|
The only thing I really seem to perceive in Silver is a tendency to hop onto twitter and act like a dick.
|
# ? Oct 25, 2018 22:32 |
|
|
# ? May 25, 2024 08:52 |
|
Lotta pretend statisticians in this thread making arguments that show they have no understanding of statistics. Otherwise known as Thursday.
|
# ? Oct 25, 2018 22:33 |
|
Frightening Knight posted:Interesting. I would sincerely love to know more, but I doubt I would understand the math involved sadly. no math involved, it's a classic dumb pundit trick. by way of demonstration, I own three racecars in a race. car 1 comes in 10th place. car 2 comes in 2nd place. car 3 explodes into the crowd killing half a dozen people. next race, car 1 comes in 2nd place, car 2 explodes into the crowd killing half a dozen people, and Car 3 comes in 16th. anyone who tells you "Forums Poster Yeowch!!! My Balls!!! Takes Second Place Twice In Stunning Display Of Racing Acumen" is absolutely full of poo poo
|
# ? Oct 25, 2018 22:47 |
|
FiveThirtyEight had Trump's chances at around 12-13% in mid-October, it was only on 10/30 that they had Trump's chances over 20% (Comey's memo was on 10/28), and not until 11/3 that they had him at the roughly 1/3rd chance of winning everyone quotes (which dropped back down to 30% the day of the election). For contrast, NYT had Trump at 7% in in mid-October, 10% on 10/30, and 15-16% on the last few days of the election. I think the main reason all the poll-analyzers get so much crap for 2016 is because until late October, they all showed a virtually-guaranteed Clinton victory, and it was only in the last 2 weeks that they showed any movement towards a competitive race. So, most people interpret that as either either the race did genuinely tighten only in the end (Comey), or the poll-readers hastily corrected themselves after deluding their audiences for the majority of the race.
|
# ? Oct 25, 2018 23:14 |
|
Euphoriaphone posted:FiveThirtyEight had Trump's chances at around 12-13% in mid-October, it was only on 10/30 that they had Trump's chances over 20% (Comey's memo was on 10/28), and not until 11/3 that they had him at the roughly 1/3rd chance of winning everyone quotes (which dropped back down to 30% the day of the election). it was the polls themselves that tightened at that stage, not the aggregators changing their interpretation of them. The only day the poll numbers actually matter is on election day, so getting mad that they didn't have the exact election outcome pinned by mid october is absurd. Also, a 10-20 percent chance of losing is not "virtually-guaranteed" God drat you guys are bad at statistics and don't even realize it.
|
# ? Oct 25, 2018 23:21 |
|
Heck Yes! Loam! posted:
I have basically no math education, but my extensive hours with X-COM have made me extremely leery of taking risks that have a 20% chance of failure.
|
# ? Oct 25, 2018 23:25 |
|
Frightening Knight posted:My primary issue with Swan is that he defended Nate Parker, which is lol worthy. I also heard he has interesting opinions about gay people but I have no proof of that. Looking at his wiki page on his views, yikes.
|
# ? Oct 25, 2018 23:28 |
|
Prester Jane posted:I have basically no math education, but my extensive hours with X-COM have made me extremely leery of taking risks that have a 20% chance of failure. X-Com chances are a much better representation of actual chance than people's brains can manage on their own, so it is a good metric. Even if you have a 99% chance of doing something, that 1 % will gently caress you more than your brain think it would. People are even worse at this when dealing with statistics that aren't as certain, like 80%/20% odds. Their brain latches on to the 80% and ignores the 20. mother fuckers, thats a 1-5 chance of failure. I wouldn't put anything important on those odds, least of all an election.
|
# ? Oct 25, 2018 23:34 |
|
Saagonsa posted:Looking at his wiki page on his views, yikes. Yeah. Like I like him as a shitposter, but he’s pretty bad on a bunch of issues.
|
# ? Oct 25, 2018 23:43 |
|
Heck Yes! Loam! posted:
Seriously, a turn at Russian Roulette only kills you 17% of the time. Winning virtually guaranteed*!
