It’s unclear to me: How many delegates does your tracker say, for instance, Biden is going to get in Alabama?
That’s not what it’s tracking.
It’s tracking the PredictIt market for Alabama, which is not a market that is is concerned with delegate counts.
Well, then I guess I just demonstrated why polls are clearly and obviously superior to PredictIt,
Not really.
A market concerned with predicting winners has predictive value when you are asking the question, “Who will win?”
A market concerned with predicting delegates has predictive value when you are asking the question, “How many delegates?”
It’s pretty silly to blame one market for not being a different market altogether.
Here’s a question: How many delegates do Alabama primary polls say, for instance, Biden is going to get in Alabama?
As of right now, there are two. One from July 2019 and one from March 2019.
Do the polls or prediction markets give us a clearer picture of what to expect from Alabama on March 3?
Another post gave evidence questioning whether Silver actually thinks that. And if he did think it, I believe he’d just be wrong. He is not infallible.
The ‘scientific’ argument, really just common sense argument is simple. People have less reason to make bets thoughtlessly, which would cost them real money, as compared to answering poll questions thoughtlessly, which costs them nothing. Also the key weakness of polls, especially ones far in advance of elections, is difficulty correlating the answers of the sample of people who respond to the poll to the actual electorate, in terms of ‘likely voter’. Polls far from elections don’t always even try to do that, but are of ‘registered voters’. That doesn’t directly enter into betting odds. The people willing to bet tend to automatically be ones with relatively more knowledge and concern about the election.
Obviously a % bet on a winner tells you nothing directly about the number of delegates a candidate is likely to amass. So if it’s Silver saying a projection of delegate numbers is not based on bets which aren’t on delegate numbers, ‘duh’ on that one I think. 
The big potential weakness in betting odds is that not a really large amount of money is bet in the US. The reason being basically tax treatment. A betting win is taxable income, a loss is not deductible against other trading gains. So it’s basically a mug’s game compared to ‘betting’ on for example the direction of the stock market, a currency’s value etc where a win is taxable but a loss is deductible against other gains. You have to be good at an unlikely level to win out after tax where gains are taxable but losses aren’t deductible.
People who bet on politics therefore tend to be some combination of those with too low incomes for that to matter, not tuned in enough to realize that problem (so you’d wonder how generally tuned in they are), bet non-serious amounts of money compared to their means, foreigners, etc. IOW not the best population to maximize the ‘wisdom of the crowd’ effect you get in financial markets where the most tuned in people bet astronomical amounts of money. And even financial markets are far from infallible obviously, nobody knows the future. But political betting odds still give you useful info and I wouldn’t automatically believe a Nate Silver probability estimate over betting odds (subject to statistical proof I should). Also published odds respond to events more quickly than polls or estimates based on polls. I shake my head more often to posts on this forum quoting obviously stale polls to make their points than ones quoting current betting odds.
A lack of polls is a good reason to turn to the betting markets. But as mentioned above this is a self selected group of people guessing. There is no way the likely 99% white internet gamblers have a finger on the pulse on the Alabam vote. Just don’t buy it.
You are not required to buy it.
The accuracy of prediction markets (on average) is a direct consequence of theefficient market hypothesis. That’s why people pay attention to betting markets. That’s the answer to this thread.
So why don’t you average some betting markets for your twitter bot? Because I see zero worth in tracking one group of gamblers’ bets. Find that extremely tough to view as some “effecient market” or even reflective of one.
I track PredictIt because that’s the market I participate in. The efficient market hypothesis does not require you to view PredictIt as an efficient market in order to be applicable.
If you really think PredictIt is inefficient, there is a financial benefit to identifying and exploiting those inefficiencies. Best of luck.
Yeah, I hearbthat comeback ll the time. The problem is, I think it’s erratically wrong so I can really make money casually assuming it’s wrong in particular direction. I’m sure with study, you could make money betting against emotional/wave sentiment there. I’m not making that my hobby.
@ the OP: People pay attention to betting markets because, more often than not, the majority is correct.
If 87% of gamblers believe the Patriots will defeat the Jets, more likely than not, the Patriots will win. Sure, the Jets could pull off an upset, but the economical money is on the Pats.
Almost all of those are real words, I’ll give you that.
Yes, “Yeah, I hearbthat comeback ll the time.” should be “Yeah, I hear that comeback all the time.”. Otherwise, I wouldn’t understand any other comprehension problems.
