https://outsidetheasylum.blog/prediction-markets-are-not-polls
Important considerations I failed to mention.
Supporting links I should include.
Errors I should fix.
Aesthetic and phraseology changes I could make to improve understandability and approachability.
Anything else.
1,000
3.001) Improve the background/description of public opinion survey design.
2) Polls and prediction markets are trying to accomplish different things in different ways; quantifying public opinion versus conventional wisdom. The participants are different: a random sample of population v. self-selection.
3) Polls are more than "a quick vote based on their gut reaction," and that's why gambling odds, forecasting models, and prediction markets move in response to them.
4) An example would be really helpful, particularly if you weave it through the entire article. Maybe something like an election where a poll comes out where a poll comes out with the Republican candidate up 45% to 42%. An individual trader makes an uninformed bet of $50 on the Republican but prediction markets move the Dem from 62% to 67%. It seems irrational at first but the market as a whole is factoring in electoral trends and how the undecided voters will break. Essentially use the example to explain polls, prediction markets, and their unique purposes.
5) Include a screenshot of the Manifold, Kalshi, or PredictIt interface with a chart of prices over time. This would give uninitiated readers a better sense of how the markets work.
6) The target audience for this article is not clear. If you're looking to persuade skeptics, government officials, or investors, then this line is very bad: "Prediction markets aren't perfect in this respect; some people may be more comfortable gambling than others, and richer people will be more able to affect the probability." Describing prediction markets as "gambling", and advantaging the rich, is very off-message. If Manifold ever ends up before the CFTC, you will be quoted.
7) Yes, people do this:
A simple "proof" that prediction markets must be at least as accurate as polls is that if they aren't, someone could just run a poll. bet on its results, and turn a profit.
8) If you're comparing accuracy, there should be more than a hypothetical. (And I agree with you on the point)
9) It still is unclear what you were trying to do with the article. Bring more people to the various platforms? Persuade skeptics about prediction markets? Trash the polling industry? General education? I'd suggest figuring out the objective and then reverse-engineering what you want to say.
10) You should run the text through Grammarly. It could help with grammar and clarity.
11) Typos:
A) Paragraph 9: change "weight" to "weigh"
B) Paragraph 24: change "spent" to "spend"
C) You use the British spelling of "judgment" in several places; I'm not sure if that matters but wanted to flag it for you.
Minor point, but “if I place positive expected value bets in a lot of different markets, the chance that I lose all of them approaches 0” isn’t necessarily true. E.g. maybe you place EV bets that pay out with probabilities 1%, 0.5%, 0.25%, 0.125% etc. One way to fix this is by adding “each of which individually succeeds with non-negligible probability” to the middle of the sentence.
2nd paragraph change first word to "For" vice "To"
As a justification for the limitations on polls you should point out that almost all polls have their numbers adjusted from the raw scores for various reasons. FiveThirtyEight has had numerous discussions about this in their election prediction articles but most of that seems buried since ABC took over. You can find a ton of information on polling at natesilver.net
Astral Codex Ten has tons of good stuff about prediction markets (mantic mondays)
Consider reducing the number of contractions used
Several times you use "they" and similar when specifically identifying y=would makje the article clearer.
So many great points here but you've buried the lede. I think it's "in a prediction market participants are incentivized to be right." Move this to the first sentence and see where that takes you. This would also help you find a more compelling title.
The whole first para maybe more can go. If I don't already know "prediction markets are growing in popularity" or what one is, I certainly won't care how similar it is or isn't to a poll. I might like to know how they differ from other markets or why they weren't popular before, but those would be different articles.
Agree with previous: who is intended audience? There must be someone out there under the impression that a prediction market does the same thing as a poll but I haven't met any.
Worth bearing in mind but doesn't need direct mention: there are a billion reasons to make a market on Manifold besides wanting to see the answer to a question or attract traders/liquidity. For instance I have some markets that are demographic probes -- even the fact that few or no traders care about them can tell me a lot about my audience here. Another example, I could see someone making a market to raise awareness of something in the question or description.
I think the key point that could be brought out more clearly is that prediction markets allow and incentivize participants to scale their participation according to how reliable their information is. You do mention this, but the "an incentive to be right" section emphasizes incentives to improve their information (as opposed to incentives to only trade if their existing information is good) and the "good predictors outperform bad predictors" section emphasizes the evolutionary argument that participation is weighted toward past success (as opposed to weighted toward present confidence).
The wisdom of crowds applies to some extent, but I think more often, a prediction market works by incentivizing the crowd to stay away while a small number of wise traders sets the price.
With a poll it is one person one vote. With prediction market it doesn’t work like that, you don’t bet much unless you are sure. Those that are confident of their beliefs bet more while if you have no idea you don’t get involved. Thus the market reflects the confidence of the participants in a way that a poll does not.
Another thing I thought...you link to your market at the end of the article, but you don't mention that you get trader bonuses for having people trade on your market, which takes away from the article. Allow me to elaborate: the user flow from a total newcomer to Manifold markets, reading your article, seeing that market being shared at the end just sounds...fishy if you put yourself in their shoes, because they have no knowledge of who you are and your total activity on Manifold and are not likely to glean that in short order. If they knew that this market is really a tiny amount of your total markets and really pretty illiquid in the grand scheme of things, they would see that this is clearly not worth it for you. Overall another way to look at it is, if you believe in what you are writing and want to support your thesis well, make sure you are transparent and don't even give a whiff of it potentially being some kind of ploy for something else, however inconsequential that something else may be, as it leaves room for doubt readers' minds that you may not have anticipated, since hypothetically anyone could read this article since it's on your blog. My approach whenever I write something and link to my own market, is I try to remember to self-deprecate and mention that I do get trader bonuses and/or referral bonuses if people trade on my markets, but that people can also simply sign up and create their own question and do the same.
The article from Jack that you cite does not support the thesis that, "[play money market] accuracy relative to real-money markets have been favorable." Jack's article fails to mention that Manifold markets could have closed after the result came in. The way he is calculating the accuracy is based upon the probability of the prediction market at the time of market close, rather than the average probability over time. So if a Manifold market closed after the election was over and ballots were counted, obviously people were going to jump in and bet toward the outcome, which would have upped Manifold's market score. I do not think this would be true for the real money prediction platforms...correct? Though he does mention that this was just accounting for one election cycle so we shouldn't read too much into it. I think that your main thesis is that prediction markets are different than polls in that they provide a mathematical mechanism for reputation. I would think that there would be piles and piles of research showing this going back to the 1980s by academics, corporations, etc. and I think Jack's article detracts from your point. I do not think Jack was trying to say what you were claiming there in your article, he was not trying to compare Manifold to real money markets anyway, he was trying to compare and contrast prediction markets to polls.
Footnote 7 could have more caveats. I wouldn't dismiss someone if they disagreed but:
didn't find it worth their time to sign up and bet for little gain
were risk averse
were morally opposed to proliferating more ~gambling in society
didn't trust the website
were in a dire financial situation
etc
There are enough caveats imo that I'd soften the statement in the article rather than trying to cover every caveat
Something useful to add would be a summary or conclusion that states something like: People paying for either side of these prediction markets have something to lose (the money they paid for the shares), while polls are, for the most part, opinions. Also, prediction markets sift through uneducated people, as those buying market shares will be motivated to research the topic, again, since they have something to lose. I don't know. I hope it helps!