|
# ? Oct 25, 2018 23:54 |
|
Nate Silver is the middle of the pack of the popular forecasters with a long track record (i.e., there was no upshot in 2012). People who don't know statistics don't understand probability, but people who know just a little bit of statistics think that being less wrong is more important than being more right, which isn't true. Sam Wang and Drew Linzer are both better than Nate Silver. Yes, even after 2016. The best way to measure who good a forecast is is through Brier scores. Brier scores take into account not only whether the prediction was correct, but the certainty involved. In 2016, which was by far the best historical performance by 538, they barely edged out Sam Wang: https://www.buzzfeednews.com/article/jsvine/2016-election-forecast-grades But in 2012, Sam Wang and Drew Linzer did quite a bit better than Nate: http://www.rationality.org/resources/updates/2012/was-nate-silver-the-most-accurate-2012-election-pundit As did Sam in 2008: http://election.princeton.edu/2008/11/11/post-election-evaluation-part-2/ Now, the differences in Brier scores are relatively minor. They historically favor Sam Wang and Drew Linzer over Nate, though Nate did get an edge over them in 2016. But as the first link shows, Nate only pulls ahead in 2016 in the last few days. Which brings be to why Nate is not the best: his models are intended to be highly volatile. It is great for driving traffic to his site, but terrible for actual long term statistical analysis. Likewise, Nate's model is intentionally underconfident, because even people with an undergraduate understanding of statistics will make the mistake of penalizing underconfidence less than overconfidence. Sam Wang's model, even in 2016, had a much smaller mean absolute error of the predicted vote shares, if you aggregate it over the last 100 days (instead of just looking at "day of" predictions) https://mpra.ub.uni-muenchen.de/83282/1/MPRA_paper_83282.pdf Nate did better in 2016 because his models are underconfident by design (which will make him always less wrong in an upset). It is based on the not unreasonable idea that readers will penalize underconfidence less than overconfidence (we dont think much of giving Hillary just an 80% chance of winning Delaware while being horrified that someone gave Hillary a 65% chance of winning Florida). It's something that has been long known about Nate's model: https://andrewgelman.com/2010/11/03/some_thoughts_o_8/ (note that Gelman from this last link is actually pretty sympathetic to Nate, more so than Sam, but still concludes, all the way back in 2010: " Nate’s forecasts do seem underconfident! Out of the 39 races where he gave the Republican candidate between 60% and 90% chance of winning, the Republicans snagged all 38. Apparently he could’ve tightened up his predictions a lot. Wang appears to be correct.") joepinetree fucked around with this message at 02:33 on Oct 26, 2018 |
# ? Oct 26, 2018 01:22 |
|
joepinetree posted:Nate Silver is the middle of the pack of the popular forecasters with a long track record (i.e., there was no upshot in 2012). People who don't know statistics don't understand probability, but people who know just a little bit of statistics think that being less wrong is more important than being more right, which isn't true. This guys statistics. Thanks for the effortpost!
|
# ? Oct 26, 2018 01:36 |
|
Of course there's the very real probability that we live in a reality where legitimate democratic process in this country is dead and has been for the past decade at the bare minimum, and what we're witnessing is potentially the rotten fruit of those efforts.
|
# ? Oct 26, 2018 03:15 |
joepinetree posted:Nate Silver is the middle of the pack of the popular forecasters with a long track record (i.e., there was no upshot in 2012). People who don't know statistics don't understand probability, but people who know just a little bit of statistics think that being less wrong is more important than being more right, which isn't true. drat I'd forgotten what it was like to see posts this good in these parts
|
|
# ? Oct 26, 2018 03:28 |
|
VitalSigns posted:Seriously, a turn at Russian Roulette only kills you 17% of the time.
|
# ? Oct 26, 2018 04:24 |
|
Harik posted:Either something happens, or it doesn't. Therefore everything is a coinflip. Tide goes in, tide goes out. You can’t explain it!
|
# ? Oct 26, 2018 08:56 |
|
Majorian posted:Tide goes in, tide goes out. You can’t explain it! Fuckin' maths. How do they work?!
|
# ? Oct 26, 2018 15:52 |
|
Elephant Ambush posted:Fuckin' maths. How do they work?! Clown face paint and Faygo
|
# ? Oct 26, 2018 15:54 |
|
VitalSigns posted:Seriously, a turn at Russian Roulette only kills you 17% of the time. depends on if it's a revolver and you spin it after firing, otherwise it progressively gets worse odds if it's a semiauto, well then you get to be the guy from Chicago
|
# ? Oct 26, 2018 16:33 |
|
exploded mummy posted:depends on if it's a revolver and you spin it after firing, otherwise it progressively gets worse odds Even then, the odds are not 100%, you could be the 1 in a million with a manufacturing defect. The takeaway here, is always have a gun in your mouth.
|
# ? Oct 27, 2018 05:58 |
|
What's going on with the USPol closure? Is USPOL not supposed to be discussed in D&D unless you agree with centrist thought?
|
# ? Oct 28, 2018 10:39 |
|
i took it as there are too many "uspol" threads. trump should be the uspol thread but way too many posters have itchy report fingers.
|
# ? Oct 28, 2018 10:40 |
|
Groovelord Neato posted:i took it as there are too many "uspol" threads. trump should be the uspol thread but way too many posters have itchy report fingers. itchy report fingers wouldn't matter if mods didn't indulge them as far as i'm aware, any discussion of the dems' failings is verboten in the trump thread, so it's not an actual USPOL thread, just a echochamber for centrists
|
# ? Oct 28, 2018 10:43 |
|
Condiv posted:itchy report fingers wouldn't matter if mods didn't indulge them It's also just kinda unreadable for a number of reasons.
|
# ? Oct 28, 2018 10:44 |
|
oh you're right. it's kinda dumb to go "well this thread generated 400 reports" as if that means a majority or any of the reports were justified. i legit can't imagine reporting a post.