But every predictor of the future will be ‘randomly wrong’, at least, because nobody knows the future. The test is whether something is unbiased predictor, or usually is. The second statement ‘I’m sure with study you could make money off it’ IOW is a much stronger claim, one that needs evidence, and you offer none.
As I said above, among market mechanisms one might point out that election betting sites attract a minuscule amount of money compared to say financial markets ‘betting’. Again I believe this is largely because tax treatment of ‘gambling’ (so defined in the US tax code) is punitive. Therefore it might be reasonable, especially, to believe the amounts bet per person are more likely to be amounts the person can easily afford to lose (other than problem gamblers, but they tend to be bad gamblers) and might somewhat weaken the basic argument in favor of market mechanisms as predictors: real money at stake makes people reveal their actual beliefs about reality, and the smartest money tends to dominate, whereas talk is cheap.
However again on ‘worthless’, the people betting on elections can read Nate Silver’s site just like you can. Prices discovery IMO is extremely unlikely to be dominated by people who don’t access easy sources like that, or else again some ‘smart money’ would ‘make it their hobby’ to exploit obvious inefficiencies. So I see little reason to believe betting market answers are worse than Silver’s predictions of the same things in the ‘steady state’, when nothing major has happened since fresh polls. And in a more fluid situation where more polls are stale, betting odds sites are a more valuable source of info than prognosticators who if honest will just say ‘we need more recent polls’.
I’m gonna reread this thread when I’m a little more awake, so I can better absorb some of the great points being made both for/against relying on the betting markets as prediction tools.
That being said, the fact that Trump is still in a literal dead-heat with Biden is gdamned depressing.
The book Superforecasting: The Art and Science of Prediction by Philip E. Tetlock and Dan Gardner addresses this question. I will try to sum up the argument from memory, though I know my explanation won’t be as thorough or accurate as the original explanation.
The idea behind betting markets is that you can tap into the wisdom of crowds. Let’s say you have one tidbit of information that isn’t publicly accessible – for instance, you know that Candidate A’s spouse isn’t totally on-board with the campaign, and if the candidate starts to struggle, the spouse will be pressuring him to drop out rather than encouraging him to double down. That information on its own is of minimal value: perhaps you would adjust that candidate’s estimated chance of victory down slightly, but only slightly.
Let’s say another person knows that Candidate B is in declining health, and that the Candidate will not be able to attend as many fundraising events as his competitors as a result. That’s another tidbit of minimally valuable information. Another person knows that Candidate C is in cahoots with an influential person in the party, and will receive an endorsement and the support of the influential person in three weeks time: that’s another small tidbit.
If you gather together the predictions of enough people, each with access to slightly different information, then your overall picture becomes more accurate than it would be if you only had access to publicly available information. Poll aggregation will show you the public’s personal preferences, but betting markets have the advantage of studying not only what people’s current preferences are, but what factors are in place that would lead to changes in the candidates’ trajectories.
(Spoiler: The book reveals that superforecasting groups have the highest prediction rate of all, so if you want to learn how to make predictions that are more accurate than both poll aggregation and betting markets, I recommend the book.)
There’s no reason to think the average bettor has access to such insights. There’s little reason to think the average bettor on those sites even lives in the same country as the candidate.
Right now, the RCP average of the betting sites has both Clinton and Cuomo ahead of Sanders.
https://www.realclearpolitics.com/elections/betting_odds/democratic_2020_nomination/
What clever insights do you figure they might have that us plebes don’t?
One obvious way that the betting markets are inefficient is that they’re run by companies who are making a profit, and that profit comes out of the bettors’ pockets, and they’re competing against other markets, and so at some point, even a positive-expectation bet becomes negative. Even if I am absolutely certain that Clinton won’t be the nominee, the return on betting against her might be lower than that from other investments that I could make over the same timespan. And even if literally all investors believe that she has literally no chance, they might all have better investments available, too. So even though nobody thinks she has any chance at all, nobody makes the bets that would move the needle in that direction, and so she remains on the betting market lists as having a nonzero chance.
Either I’m misunderstanding you, or you’re misunderstanding me. I’m not saying that the *average *bettor has access to such insights. I’m saying that some do, maybe not a lot, but enough that those insights get factored into the bets that they make. If you take an aggregate of polls, then you are taking an aggregate merely of voter preferences, but if you take an aggregate of bets (which is what a betting market it), then you are aggregating human knowledge rather than just human preference.