Ed FiveThirtyEight GE Forecast

[quadraginta]

Well, always seemed to me that being betting odds, they should simply be reflecting which side people were putting their money on like any sporting bet. But many posters here have offered that bettors are inherently conservative so will take the smart bet rather than their emotional choice. Seems logical... there's a whole lot more data available in the fifty electoral vote states than there is data on Notre Dame's defense or whether Montezuma's Revenge likes a sloppy track in the feature at Woodbine.

As before, I am generally in agreement with you.

However, regarding the highlighted statement I'm not so sure. If it wasn't for emotion there would be no successful state lotteries.

And if conservatives were less inclined toward emotionalism then Trump would still be a blowhard reality TV personality who was watching all the rest of his wealth evaporate from a string of stupid business decisions.

There's no way they could rationally believe he is the smart bet to put in the White House.

 

[xjx388]

I totally disagree with the idea that bettors are somehow more rational and less emotional when it comes to placing their bets. If this were the case, Las Vegas would be a ghost town because there are very few bets that are rational to make. Sports betting and odds are not reliable indicators of who will win or what the score will be. I will grant that people that bet on sports or election outcomes have more of an interest in looking at the situation rationally but they are, at best, making educated guesses.

 

[Skeptic Ginger]

That's backwards. It's the house that is the rational entity in gambling.

 

[Tony Stark]

That's backwards. It's the house that is the rational entity in gambling.

Because they rig things in their favor. For casino games, the odds are in their favor. For sports (or election) betting, they attempt to make it so equal money is bet on both sides so they win no matter what happens.

 

[Ladewig]

I suppose there's spread-betting on the Electoral College votes? That's where I'd expect the pros to have an influence.

Here's one place to make bets on that

https://www.predictit.org/Market/23...-the-winner-of-the-2016-presidential-election

You don't even need to predict which side will win. All you have to do is come up with your idea of the number of electroal votes cast for the winning party.

You can also bet on which way each state will fall and all the House and Senate seats.

The up side is that you can make bets for as little as two or three dollars. The down side is the vig is pretty steep (in this case, 10% of profits and 5% to cash out).

ETA
the other upside is that if the odds start moving in your direction, you can cash out of the bet and take a small guaranteed profit instead of taking your chances on a higher payout when the bet is resolved.

 

[varwoche]

I'm not sure how 538 could be accused of having left bias ...

Nor the accusations of right bias that certain Bernie supporters foisted hereabouts.

 

[phiwum]

I'm not sure how 538 could be accused of having a left bias; they explain their methodology on their website and how often they adjust it based on new information. Not to mention Nate Silver has a 100% success rate at predicting the last four or five Presidential elections (starting with the first one he ever predicted, as I recall) and something like a 95% or higher success rate at predicting Congressional elections. That to me says he's not paying attention at ALL to media spin and hype and is focusing exclusively on the numbers, based on which polls have been shown to be the most accurate based on their history. Dude's a freakin' statistician, not a pundit or talking head.

I'm not much on statistics, but this statement puzzles me.

I assume that he often forecasts less than a 95% chance for the winning candidate in congressional elections. And yet, the one he gives the advantage to wins about 95% of the time. Doesn't that suggest that his probabilities are off?

If I repeatedly say R has a 75% chance to win this race, then surely we expect R to lose one out of four times. If he doesn't, then my forecast is off.

 

[Foolmewunz]

I'm not much on statistics, but this statement puzzles me.

I assume that he often forecasts less than a 95% chance for the winning candidate in congressional elections. And yet, the one he gives the advantage to wins about 95% of the time. Doesn't that suggest that his probabilities are off?

If I repeatedly say R has a 75% chance to win this race, then surely we expect R to lose one out of four times. If he doesn't, then my forecast is off.

No. He's calling each individual race. If a race is 75% towards candidate X then three out of four times THAT RACE WAS RE-RUN, X would win. But the next time the race is run, it's still 75% in X's favor.

