|
canepazzo posted:Uhm, the USC poll must be broken. So, equivalent to an 8 point lead elsewhere?
|
#
?
Oct 24, 2016 08:17
|
|
|
|
| # ? Jun 12, 2024 06:41 |
|
|
ascii genitals posted:I've had plenty of 1% probability things occur to me playing poker, much less 15%. This election going to Trump would not be like getting hit by lightning--I get what you're saying, but the chance is the chance. The thing is, in things like poker or football, opportunities for rare things are constant, so they still happen every now and then. This election, though, is happening once. It isn't going to repeat until Trump eventually wins; his low chances, appropriately enough, mean that he's really, really unlikely to win this. Also, most non-538 aggregators are giving Trump far lower chances than 15%; Nate himself admits that his model is deliberately conservative and includes the possibility of, as he puts it, everything we know about polling being wrong. Other sources are putting Trump's victory chances at <5%.
|
|
#
?
Oct 24, 2016 08:20
|
|
|
Charlz Guybon posted:So, equivalent to an 8 point lead elsewhere? 538 adjusts a tie to a 4 point lead for Hillary, so yeah, about 5-6 lead elsewhere. On the low end of national average, but still more in range I'd say.
|
|
#
?
Oct 24, 2016 08:21
|
|
|
Gyges posted:Really, instead of trying to find analogies like field goal misses and the like, we should be focusing on what the chance of Trump winning really is. It's the statistical chance that everything we know is wrong. It's not an event that rests on some place kicker holding the ball wrong, or the field being wet. It's more like being struck by lightning or being hit by a meteorite. It's more likely than those things though. For trump to win he needs some combination of the race to close, polling to be systematically off and turnout to be better than expected. This is unlikely but it's not super so.
|
|
#
?
Oct 24, 2016 08:22
|
|
|
Roland Jones posted:The thing is, in things like poker or football, opportunities for rare things are constant, so they still happen every now and then. This election, though, is happening once. It isn't going to repeat until Trump eventually wins; his low chances mean, funnily enough, that he's really, really unlikely to win this. Yeah you play poker enough and you'll get a royal flush. Also I don't think you can compare poker and election statistics. Football maybe since there's predetermined factors involved such as how lovely the 49er defense is.
|
|
#
?
Oct 24, 2016 08:25
|
|
|
hexenmexen posted:
Ah yes, the leaked video of helicopter pilots laughing while gunning down a van full of children and reporters among other things vs. leaked email about risotto recipes and someone calling Chelsea Clinton spoiled. Hmm yes, yes, these things are about the same. Tom Guycot fucked around with this message at 08:29 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 08:26
|
|
|
Having played a lot of poker myself, Trump winning would be akin to having played poker for years and years and suddenly losing because you didn't know there's actually a fifth suit. I.e. everything you thought you knew about the system is suddenly wrong.
|
|
#
?
Oct 24, 2016 08:34
|
|
|
Antti posted:Having played a lot of poker myself, Trump winning would be akin to having played poker for years and years and suddenly losing because you didn't know there's actually a fifth suit. I.e. everything you thought you knew about the system is suddenly wrong. How so?
|
|
#
?
Oct 24, 2016 08:37
|
|
|
I mean there are a number of things going on that could potentially cause polls to be off in ways they previously weren't in elections. Cell phone users, high undecided, non-traditional voting blocks. These could tip either way but they probably make 2016 more volatile than previous elections.
|
|
#
?
Oct 24, 2016 08:43
|
|
|
It would mean polls showing an average of a 7 point lead and beyond-MOE leads in enough states to hit 270 and a GOTV and a money advantage, none of them actually matter. Everything we know about elections and polling would have to be wrong. I'm not saying it's as unlikely as the obvious absurdity of an old game suddenly having different rules, but I'm trying to use an imperfect analogy to demonstrate the mental space you have to be in here. Because this is not a single hand of poker where Trump is drawing to an inside straight. Bip Roberts posted:I mean there are a number of things going on that could potentially cause polls to be off in ways they previously weren't in elections. Cell phone users, high undecided, non-traditional voting blocks. These could tip either way but they probably make 2016 more volatile than previous elections. The world is not that different from 2012. As per Sam Wang this is actually a fairly stable election. We've actually wound up exactly where we were in February, and I think in retrospect this constant freaking out will be seen as charming and nostalgic.
|
|
#
?
Oct 24, 2016 08:46
|
|
|
Antti posted:The world is not that different from 2012. As per Sam Wang this is actually a fairly stable election. We've actually wound up exactly where we were in February, and I think in retrospect this constant freaking out will be seen as charming and nostalgic. I'm not freaking out here, I think the most likely systematic polling error is from the ground game mismatch between Clinton and Trump pushing her into landslide territory. The trump silent majority appearing from the woodwork to save the day is remotely possible but I don't think it's nearly as likely.
|
|
#
?
