: http://fivethirtyeight.com/interactives/pollster-ratings/ TBH the major issue that would help a lot is better assessment on confidence by trying to normalize polls to see how well they represent segments in society. ACS data is more than enough for this - figure out who you polled and get a sense of how descriptive it is of society as a whole based on whatever parameters you define for demographics. It's not hard at all but tedious to do the first time. Once you have a script for it (and if you can reference ACS via its API), its automated for the rest give or take entering in the poll data manually... I've been to lazy to write a blog post for this. Meh. e: when I was doing some probabilistic systems analysis for government/aerospace, the work was all in excel. except for VB being the worst language ever created, excel is useful for rapid prototyping for people with minimal programming background. it's bad for performance and updating, but it generally works and everyone knows how to use it. i typically prototype statsy stuff or models with a lot of parameters I want to quickly tweak in excel before porting the real machinary to mathematica (because im a dirty physicist) or py. guidoanselmi has issued a correction as of 19:33 on Mar 14, 2016 |
|
#
?
Mar 14, 2016 19:28
|
|
|
I think Silver is ironically doing the exact thing he eviscerated the conservative news bubble for doing in 2012 (letting wishful thinking influence your observations)
|
|
#
?
Mar 14, 2016 19:54
|
|
|
The Revenant posted:A guy with a BA in Econ and an excel that takes 10+minutes to load has become some kind of renowned voice of very serious people. Why. College is useless. I mean I expect this kind of "just a bachelors in economics" from baby boomers or a human resources employee but the measure of a man is not the piece of paper that some state college printed out. I have a bachelors in mathematics and nobody who matters asks why I'm rich in completely unrelated fields.
|
|
#
?
Mar 14, 2016 20:56
|
|
|
loquacius posted:I think Silver is ironically doing the exact thing he eviscerated the conservative news bubble for doing in 2012 (letting wishful thinking influence your observations) Yeah no poo poo. Harry Enten wrote today that if Trump doesn't win Ohio he is FINISHED.
|
|
#
?
Mar 14, 2016 20:57
|
|
|
Bernice Anders posted:Yeah no poo poo. Harry Enten wrote today that if Trump doesn't win Ohio he is FINISHED. How exactly does it keep working out that Clinton's lead makes the delegate math basically insurmountable for Bernie but Trump is constantly teetering on the brink of disaster despite winning basically all the primaries I must know and then I can be a statistics guru as well
|
|
#
?
Mar 14, 2016 21:11
|
|
|
loquacius posted:How exactly does it keep working out that Clinton's lead makes the delegate math basically insurmountable for Bernie but Trump is constantly teetering on the brink of disaster despite winning basically all the primaries Some of it has to do with the way each primary awards their delegates, a lot of later states are wta in the Republican side, whereas the democrats award proportionally
|
|
#
?
Mar 14, 2016 21:17
|
|
|
loquacius posted:How exactly does it keep working out that Clinton's lead makes the delegate math basically insurmountable for Bernie but Trump is constantly teetering on the brink of disaster despite winning basically all the primaries The delegate math is different on the Republican and Democratic sides. Dem delegates are proportional throughout, so once you fall behind it's hard to catch up, but Republican delegates are a weird mix of unbound, proportional, and winner takes all by state or by congressional district, so a lot of the candidates' path to 1237 is very swingy and depends on their performance in large WTA states (like Ohio). Trump's performing at around his target rn, but losing the 66 delegates from Ohio would put him under by quite a bit. 538's delegate targets don't take things like Pennsylvania into account (where 54 of the 71 delegates are unbound), so that's another ~40 delegates off Trump's road to 1237. And on top of that, a bunch (a majority?) of Trump's delegates will actually be selected by the party establishments in each state, so while they'll be bound to him on the first ballot, they'll vote for someone else as soon as they're unbound. Trump absolutely needs 1237 on the first ballot to win.
|
|
#
?
Mar 14, 2016 21:18
|
|
|
loquacius posted:I think Silver is ironically doing the exact thing he eviscerated the conservative news bubble for doing in 2012 (letting wishful thinking influence your observations) It is increasingly more obvious that that is what is happening here. It's kind of depressing.  c'mon, Nate, it's over, man. You know this.
|
|
#
?
Mar 14, 2016 21:23
|
|
|
loquacius posted:I think Silver is ironically doing the exact thing he eviscerated the conservative news bubble for doing in 2012 (letting wishful thinking influence your observations) i still think Paris and San Bernardino has completely hosed this election cycle and we wouldn't be discussing trump right now if it weren't for that, or at least he wouldn't have been able to get away with the whole fascism deal
|
|
#
?
