Hacker News Reader

Comment by yellowstuff | Hacker News Reader

I looked into the economics of PredictIt a few years back.

* Fees are high. They charge 10% of profits and and additional 5% for withdrawals.

* Markets were illiquid. There aren't enough shares available to put much money to work. There's an $850 limit per bet, but it's hard to reach that amount.

* Bets can take a long time to settle. So even if you are 90% certain of something trading at 50%, you can't put a ton of money into that view, nor can you necessarily get paid quickly.

* Occasionally bets are determined by dumb technicalities. As I recall the bet for Powell being the next Fed chair settled on No, because the date referenced in the bet was a holiday or something, and he was confirmed on the next business day.

If you are smart enough to make significant money on PredictIt aren't you smart enough to make more money doing almost anything else?

View on HN

> Fees are high.

Yep. I was skeptical about if they were beatable initially.

> Markets were illiquid.

Not always true. This varies a lot between markets. If you look at the Democratic nominee markets, you can buy as many shares you like.

> $850 limit

This is per bracket. So you can max No on several candidates in a market if you want, and PI also refunds the amount that is impossible to lose since only one can resolve as Yes. There are also a lot of correlated markets that are effectively the same bet. You can bet on Biden being the nominee at least 4 different ways if you want.

> Bets can take a long time to settle.

You can sell at any point in time. No need to wait. There are long and short term markets.

If there was something 50% that was 90%, I'd have no problem taking that bet. But it's something like a 2020 election you can always wait until closer to the election. Though some markets appreciate during that span of time.

> dumb technicalities

> Powell being next Fed chair

Always read the rules. All predictions are useless without a time.

Powell market was free money. The good players were telling people the correct answer in the comments for months but people didn't want to listen. Amazing.

> aren't you smart enough to make more money doing almost anything else?

That's what I was aiming for before the economic shock, but it's not like people think "Oh you won money at bet on politics? Here is job"

Any advice on how to get better at making predictions?

There's an awesome book about getting better at predictions called "Superforecasting"

https://www.goodreads.com/book/show/23995360-superforecastin...

The idea is to make lots of time-limited, falsifiable predictions with percentage estimates attached and check your track record.

I second stevenwliao's recommendation of Superforecasting by Philip Tetlock. It's a good intro. It might seem unsatisfying in that there is no single thing that makes one good, and he basically refers to it (appropriately) as good judgement. But it highlights a lot of basic areas.

Practice is also good. If you're not used to probabilistic thinking you'll need to develop that intuition and calibration.

Anything about how to think about things better is going to be useful. There's a Coursera course called Model Thinking that is might be useful. Just being curious about things in general and pushing yourself outside of your normal areas of competency/interest.

It might seem weird, but I find Twitter to be pretty essential these days. There are a lot of smart people freely sharing information and some don't mind answering questions.

>If you are smart enough to make significant money on PredictIt aren't you smart enough to make more money doing almost anything else?

Don’t underestimate the value of domain specific knowledge and the costs of lacking it. Just because you know how to count cards doesn’t mean you know how to deal cards, how to run a casino, or most importantly, how to avoid getting caught.

The domain knowledge is freely available for the overwhelming majority of markets but it's usually not enough to provide an edge.

Sure, but you usually don’t know what you don’t know. More pertinent to jobseeking, you don’t know who you don’t know.

Reply on news.ycombinator.com