It's like the old 1000-1 odds that you will not get hit by a car when crossing a street. You do not count 999 crossings and put your papers in order. That 1000th crossing is an individual act and the chance of you getting run over is yet again 1000-1.

 

[Foolmewunz]

As before, I am generally in agreement with you.

However, regarding the highlighted statement I'm not so sure. If it wasn't for emotion there would be no successful state lotteries.

And if conservatives were less inclined toward emotionalism then Trump would still be a blowhard reality TV personality who was watching all the rest of his wealth evaporate from a string of stupid business decisions.

There's no way they could rationally believe he is the smart bet to put in the White House.

ETA: You can't discuss "bettors" or "gamblers" and include lottery players.

I totally disagree with the idea that bettors are somehow more rational and less emotional when it comes to placing their bets. If this were the case, Las Vegas would be a ghost town because there are very few bets that are rational to make. Sports betting and odds are not reliable indicators of who will win or what the score will be. I will grant that people that bet on sports or election outcomes have more of an interest in looking at the situation rationally but they are, at best, making educated guesses.

I think we're all having a weird discussion, here. People are expressing their intuition as to why it's irrational to think gamblers would get things right, but the history bears out that they have been getting things right,.... repeatedly.

When the polls called the 2004 election as an average 1.2% for Bush several polls, including Fox, had Kerry winning. Exit polls on election day actually had him winning. Intuitive logic would have that the betting pools would be close, too. They weren't. Bettors on BetFair had 91% of their money on Bush. They also got fifty out of fifty states correct on the individual state pools.

Before there were polls, the bookies were the published predictions. The newspapers had stringers and associate papers in other cities out there getting the odds from the bookies. These were reported to the public in the major papers all across the country. Bettors pegged every election from 1874 to 1940 except one. They bet wrong on Charles Evans Hughes v. Woodrow Wilson. Other than that, they called all of them.

 

[Cain]

Professional gamblers take advantage of sucker money, like in the case of the Smoke-a-Bowl Superbowl between Seattle and Denver. Homers and die-hard, straight-ticket-voting hacks put money on teams/people they want to win, but there's obviously no guarantee that betting markets will get things right. I remember the Reds beat the A's in the 1990 World Series.

If Predictit sells share of Hillary getting federally indicted for twenty-two cents a piece, then a straight-up bet with a fellow forumite is probably a good deal. Still, the people who made a killing on Trump securing the Republican nomination were almost certainly not professional gamblers. They were hard-core true-believers who bought it for cheap. It took awhile for the betting markets to shift to Trump (and it took awhile for Nate Silver as well). Everything is only obvious in retrospect.

Professional gamblers also have stories of when sure-things went to **** (because they always do).

 

[Cain]

When the polls called the 2004 election as an average 1.2% for Bush several polls, including Fox, had Kerry winning. Exit polls on election day actually had him winning. Intuitive logic would have that the betting pools would be close, too. They weren't.

Maybe intuition suggests close polling means the betting would be close, but people can easily see why that intuition is mistaken. Nate Silver's site gives Clinton a 99% chance of winning California, but she's not polling 99% to 1%. The difference in a relatively narrow contest, assuming a preponderance of polls consistently show the race going a certain direction, gets magnified in terms of odds of winning.

Not surprisingly, the polls influence the betting markets, especially sites like Silver's, and I'd bet that when early data came in on 2004 election, if betting was still allowed, it would have moved the line.

 

[Tsukasa Buddha]

"Predicted X of Y" is not a good measure. This tries to compare harder stats.

 

[C_Felix]

Because he shows the Republican candidate as losing in the polls when the Republican obviously is really winning.

Did I miss the sentence, paragraph, section, where Mr. Silver's methods were questioned?

 

[Mader Levap]

Right now, 538's forecast is 76.2% chances of win for Clinton.