Mar 14, 2016 21:23
|
|
|
loquacius posted:How exactly does it keep working out that Clinton's lead makes the delegate math basically insurmountable for Bernie but Trump is constantly teetering on the brink of disaster despite winning basically all the primaries The simple reason is that because the Democrats only have two candidates, one of the two must get a majority. Because the Republicans have multiple candidates, it is possible for nobody to get a majority. Nobody thinks anyone can overtake Trump anymore, merely deny him 51%. This is aside from the delegate rules that make it harder to catch up in the Democratic primary.
|
|
#
?
Mar 14, 2016 21:27
|
|
|
loquacius posted:I think Silver is ironically doing the exact thing he eviscerated the conservative news bubble for doing in 2012 (letting wishful thinking influence your observations) I think it's more likely that now that he needs to carry the site he's doing more and more clickbaity poo poo, and saying trump could lose is better at getting clicks.
|
|
#
?
Mar 14, 2016 21:29
|
|
|
evilweasel posted:I think it's more likely that now that he needs to carry the site he's doing more and more clickbaity poo poo, and saying trump could lose is better at getting clicks. I think he's just an idiot. He very clearly let "NATE SILVER POLLSTER WIZARD!!" get to his head from 2008 and then let his preconceptions become the base for his predictions instead of the polling that got his wizardry declared.
|
|
#
?
Mar 14, 2016 21:33
|
|
|
Bernice Anders posted:Yeah no poo poo. Harry Enten wrote today that if Trump doesn't win Ohio he is FINISHED.
|
|
#
?
Mar 14, 2016 21:37
|
|
|
MaxxBot posted:Has internet polling improved at all and can it realistically replace traditional phone polling? That seems like the only option at this point. Internet polls were pretty bad during the UK election, so it's a bit of a mixed bag. It's clearly the future, but it's not there yet. Landline polling is obviously garbage, but cell phone polls are expensive so most people don't want to do them, especially at the state level. And even the best telephone polls still have single digit response rate. edit: Essentially there are two challenges. One is reaching the right people and getting them to take the poll and the second is determining if they are likely to vote. Both can be challenging, particularly if certain groups (like young people) turn out in unexpected numbers. Xandu has issued a correction as of 22:41 on Mar 14, 2016 |
|
#
?
Mar 14, 2016 22:30
|
|
|
waffle posted:I think Harry Enten is a big part of the reason 538 has some really lovely stuff nowadays. He is really not very good at quantitative stuff. Here is a good example where Harry literally copied and pasted Stata output into an article. Are you loving kidding me?? Like that just screams "I actually have no idea which outputs are relevant, so I'll just throw it all out there so I look like a smarty pants. more numbers = better" lol I got a feeling he just googled how to do regressions online and c/p the result cuz he actually has no idea what's going on I did the same poo poo in my undergrad when I first took econometrics
|
|
#
?
Mar 14, 2016 23:20
|
|
|
Enten literally majored in "Government" at Dartmouth lmbo
|
|
#
?
Mar 14, 2016 23:24
|
|
|
and the RNC rules need a candidate to have won 8 state majorities, not that 538 have actually cited that to my knowledge. DNC don't have such a rule, but they've never really needed it. Trump does not look like he will get 8 majorities. What the gently caress am I saying of course he will
|
|
#
?
Mar 14, 2016 23:25
|
|
|
stephenfry posted:and the RNC rules need a candidate to have won 8 state majorities, not that 538 have actually cited that to my knowledge. DNC don't have such a rule, but they've never really needed it. That's delegate majority, not popular vote majority. He'll make it.
|
|
#
?
Mar 15, 2016 00:25
|
|
|
Bernice Anders posted:Yeah no poo poo. Harry Enten wrote today that if Trump doesn't win Ohio he is FINISHED. All he wrote is that "Trump is unlikely to reach a majority of delegates needed to clinch the Republican nomination if he loses Ohio" which is true.
|
|
#
?
Mar 15, 2016 00:46
|
|
|
Vox Nihili posted:That's delegate majority, not popular vote majority. He'll make it. Yeah that's not gonna be a problem for Trump, but it could for the other candidates 
|
|
#
?
Mar 15, 2016 00:47
|
|
|
stephenfry posted:and the RNC rules need a candidate to have won 8 state majorities, not that 538 have actually cited that to my knowledge. DNC don't have such a rule, but they've never really needed it. Do the 8 states have to be from the actual 50 states or does Rubio winning Puerto Rico and DC help him?
|
|
#
?
Mar 15, 2016 01:03
|
|
|
Maxwells Demon posted:Do the 8 states have to be from the actual 50 states or does Rubio winning Puerto Rico and DC help him? States only.
|
|
#
?
