Thread by @ErrataRob

Since it's in the news, I thought I'd comment why election betting sites aren't as accurate as you think. I mean, yes they are really cool, but there's still nuance.

The reason election betting sites are cool is because it exposes the sheer craptitude of election commentary by pundits. You can spend an hour watching a debate between two pundits arguing whether Elizabeth Warren has a chance, but never do they quantify exactly how much chance.

I work in a field that includes "risk assessment". The thing outsiders don't understand is that if you can't/won't quantify risk, then the statements you make about it are meaningless.

Thus we know with 100% certainty that pundits are full of crap, because they can't or won't putt hard numbers to vague statements. With betting sites, we can put hard numbers on it: Warren's chances of winning the Democrat nomination is 1.2%, at the moment I tweet this.

If you feel differently AND ARE COMPETENT, then now is your chance to make some easy money: bet. If you think Warren's chances are 30%, then bet against people those who think it's around 1%. This is easy easy easy money.

Another quantification is "which candidate is most able to beat Trump". Well, the numbers here say:
1. Warren
2. Bloomberg
3. Biden
4. Sanders

This is calculated by comparing "win presidency" chances vs. "win Democrat nomination" chances.

I should put numbers behind that. A candidate has to first win the Democrat primary and then win against Trump. So combine the two you get their "vs. trump factor".
Warren = 1.16x
Bloomberg = 0.60
Biden = 0.43
Bernie = 0.36

A number greater than 1.0 is (mostly) impossible. Warren has to first win the Democrat primary before winning the general election (barring some 3rd party run). Thus, here chance at winning the presidency can't be larger than her chance at winning the primary.

Arbitragers can make some quick money by selling "puts" on one and buying "gets" on the other, guaranteed to make money regardless of who wins the primary or election. The fact that arbitragers aren't doing this means that this market-based approach is not as efficient as hoped.

There are other reasons why "prediction markets" may not be as accurate as people claim: arbitraging stuff outside the market. If you are one of the 1% rich people Bernie targets, then of course you'll want to bet that he wins.
en.wikipedia.org/wiki/Predictio…

You'll want to be enough that Bernie wins, that you can take your 8-to-1 winnings and use them to pay off the additional taxes he'll levy on you. "Ah, I'll owe and additional $90,000 in taxes if Bernie wins so I'll bet $10,000. Either way, I lose $10,000".

This is called "hedging". It's what market participants do all the time, hedging their bets against foreign currency fluctuations, oil price shocks, labor unrest, and so on. People think of "hedge funds" as a clever way to make money -- they in fact smooth out potential losses.

BTW, this is a good point: the sites charge fees, so perfect arbitrage isn't possible. The imbalance in Warren's numbers may be due to this.

https://twitter.com/dfranke/status/1234964560744001536

Another thing about the accuracy of these predictions is timing. The prediction markets are very accurate the day before the election, but increasingly inaccurate the further in advance you go.

In other words, instead of measuring the charts based upon what they predicted moments before, measure how well last week's numbers predicted this week's numbers:

These are "betting markets" -- the key word being "markets". You can buy and sell bets. Arbitragers can thus bet that Biden's chances will improve or fail over the next week, buy in now on one side, then sell that bet next week -- regardless of who eventually wins.

The upshot is that if somebody in the bar bets you their favorite candidate will win, first check the website, and if the numbers are favorable, take that bet. They are usually given 50%/50% odds on their nonsense, which half the time you can get better odds on the market.

Yes, yes, in those screenshots I may have accidentally outed my secret perversions. Or maybe my actual perversion is that I'm an Internet troll. #SexyVixenVinyl

So if all you care about is getting Trump out of office, who should you vote for in the Democrat primary? The betting markets say "Bloomberg" and maybe "Warren" (her numbers are too small to be accurate). If you have good reason to believe otherwise, well, put your money down.

(yes, yes, I too agree Bernie's chances are better than what the market predicts -- but don't @ me, just go online and bet and make money from your superior understanding).