Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Prediction markets have an elections problem

asteriskmag.com

Post reply on HN
Print view View original

151–160 of 201 posts

Re: Prediction markets have an elections problem

#151

Post by somenameforme »

quickthrower2 wrote: post #137
bluedays wrote: post #110

Earlier quoted context omitted.

How does that work? Being a professional gambler, I mean. I'd love to know more about this.

The most open information is about poker. See the forums. Many of the concepts there apply to other sports. The idea that you don’t know if you are any good until you played 10000 hands etc. Trying to make money off lots of little edges. In a sport you probably need to model every game very well, use AI/ML (mostly linear regression though). Honestly it is boring frustrating work IMO.

10,000 hands in poker is nowhere even close enough to get a significant sample, unless you're playing extremely weak players where you just have a girnomous edge. But those sort of games are very uncommon. There were plenty of instances of strong, winning players going through swings that lasted hundreds of thousands of hands. A more reasonable baseline sample would be ~500,000 hands.

Frustrating is definitely a great way to describe the game, and emotionally crushing. Like imagine one is playing a 5/10 no limit hold'em game - so people are generally buying in for about $1000, unless they're from Spain. An extremely good player might have a longterm winrate of around 3bb/100 there. In 100 hands, there's every chance you'll be involved multiple pots, easily for thousands of dollars. And your long-term expectation is about $30 out of all of that. So you'll go through short term swings where you're up tens of thousands of dollars in very short periods of time, swings where you're down tens of thousands of dollars - all to get your $30 per 100. And you never become completely numb to the money, even after playing millions of hands.

Re: Prediction markets have an elections problem

#152

Post by somenameforme »

jrm4 wrote: post #47

There's no problem here at all; except perhaps the wildly absurd fantasy of "perfectly rational (and/or intelligent) markets," which the popularity of e.g. lotteries and casinos immediately disproves. Not much to see here.

The magic of a prediction market with a reasonable sampling of people is that the players don't need to be rational, for things that are publicly decided. Imagine every person who went to vote, also bet in a prediction market based on nothing except who they wanted to win - the prediction market would be right 100% of the time. And it'd similarly be right most of the time, with a good random sample.

That said, even for things that aren't publicly decided - the 'wisdom of the masses' [1] is very real and kind of weird phenomena. If you get people to try to guess how many beans are in a large jar, you'll get some people making really stupid guesses like 100, or a million. Yet when you average out the result over all answers, it tends to be extremely close to the right answer. At least unless you let people collaborate, then the effect disappears and we become stupid again. Kind of explains a lot of things.

[1] - https://en.wikipedia.org/wiki/Wisdom_of_the_crowd#Examples

Re: Prediction markets have an elections problem

#153

Post by caf »

mattmaroon wrote: post #67

I am a former professional gambler and now I don’t gamble much, because I have neither the time nor interest to find good wagers, but what I do is on predictit. Being a great gambler mostly means not gambling until you have an edge, and what he says about and market being less accurate than 538 is true in my experience. I’ve been betting on there every couple years for maybe ten or so, it’s held up well. As many ment…

The article does mention the time-to-settle problem in footnote 9.

Re: Prediction markets have an elections problem

#154

Post by caf »

codexb wrote: post #126
j_maffe wrote: post #53

Earlier quoted context omitted.

The article shows that the average predicted likelihood is frequently illogical. That is not zero-information.

The author claimed that 90% was illogical; that it should have been closer to 98-99%. What the author failed to realize is that there are overhead costs in betting -- 10% rake, 5% withdrawal, etc. that cut into that 98-99% margin.

The author didn't fail to realise any of that, because he pointed all of those things out (and more) in the article.

Re: Prediction markets have an elections problem

#155

Post by mikeyouse »

dataflow wrote: post #144
mikeyouse wrote: post #141

Earlier quoted context omitted.

Seems like fraud to spell out the terms of the bet in plain language and then enforce those terms? The annual rate of increase in real gross domestic product (GDP) in one or more quarters prior to or including the fourth quarter of 2020, as rounded to the nearest tenth of a percent, per the advance, second or third estimate, as published subsequent to the launch of this market for any such quarter by the Bureau of Ec…

Are you seriously claiming the average person's interpretation of "Will the US economy hit 5.0%+ GDP growth by year-end 2020?" would match the "spelled out terms"?

I'm seriously claiming that if you don't read the actual terms of a legal agreement, you shouldn't wager on the outcome of that agreement. The headline statement leaves a ton of ambiguity, so any reasonable person would read the short paragraph that describes in full detail how the outcome will be measured and then bet accordingly. Caveat Emptor and all that.

Re: Prediction markets have an elections problem

#156

Post by dataflow »

mikeyouse wrote: post #155
dataflow wrote: post #144

Earlier quoted context omitted.

Are you seriously claiming the average person's interpretation of "Will the US economy hit 5.0%+ GDP growth by year-end 2020?" would match the "spelled out terms"?

I'm seriously claiming that if you don't read the actual terms of a legal agreement, you shouldn't wager on the outcome of that agreement. The headline statement leaves a ton of ambiguity, so any reasonable person would read the short paragraph that describes in full detail how the outcome will be measured and then bet accordingly. Caveat Emptor and all that.

No. If this was in an advertisement I'm pretty damn sure it would violate truth-in-advertising laws. Clarifications are one thing, contradictions are another. Most "reasonable" people reading it come away with an understanding of it that contradicts the fine print. Which is exactly what happened, and hence why so many people lost money. You don't get to put contradictory terms in a contract then blame it on "caveat emptor".

Re: Prediction markets have an elections problem

#158

Post by caf »

hackerlight wrote: post #87
fulladder wrote: post #35

The author does not understand bid-ask spreads, does not understand distortions created by fees, does not understand that the last trade price is not the same as what you can get at present, and just generally does not understand the realities of trading in markets. This article is a zero-information void.

Also isn't there a small bet size limit on predictit? That would screw up price discovery because a single informed trader can't correct any inefficiencies.

Bet limits are also mentioned as one of the factors in the article.

Re: Prediction markets have an elections problem

#159

Post by caf »

ummonk wrote: post #109

The article briefly mentions causes for market inefficiency such as the time value of money and platform fees, but then inexplicably goes on to ignore them as potential reasons for inefficient markets. Plenty of arbitrage opportunities exist on PredictIt that can’t be taken advantage of either because the platform’s cut of any profit you make on a bet will negate the arbitrage or because resolution is so far off that…

That is not how I read it at all. To me, it reads as "there are likely many reasons these markets aren't operating efficiently".

The time value of money and fees certainly are part of it, but they don't explain things like, for example, the weight of money in the Trump-by-280+-EV contract.

Re: Prediction markets have an elections problem

#160

Post by silisili »

oramit wrote: post #37

The 2020 elections were such a shitshow on PredictIt. So many trades were skewed by partisan bias that it was shockingly easy to make money. My favorite trade that year was for some reason the "market" was pricing in a lower turnout in 2020 vs 2016. I grabbed a strip of the over bets at big discounts. I still to this day don't understand what people were thinking with their bets there. A hugely contested election dur…

I think 2016 was easier wrt odds, because few thought Trump had any real chance.

I made pretty good money just betting on him to win PA and FL.

Post reply on HN