Oct 24, 2016 08:51
|
|
|
wizard on a water slide posted:The difference for me is less in the content itself - I have no doubt that Clinton may be guilty of war crimes or something, a possibility I leave open for any other US President or Sec State as well - and more to do with the source. Julian Assange is a bigot, a propagandist pretending to be objective, probably a sexual predator, and almost definitely in bed with Russia, none of which is true of Manning (AFAIK). That all makes it difficult for me to find him sympathetic or credible. Chelsea Manning is far from an angel as well. The messenger in that situation is heavily flawed.
|
|
#
?
Oct 24, 2016 08:52
|
|
|
Roland Jones posted:The thing is, in things like poker or football, opportunities for rare things are constant, so they still happen every now and then. This election, though, is happening once. It isn't going to repeat until Trump eventually wins; his low chances, appropriately enough, mean that he's really, really unlikely to win this. You need to go study basic probability if you think the number of times you roll the dice has an impact on what the dice may show on a single roll. Doesn't matter that the election happens once. The odds are the odds. Antti posted:It would mean polls showing an average of a 7 point lead and beyond-MOE leads in enough states to hit 270 and a GOTV and a money advantage, none of them actually matter. You're making the assumption that the election is being held today here. It's not being held today. It's being held in 14 days, and things can happen in that time. The percentages you're trying to quibble with here are actually attempting to the account for the fact that there's still 2 more weeks of campaigning. Hillary's lead will most likely persist until election day. However, presidential races usually tighten a bit. New information can come to light that can cause the numbers to swing. There are things that can happen. Trump is not a normal candidate, but we have no way to quantify how that might affect this presidential race since it hasn't happened yet. There's no way to build a model that takes that into account. So the current models are biasing towards the status quo, and that might be making Trump's chances look rosier than they are for real. There's no way for us to know how much, though. It's possible Trump could win. It would require something to happen to shift his polls to within MoE of Clinton and then for Trump to outperform his predicted turnout while Hillary underperforms hers. The models are accounting for something like that scenario happening. As we get closer to the actual election the probability space will tighten and you'll see a lot of projections have less uncertainty. ErIog fucked around with this message at 09:02 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 08:55
|
|
|
Bip Roberts posted:I'm not freaking out here, I think the most likely systematic polling error is from the ground game mismatch between Clinton and Trump pushing her into landslide territory. The trump silent majority appearing from the woodwork to save the day is remotely possible but I don't think it's nearly as likely. The dead vote Democratic though.
|
|
#
?
Oct 24, 2016 08:56
|
|
|
Even Nate's now-cast has way too much improbable poo poo built into it, as if this election will somehow revert to "generic candidates, electorate votes based on economic factors", which is just lol
|
|
#
?
Oct 24, 2016 09:00
|
|
|
ErIog posted:You need to go study basic probability if you think the number of times you roll the dice has an impact on what the dice may show on a single roll. Yes, and the odds are <5% on some non-538 things, with 538 being deliberately (and even by Nate's own admission bordering on unnecessarily/unrealistially) conservative. The post I was replying to said that their seeing rare things happen in poker fairly often meant we shouldn't discount the chances of Trump winning, but the situations aren't remotely equivalent. Just because you may have seen many 1% poker hands in your life doesn't mean that in this particular game you're going to see a royal flush, it just means you watch/play a lot of poker. Trump's chances are tiny. He could win, but he almost definitely won't; comparing this to football or poker or whatever doesn't work. Roland Jones fucked around with this message at 09:05 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 09:01
|
|
|
WeAreTheRomans posted:Even Nate's now-cast has way too much improbable poo poo built into it, as if this election will somehow revert to "generic candidates, electorate votes based on economic factors", which is just lol The nowcast is a toy model for the record. It's meant to work more like a tracking poll than be predictive.
|
|
#
?
Oct 24, 2016 09:02
|
|
|
Bip Roberts posted:The nowcast is a toy model for the record. It's meant to work more like a tracking poll than be predictive. Nate is a joke statistician
|
|
#
?
Oct 24, 2016 09:03
|
|
|
Bip Roberts posted:The nowcast is a joke model for the record. Nowcast is literally the same model as Polls-Only they just say the election day is today instead of November 8th. It's not really a joke model, but it's of limited usefulness the further out you are from an election. ErIog fucked around with this message at 09:06 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 09:04
|
|
|
Trump has 7 deuce off suit and he's going all in on a preflop and everyone is calling because they want to push out Hillary who's big blind
|
|
#
?
Oct 24, 2016 09:09
|
|
|
I'm pleased arzy'ing has moved on to the level of "well, its theoretically possible Trump could still win!!"
|
|
#
?