I predict this number will start falling just before second debate and will jump few days after it, but less dramatically. I doubt it will ever achieve or surpass record 89.2%, unless Trump actually takes dump in public or something, like some sarcastically said.

As far I am concerned, you all can start practicing to say "Ms. President Hillary Rodham Clinton".

Usual caveat: terrorist large-scale attack just few days before election on USA soil means all bets are off.

Did I miss the sentence, paragraph, section, where Mr. Silver's methods were questioned?

Nope. His argument against Nate's 2012 prediction was "he looks like gay" (therefore his predictions are untrustworthy). Well, what else you can expect from republicans?

 

[bonzombiekitty]

*sigh* all the talk of betting sites and such has gotten the last page of this thread categorized under "gambling" by my work's internet nanny filter.

 

[Tony Stark]

Right now, 538's forecast is 76.2% chances of win for Clinton.

I predict this number will start falling just before second debate and will jump few days after it, but less dramatically. I doubt it will ever achieve or surpass record 89.2%, unless Trump actually takes dump in public or something, like some sarcastically said.

As far I am concerned, you all can start practicing to say "Ms. President Hillary Rodham Clinton".

Usual caveat: terrorist large-scale attack just few days before election on USA soil means all bets are off.



Nope. His argument against Nate's 2012 prediction was "he looks like gay" (therefore his predictions are untrustworthy). Well, what else you can expect from republicans?

I dunno, in 2012, the final number for Obama was over 90%. I think Hillary could easily best that unless the model is significantly different.

 

[maxpower1227]

The now-cast just turned Arizona a tiny bit blue. Wow.

http://projects.fivethirtyeight.com/2016-election-forecast/#now

 

[phiwum]

No. He's calling each individual race. If a race is 75% towards candidate X then three out of four times THAT RACE WAS RE-RUN, X would win. But the next time the race is run, it's still 75% in X's favor.

When I wrote "we expect R to lose one out of four times," I did not mean to invoke the gambler's fallacy.

Let's put it a different way. Suppose each of R1,...,R4 has a 75% chance to win (parallel races on the same day). Then the chance that *all* of them win is just 81/256, or about 1/3.

Now if there are twenty such races, and in 95% the projected winner wins, then your odds are likely off.

(ETA: The odds of either zero or one "failed" predictions out of twenty is about 2.5%.)

Of course, the real situation is somewhat different, and I'm sure that many of the gerrymandered seats have probabilities considerably higher than 75%, but the same reasoning suggests that his projected winner should lose a lot more than 5%.

(ETA: The odds that 19 or 20 out of 20 win their race, given that all of them have 95% odds, is about 38%. Not terribly unlikely. The odds that 95% or greater win their races, given that each have 95% odds, is a bit more of a pain to calculate, involving 22 "n choose k" calculations, but it can be done. If I do it, I will update here.)

It's like the old 1000-1 odds that you will not get hit by a car when crossing a street. You do not count 999 crossings and put your papers in order. That 1000th crossing is an individual act and the chance of you getting run over is yet again 1000-1.

No doubt. I was not making that error.

What I said was that we expect the 75% favorite to lose about one out of four contests. That is the expectation, after all. It does not mean that if he's won three, then he will lose the fourth. If he's won three, then our prior expectation (he will lose one) is likely false, that's all.

 

[Foolmewunz]

Nope. It still breaks down to individual probability of the races. His current national probability for a Clinton win is in the mid-70s.

As we all know, that's based on 51 individual battles, though, and he's got odds for each state. Each race will have a 40 or 50 or 70 or 90 per cent probability.

 

[phiwum]

Nope. It still breaks down to individual probability of the races. His current national probability for a Clinton win is in the mid-70s.

As we all know, that's based on 51 individual battles, though, and he's got odds for each state. Each race will have a 40 or 50 or 70 or 90 per cent probability.

I don't know why you think that changes any of my off the cuff calculations.