|
# ? Oct 28, 2018 10:45 |
|
Yeah, I don't what the point is unless literally, all US politics is going to be devoted to Trump. Maybe you could argue both should be closed and reformed into an "battlefield" thread but lets be honest it, every other post is going to be reported in that thread because well both sides honestly have very little common ground.
|
# ? Oct 28, 2018 10:50 |
|
Ardennes posted:Yeah, I don't what the point is unless literally, all US politics is going to be devoted to Trump. it shouldn't really matter that boths sides have very little common ground. debate and arguing is kinda the point of D&D. I'd say a reversal of fortune rule is necessary if a certain side can't stop reporting posts that aren't against the rules. that or naming and shaming
|
# ? Oct 28, 2018 10:59 |
|
To be fair, maybe it is time for both sides to "learn to live with each other" and having Trump and the rest of US Politics separated doesn't make much sense rhetorically. That said, I can understand from the perspective of moderators since that thread is going to be to be lit. I guess one way to do need a pretty strictly defined rule on what is qualified to be a report.
|
# ? Oct 28, 2018 11:15 |
|
Ardennes posted:To be fair, maybe it is time for both sides to "learn to live with each other" and having Trump and the rest of US Politics separated doesn't make much sense rhetorically. That said, I can understand from the perspective of moderators since that thread is going to be to be lit. I remember a former mod a few days ago giving LK the advice of "if it's not breaking the SA forum rules, just mark the report as resolved", so hopefully that's the game plan. Edit: And no racism/slurs.
|
# ? Oct 28, 2018 11:24 |
|
From what I gathered it seems as much because the USPOL thread kinda sees disuse, between this thread, the Trump threads and C-SPAM.
|
# ? Oct 28, 2018 11:51 |
|
BadOptics posted:I remember a former mod a few days ago giving LK the advice of "if it's not breaking the SA forum rules, just mark the report as resolved", so hopefully that's the game plan. Yeah, I could see a larger thread needing some special rules on top of that but they need to be strictly defined. Ghost Leviathan posted:From what I gathered it seems as much because the USPOL thread kinda sees disuse, between this thread, the Trump threads and C-SPAM. Well C-Spam is another subforum, and this thread seems to be chiefly about very particular infighting between factions of Democrats. I think there is actually plenty of room for an American politics thread that isn't just the Trump thread. If anything Trump and the "rest" of US politics needs to be merged together.
|
# ? Oct 28, 2018 12:04 |
|
it's impossible to talk about how we got here and trump without talking about the failures of the clinton and obama admins.
|
# ? Oct 28, 2018 12:15 |
|
Groovelord Neato posted:it's impossible to talk about how we got here and trump without talking about the failures of the clinton and obama admins. That is sort of the issue that there have been two different "chambers" of discussion and they have split apart. A single thread would be like dogs and cats living together but maybe that is what has to happen? One issue is simply speed and the Trump thread moves fast, and I assume a megathread would have even most posts in it.
|
# ? Oct 28, 2018 12:18 |
|
Groovelord Neato posted:it's impossible to talk about how we got here and trump without talking about the failures of the clinton and obama admins. Even beyond that, it’s just tiring to have any discussion of republicans succeeding in their evil plans met with “ugh more nothing matters posts.” Why try to discuss politics there if only hopeful fantasy is meaningfully responded to?
|
# ? Oct 28, 2018 12:26 |
|
Trumpthread is a cult, so any meaningful discussion there is by definition impossible.
|
# ? Oct 28, 2018 12:32 |
|
Cerebral Bore posted:Trumpthread is a cult, so any meaningful discussion there is by definition impossible. Granted, I think its existence as very much "its own thing" is part of the explanation of how it works out.
|
# ? Oct 28, 2018 13:20 |
|
How did they take the Mueller Investigation "prepare for diappointment" announcement?
|
# ? Oct 28, 2018 13:21 |
https://twitter.com/thehill/status/1056521221062516737?s=19
|
|
# ? Oct 28, 2018 13:36 |
|
Ghost Leviathan posted:From what I gathered it seems as much because the USPOL thread kinda sees disuse, between this thread, the Trump threads and C-SPAM. This was why I closed USPOL. Main Painframe astutely noted that we went all day yesterday without a single post in USPOL despite a major right-wing terrorist event occurring, which signals to me that interest isn't really there right now. PPJ and I are still deciding what should happen, but essentially what is likely is that after the midterms either we will make a fresh USPOL thread or something more significant may change, depending on what we decide. I left this thread open because it serves a slightly different purpose, although it's possible we might refresh this one as well.
|
# ? Oct 28, 2018 14:38 |
|
|
# ? May 25, 2024 08:52 |
|
https://twitter.com/eoinhiggins_/status/1056432637764665344?s=21 https://twitter.com/eoinhiggins_/status/1056548219453403136?s=21
|
# ? Oct 28, 2018 17:23 |