Oct 24, 2016 09:09
|
|
|
gfsincere posted:Chelsea Manning is far from an angel as well. The messenger in that situation is heavily flawed. Yeah but she's suffering from psychological issues and being a trans person in a transphobic society. Assange is just a rapist prick.
|
|
#
?
Oct 24, 2016 09:11
|
|
|
Tom Guycot posted:I'm pleased arzy'ing has moved on to the level of "well, its theoretically possible Trump could still win!!" It's not really Arzying it's just people unloading their hot takes about Nate Silver in order to make themselves feel smart without having to do more than compare PEC to 538 and see that 538's projection is more conservative than others and recall that Silver didn't think Trump could win the GOP primary. A lot it is literally just, "well... 538 says he has a 14% chance, but I know Hillary's gonna win so I just don't understand why it's not 100% to 0% I don't understand why Nate doesn't just manually set Hillary's chances to 100% herp derp he should retroactively remove the LA Times and IBD polls because I don't like what they say"* Most of their quibbling could be answered if they would actually read something about Nate's model, but nah, guess it's just easier to assume Nate built in the chance of Hillary being a lizardperson and it coming out in a Wikileak a day before the election. It's dumb herd mentality and I'll be glad when the election is over so people can stop pretending they know anything about statistics or probability. *Interestingly they never bitch about the C+12 polls that are statistically just as likely to be just as wrong. ErIog fucked around with this message at 09:30 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 09:12
|
|
|
Tom Guycot posted:I'm pleased arzy'ing has moved on to the level of "well, its theoretically possible Trump could still win!!" The Celestial Scribe was inside of us all, all along.
|
|
#
?
Oct 24, 2016 09:12
|
|
|
Tom Guycot posted:I'm pleased arzy'ing has moved on to the level of "well, its theoretically possible Trump could still win!!" Yeah, that's basically my point. Just because he can win, and you've seen things that are technically more unlikely to happen (because you want a lot of things so there's constant chances for said unlikely things to occur), doesn't mean he's at all likely to actually win, and silly comparisons are, well, silly. If you're worried, go vote or support the Clinton campaign in your area or something. Don't convince yourself that because you once saw someone make a three point shot in the last ten seconds of the game, steal the ball, then do it again to make a miracle come-from-behind victory, or whatever weird comparison you want to make, Trump's seemingly-bad chances here are actually a lot better than the numbers say they are. That's just worrying yourself unnecessarily. Roland Jones fucked around with this message at 09:14 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 09:12
|
|
|
I've never lost faith in the Supreme Abuela.
|
|
#
?
Oct 24, 2016 09:17
|
|
|
ErIog posted:It's not really Arzying it's just people unloading their hot takes about Nate Silver in order to make themselves feel smart without having to do more than look at the Upshot page and see that 538's projection is more conservative than others and recall that Silver didn't think Trump could win the GOP primary. It's not even that. It's that Nate has a perverse incentive to pursue his strangely conservative model because it promotes the horserace and drives site visits. Nate is also not a specially gifted or qualified statistician despite the myth that has grown up around him, and his attempts at punditry lack insight and objectivity. That being said, come election day he will get most of his "picks" just right
|
|
#
?
Oct 24, 2016 09:17
|
|
|
What do you mean by "pursue his model?" He built the model like 5 months ago and hasn't touched it. It's not like he's banging on the code every day or tuning it. All they do is add polls to it and do analysis with the results. loving with it would open it up to results-oriented biasing. This is precisely why he hasn't removed the LA Times tracking poll since the model can route around it via house effect calculations and averaging. Also, your claims of, "He's trying to drive the horserace," are way off base considering the editorial content of the site has been, "Hillary's gonna win this, but we just don't know how much yet" for the past 6-8 weeks. All the analysis of the model itself have been attempts to explain why the model is too conservative. Also, I'm not trying to defend Silver as some kind of genius. It's just that people focus solely on him when there's all sorts of other models they could also be making GBS threads on. For some reason Silver is the one they force themselves to pay attention to. Nobody ever mentions the Upshot model. They'd rather bitch about the one that's on ESPN rather than the one that's on NYT. 
ErIog fucked around with this message at 09:44 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 09:20
|
|
|
Didnt Silver also have a weird economic feedback problem with his model? I seem to recall he weights economic data in his modeling because he believes it's more predictive. When the polls nationally were inside the margin of error for Trump being elected the fact that Trump was more likely to be elected meant an economic disaster was more likely in November which made it more likely Trump would be elected. I want to say he admitted in the podcast when Sam Wang called him out on why his model was acting so strange in late September.
|
|
#
?
