Hacker News Reader

Comment by rootusrootus | Hacker News Reader

It will be interesting to see how we feel about the prediction markets after today. They are making a clear bet on the winner, and if they are wrong, it will be a pretty big hit on their value.

View on HN

If you have a coin and know for a fact that it is either a fair coin or a 60 40 coin, it would take ~100 flips to be 80% sure which kind of coin it is. We will never know whether polymarket’s odds were more accurate than e. g. Nate Silver’s on predicting a single event

It's just statistics. Even if it's 60/40 you could be wrong 4 out of 10 times.

This is true, but go ask Nate Silver how that went for him in 2016. The general public has only superficial familiarity with even basic statistics. 60/40 is perceived by many as a near guarantee. More than enough to make them cry 'stolen election!' if the election does not go the way they want.

Market dynamics are subtly different than statistics of traditional polls, since you need to account for fees and slippage and irrationality. You’d also need to compare this to a poll asking “who do you think will win the election?” which is a different question from “who are you voting for?”

"Subtly different"? It's not even remotely the same thing, all they have in common is that the unit is %. The betting probability is the (expected) chance that a certain candidate wins, while polls measure the % of population that votes for a candidate. A polling lead of 20% would seal the election.

There is already plenty of historical data that would have demonstrated the quality of the predictions. This specific bet, however, could be the worst way to judge the prediction markets given that the key demographics of the gamblers, young risk taking males, are the same that strongly favor the GOP candidate. In that regard, your bet alone demonstrates loyalty which feels like a soft win. Also, when you win if you win, and if you lose, you were still right because you can cite the heavily skewed prediction market as proof the election was stolen.

But how is a prediction market any good if you need to correct for the demographics of the gamblers for each bet?

That depends on what you are using it for, doesn't it? Like anything, you cant turn your brain off and take it as gospel.

It being relevant to a couple bets doesn't mean you need to correct "each bet".

PredictIt is at 53:52 currently, wouldn’t call this clear.

That's interesting, I have only looked recently at polymarket, and it's at 62/38 right now. Robinhood is at 59/43. I wonder why PredictIt is so much closer.

Does predictit allow cash bets? seems like an arbitrage opportunity for a win-win bet.

That's to mobilize people to vote who would otherwise stay home if the gap was bigger

Does it work like that? Wont the house set the odds such that it always win? I.e. the betters set the odds but the initial seed. The house know their book, we don't.

EDIT: Oh. Visiting the site, it seems like it is not gambling? Or have I misunderstood it.

It's definitely gambling (though maybe not legally, in some cases; RH is selling event contracts, and maybe derivatives are not legally gambling, but effectively they sure seem like it...). The payout is $1 if it turns out correct, $0 if not.

They had a huge miss in 2016 and it didn't seem to have a significant long term effect.

I agree, although I read somewhere prediction markets in 2016 had Clinton by a wide margin. I don’t have a reference for that and obviously they would have been tiny markets by comparison to today.

Reply on news.ycombinator.com