Quote:
Originally Posted by HolidayInTheSun
once you admit the polls are kind of wrong, then it's a slippery slope between being off by a couple percent and being off by 10%.
this is where you're wrong imo it is not a slippery slope any more than its a slippery slope to say 'well KC may not cover -20.5 vs the Jets, so they also may not cover -14.5 so they may also not cover -3.5 so they may also not cover +4.5'
We're probably at something like KC -3.5 or something here in 2019 with KC -19.5 or whatever the current line is being the coinflip mark as an analogy, while Hillary was at say KC -13.5 or something the day before the election. Obviously one 'miss' on the cover is far more likely than the other
If the slippery slope argument had any merit, Trump to win the popular vote would be a snap bet at +700 or whatever it is, because 'hey if it can miss by 5% it can miss by 9% just as often' etc
At a certain point events go from being likely, to not very likely, to very unlikely, to nearly impossible and so on. Right now we're at the very unlikely area in terms of a polling error being large enough to overcome the margin of error, whereas in 2016 it was in the 'not very likely' territory
Trump was 25-30% to win in hindsight down 4 or so w the way the electoral college works in 2016 and I made a few error that cycle, mostly being that I split my bet between Hillary to win and Hillary to win the popular vote when the odds were similar at -3xx from memory (I think I got about that price on both EC vote and pop vote can check my spreadsheet if anyone cares)
If Biden to win pop vote wasn't -600 and was say -300, i'd just be all in on that for ~50% of my roll instead but I think the ROI is higher getting -140ish on a -400 shot than getting -600 on a -1500 or -2000ish shot.
The polls missed by 1.5% last cycle nationally and Clinton's position got 2% worse with the Comey letter in the polls in the final days before the election
If Biden was 2% worse off, he's still in a far better position than Hillary, but he would then STILL win with a 1.5% polling error. Right now, the polls would need to be off by over 5% for Biden to lose, so assuming there is no 'Trump declares all postal ballots illegal with no evidence of fraud and supreme court upholds that' in which case we don't have a Democracy anymore, we'd need a historically large polling error for Biden to lose - again, he is leading by outside the margin of error in 270 EVs worth of states (PA WI MI gets him to 270) and then he has fallbacks if he loses one or two of those if he wins FL AZ NC etc etc (obv by GA TX OH etc the election is already well and truly over, same if Trump is winning Nevada or New Hampshire or whatever)
The Margin of error is typically 5% or so in large sample polls (which is basically what a weighted polling average is) with about a 95% degree of certainty. Individual polls suck, polling averages are a lot better than people seem to think they are
If you take a sample of 200 of my bets I probably win at about my 4% roi a bit, by 1000 bets the sample is semi meaningful by 10,000 bets you're prob gonna see me winning at between 2-6% roi because the sample size got meaningful and we got through the variance.
The certainty is slightly lower this cycle than usual specifically due to the possibility Trump will try and cheat by invalidating postal votes - PA has the double envelope rule, and WI has the must be received by close of polls regardless of when sent, so Dems will lose 1% or so in PA there but also they can be received later in PA so they're harder to suppress there. In Wisconsin Dems might lose 1% or so as well with the ruling today but all up Trump's efforts to **** with postal voting haven't been successful, my postal vote got there easily in Florida and incrementally with ballot tracking I don't know of anyone yet whose vote has been 'lost' in a swing state as a registered Dem, and 45% of the 2016 turnout or something insane like that has already voted (haven't checked todays numbers) so the Dems have banked a reasonably large lead so far that can't be taken back or invalidated barring massive cheating/abandonment of Democracy etc in which case we as in the entire west have much bigger problems than losing one wager
Anyone who cares about the Hunter Biden stuff is already voting for Trump, it may swing 0.x% but not enough care about it at all and there's some doubt over the legitimacy of some of it, but even if it's all accurate as told by Fox News, Hunter did some drugs and maybe had Joe set up one improper meeting or something which no one gives a **** about while Trump is in the oval office forcing the secret service to stay at his resorts and raking in millions while he golfs and refusing to release his taxes and all that and Trump has a secret chinese bank account himself and shady ties to Russia and whatnot. No one in either base is going to go 'oh **** Biden/Trump is shadier than I thought better vote for the other guy' and very very very few swing voters care about that this cycle - they care about healthcare, riots, political polarization, Trump's personality/divisiveness and a bunch of other things. Trump's approval rating has been more static at about 40-50 approve-disapprove through his whole term than any other president ever basically because people's minds are made up about him. The base love him, the left hates him, and independents don't like him as much as they did four years ago partially because he had a fairly bad first term and partially because Biden is less divisive and has less of an offputting personality than Hillary did. Also, Democrats were complacent in 2016 assuming a win was assured and are very much not so this year - case in point, I made the effort to vote (and didn't in 2016 only pres year ive missed after I requested my ballot and it didn't come I didn't follow up because I assumed Hillary was a lock) and I have friends voting Dem who skipped 2016 or voted third party etc it's all incremental but the numbers suggest 60%+ turnout as opposed to 55% last time and Trump hasn't grown his base or improved his popularity with independents - he'll get a similar number of votes to 2016, but Biden is going to get more votes than Hillary did. Remember, Trump roughly matched Romney's numbers it was Hillary's vote that declined, Trump didn't unlock a huge new base of support.