Mar 15, 2016 01:07
|
|
|
Maxwells Demon posted:Do the 8 states have to be from the actual 50 states or does Rubio winning Puerto Rico and DC help him? Rule No. 1 posted:(b) For the purposes of this rule and all other rules, "state" or "states" shall be taken to include American Samoa, the District of Columbia, Guam, Northern Mariana Islands, Puerto Rico, and the Virgin Islands, except in Rule No. 14, and unless the context in which the word "state" or "states" is used clearly makes such inclusion inappropriate. Rule No. 40 posted:(b) Each candidate for nomination for President of the United States and Vice President of the United States shall demonstrate the support of a majority of the delegates from each of eight (8) or more states, severally, prior to the presentation of the name of that candidate for nomination. PR and DC count.
|
|
#
?
Mar 15, 2016 01:09
|
|
|
Jewel Repetition posted:States only. 
|
|
#
?
Mar 15, 2016 01:09
|
|
|
|
|
#
?
Mar 15, 2016 01:12
|
|
|
Xandu posted:Internet polls were pretty bad during the UK election, so it's a bit of a mixed bag. It's clearly the future, but it's not there yet. Landline polling is obviously garbage, but cell phone polls are expensive so most people don't want to do them, especially at the state level. And even the best telephone polls still have single digit response rate. I would add that elections are also much more fluid, dynamic these days. If a candidate says or does something, it is being fact-checked and reposted, dissected in real time. Social media makes it much more possible for people to spread a message and influence elections today, particularly people who would be otherwise without power to shape things. It's much harder to run misinformation campaigns without being penalized.
|
|
#
?
Mar 15, 2016 01:17
|
|
|
waffle posted:I think Harry Enten is a big part of the reason 538 has some really lovely stuff nowadays. He is really not very good at quantitative stuff. Here is a good example where Harry literally copied and pasted Stata output into an article. Are you loving kidding me?? Like that just screams "I actually have no idea which outputs are relevant, so I'll just throw it all out there so I look like a smarty pants. more numbers = better" Lmao holy poo poo this is bad.
|
|
#
?
Mar 15, 2016 01:24
|
|
|
Nate Silver is clearly phoning it in this year. His model works really well in the general election on presidential years when turnout is high but for everything else it seems to have very weak predictive value. Same for the other aggregates - 538, Sam Wang, etc. And I LIKE Sam Wang. But he hasn't done much this year (because he knows he can't).
|
|
#
?
Mar 15, 2016 01:48
|
|
|
Nobody knows anything ever
|
|
#
?
Mar 15, 2016 01:56
|
|
|
Typo posted:lol I got a feeling he just googled how to do regressions online and c/p the result cuz he actually has no idea what's going on I did the same poo poo in my undergrad when I first took econometrics This is exactly what it is. It reeks of an undergrad clubbing a pset because they don't know what's going on.
|
|
#
?
Mar 15, 2016 02:05
|
|
|
i read a gerrymandering article by sam wang and was kind of disappointed by how seriously he was treating the approach of diagnosing the severity of gerrymandering by seeing how far the mean state result is from the median district. gerrymandering isn't the only possible source of skew! just like minimizing surface area or distance from the centroid, these methods that only consider one or two dimensions of one or two variables just aren't one-size-fits-all metrics for quantifying gerrymandering. they can be useful as tools for courts in setting standards for unacceptable gerrymandering, but i don't think that's what sam wang was doing. i may have misinterpreted most democratic countries (and a few states, republican and democratic) have panels of experts take a few months and draw up districts that embody communities of interest in an apolitical manner, and sometimes those districts don't look awesome geometrically, or the partisan distribution is such that there's a skew (even considering competitiveness can lead to this - arizona's commission is told to draw districts that are competitive in elections, which, i've seen argued, results in overrepresentation of the minority party). if there is a problem, it needs to be addressed as part of a holistic analysis, not by creating a simple geometric or statistical rule which then just becomes another constraint for the gerrymanderer to work around
|
|
#
?
Mar 15, 2016 02:10
|
|
|
Mental-Rectangle posted:Nate Silver is clearly phoning it in this year. His model works really well in the general election on presidential years when turnout is high but for everything else it seems to have very weak predictive value. Same for the other aggregates - 538, Sam Wang, etc. And I LIKE Sam Wang. But he hasn't done much this year (because he knows he can't). Phoning it in on a landline.
|
|
#
?
Mar 15, 2016 02:47
|
|
|
|
| # ? Jun 1, 2024 21:52 |
|
|
Nate Bitchboy silver is clearly phoning it in but I'm dialing it in by which I mean my fists on his face.
|
|
#
?
Mar 15, 2016 06:09
|
|