Oct 24, 2016 09:37
|
|
|
Dick Milhous Rock! posted:Didnt Silver also have a weird economic feedback problem with his model? I seem to recall he weights economic data in his modeling because he believes it's more predictive. Economic indicators are only taken into account for Polls Plus. So it would not have affected the default Polls Only view or the LolCast. I've also listened to all his podcasts, and while I do remember him mentioning a thing between him and Sam Wang I don't remember him saying anything about a feedback problem. I'd definitely listen to that podcast if you can find it.
|
|
#
?
Oct 24, 2016 09:41
|
|
|
ErIog posted:What do you mean by "pursue his model?" He built the model like 5 months ago and hasn't touched it. It's not like he's banging on the code every day or tuning it. All they do is add polls to it and do analysis with the results. quote:pursue It should be pretty obvious that I'm using the latter definition, but there you go. And it's completely wrong to say that the editorials are focused on why the models are too conservative. There will usually be a couple of sentences, maybe, which follow a sort of "oh this could be overly conservative, but that's because our special model just has too much top-secret poo poo built into for you to comprehend". You completely missed my point anyway - it's not that Nate doesn't genuinely believe in what he's put together, or that the model is statistically bogus, it's just that it errs towards "status quo" results in a way which just so happens to dovetail with the media narrative (which has now fallen apart of course), and has the perhaps-unintended consequence of increasing clicks for 538, their owners, and their advertisers. Again, most of his calls are going to come in just fine. He's fine at aggregating polls. He's just not much of a statistician or a serious political pundit. WeAreTheRomans fucked around with this message at 09:51 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 09:48
|
|
|
So is your contention that Nate' s just supposed to go, "Model's canceled! Sorry! It's not exactly the same as PEC so gently caress it!" I just don't understand what you want Nate to say at this point, but it doesn't really matter since it doesn't seem like you even read the site. WeAreTheRomans posted:He's just not much of a statistician or a serious political pundit. Wow, you're a person who really notices things! The sportsball analyst who works for ESPN might not have his finger as squarely on the pulse of the election as dude from Princeton. Well, why I never! ErIog fucked around with this message at 09:53 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 09:51
|
|
|
Tom Guycot posted:I'm pleased arzy'ing has moved on to the level of "well, its theoretically possible Trump could still win!!" I had an Arzyesque dream last night, in which I checked Nate's page and Trump was up 52-47 or so. I woke up and just laughed it off.
|
|
#
?
Oct 24, 2016 09:53
|
|
|
ErIog posted:You need to go study basic probability if you think the number of times you roll the dice has an impact on what the dice may show on a single roll. That's true, but you're trying to extrapolate anecdotes based off repeated sampling to a one-off event. Each independent poker hand may only have a 1% chance of seeing your particular event, but over a set of N poker hands the chances that you see that 1% event increase to the point where you have a 50/50 to have experienced the event after about ~70 poker hands. This election, with these particular people and these particular parameters only happens once.
|
|
#
?
Oct 24, 2016 10:00
|
|
|
ErIog posted:So is your contention that Nate' s just supposed to go, "Model's canceled! Sorry! It's not exactly the same as PEC so gently caress it!" 538 is the most popular projections site. Criticizing it for having a limited understanding of probability and that limited understanding influencing the horse race is completely in bounds. straight up brolic fucked around with this message at 10:05 on Oct 24, 2016 |
|
#
?
Oct 24, 2016 10:03
|
|
|
Billy's not happy about the USC/LAT poll either  https://twitter.com/mitchellvii/status/790475613090045952
|
|
#
?
Oct 24, 2016 10:07
|
|
|
Did CelestialScribe ever end up toxxing for Trump?
|
|
#
?
Oct 24, 2016 10:08
|
|
|
canepazzo posted:Billy's not happy about the USC/LAT poll either this is it, the most pathetic tweet
|
|
#
?
Oct 24, 2016 10:14
|
|
|
ErIog posted:What do you mean by "pursue his model?" He built the model like 5 months ago and hasn't touched it. It's not like he's banging on the code every day or tuning it. All they do is add polls to it and do analysis with the results. A small point, but I disagree that 538 has had a "Hillary's gonna win" angle since August. I mean, he was writing stuff about Trump's "1 in 4" chances really recently. The model is the model, conservative as it is, but there are always ways to spin it, and 538's bias is definitely more like "It looks like Hillary's winning now, but..." Remember the "Hillary's Ahead like John Kerry in July" piece? And that article, like many others, ended with a "we'll know more in a month" addendum. And then a month goes by and it's like "well, Hillary's still fairly ahead, but..."
|
|
#
?
Oct 24, 2016 10:18
|
|
|
|
| # ? Jun 12, 2024 06:41 |
|
|
ErIog posted:Wow, you're a person who really notices things! The sportsball analyst who works for ESPN might not have his finger as squarely on the pulse of the election as dude from Princeton. Well, why I never! Many people are treating him like he does, that's the whole point you dummy
|
|
#
?
Oct 24, 2016 10:20
|
|