No swing voter is going to vote based on who they think is more corrupt this cycle, they're going to vote on 'do I want another four years of Trump yes/no', it's a referendum on the incumbent because Trump has not made it anything other than a referendum on the incumbent, trying to paint Joe as a dementia patient doesn't work with anyone outside the base, because people see the debates and while yeah, he may have lost a step compared to ten years ago he's obviously still got a functional brain as we saw in the debates.
Anyway sorry thats a bunch of ramble but the flaw in your thinking is that overcoming a 4% deficit in the polls by overperforming by 1.5% and losing the pop vote by 2.5% but winning the EC by about 1% in the key swing states is not the same as overcoming a 6% 8% or 10% deficit at all. Remember, the polls can be wrong in BOTH directions too they won't always overstate Democratic support they may overstate Republican support as well.
So first we need there to be a 5%+ polling error, then we need it to ALSO be in favour of Trump not against him (if it's against him he's losing a landslide) then Trump ALSO needs the vote to be distributed correctly in the appropriate states like it was last time if he can get the popular vote within the 3-4% he needs it to be to be live to win the electoral college
If he loses the pop vote by 5%+ he's just not going to win the electoral college. By comparison, if Clinton had 2.5% more of the vote nationally, she wins FL MI WI PA and the election
Again i'm not saying it's impossible Trump wins, but he's a lot closer to 20% to win than the 33% or so the market has right now and Biden is a lot closer to 80% than the 67% or so the market has give or take. The market is wrong, it is undervaluing both fundamentals and weighted polling averages and now that high turnout is a lock given the early voting, Trump is drawing to 'trump cheats' equity or 'historically huge polling error which is many many times less likely than 2016 margin'
I would snap lock in a 'polls are wrong by the same margin as 2016 in Trump's favour' outcome if I could today, and that would lead to a Biden victory. Trump needs a polling error about three to four times the size of the polling error in 2016, or to improve his position by several percent in the averages this week. Neither is likely.
This could be as good as the second best bet of the past five years on a liquid market after May/Mac which was -3xx on a spot that should have been between -2000 and -5000 or so - feel free to laugh at me in a week and change if i'm wrong, but I have a -14x average price on a spot that should be -400 or so and I probably don't have enough on it quite at 10% of the bankroll, but more than 25% or so would be too much
For what its worth my bet size is about 1.5x what I had on May/Mac, but my roll is several times the size of what it was back then so as a percentage of roll it's smaller, but again I was probably too conservative with my sizing then as I am being right now as well (half kelly would suggest 20% not 10% if my calculations are correct re actual odds vs available prices etc)
Saying '2016 proved the polls can be wrong' is like saying 'The Cardinals actually beat the Seahawks by 3 when the spread said they would lose by 3.5, therefore the current betting line is wrong by 6.5 in the same direction on all Seahawks and Cardinals games in the future'
On average, the polls will be accurate, with the most likely result being about the let's say 9% or whatever margin the polls suggest, then 8 and 10, then 7 and 11, then 6 and 12, and so on until eventually yes, the underdog can win a small percentage of the time, just as an NFL +10.5 favourite sometimes but not very often wins a single match (1 in 5-6 times or whatever), but just because they won that one match doesn't mean the spread was 'wrong' by 10.5 or more and will be every time in the future in the same direction
Obviously if Trump is only down 4% or so in the polls on election day he's going to win a LOT more often than if it's 6% which in turn is a lot more often than if he's down 9% and so on just as a +3.5 point dog is going to win more often than a +7.5 point dog and in turn more often than a +19.5 point dog in the NFL
If I had to pick an exact NFL game from this weekend to show where Trump is right now it would be about he's the Giants against the Buccaneers, but if his polling deteriorates by a few percent he becomes the Jets against KC, but if he improves it by 2-3% he becomes the Vikings or Bengals, and if he improves it by 5% or so to trail nationally by 3.5-4 or whatever then it gets a lot closer to a pickem, or if he's polling even nationally he becomes a pretty substantial fav and at that point he's the Packers or maybe even a slightly bigger fav than that
Last edited by SwoopAE; 10-27-2020 at 08:30 AM.