artlu 🎩 on Warpcast

Content pfp

https://warpcast.com/~/channel/thomas

0 reply

0 recast

0 reaction

Thomas pfp

@aviationdoctor.eth

I've refined my earlier take on the Polymarket vs. election polls debate by drawing the following Venn diagram. I still believe that comparing the predictive power of polls to that of Polymarket is a category error. My argument has two main prongs: 1/ Polymarket is permissionless, and thus has the potential to reflect not only the voting intentions of US voters (blue circle), but also the preferences (orange circle) and the predictions (green circle) of everyone; all three are adjacent, but distinct concepts. If these were included in a poll, they would require three different questions to capture properly. 2/ Out of every 100 humans alive today, only 4 are American, of which 3 are eligible to vote, and 2 actually turn up to vote on election day. The Polymarket population is vastly larger and more heterogeneous than just the eligible and active US voters who actually get to decide the next election. If that population was included in a poll, it would require demographic segmentation to control for properly.

5 replies

9 recasts

14 reactions

artlu 🎩 pfp

but isn't the argument / pitch to users as well as consumers of the results, that the contract must resolve on a resolution tied to a real-world outcome? not asking "what do you prefer as a human being" but "what do you think a subset of human beings will do within a societal framework" and that markets are complete enough to force the prices to converge to a meaningful measure for predicting results ahead of time? It kind of seems like you're saying, the noise is a feature not a bug. And I don't get it, yet.

1 reply

0 recast

1 reaction

Thomas pfp

@aviationdoctor.eth

Yes, the market ultimately resolves to a real-world outcome. But a spread / delta between the final market’s position & that actual real-world outcome remains until the moment of resolution. In other words, the convergence between predictions and outcomes is an abrupt step function at the time of resolution, instead of being gradual & smooth. For example, the last bet might be Trump 70%, & a minute later the resolution is Trump 100%. That difference comes from the fact that the bets are not 100% representative of voting intentions. The noise includes people like this French trader who was just outed as betting $45M on Trump, enough to sway the market, even though he can’t even vote (and even if he did, he could cast only one vote, not disproportionately as many). So this single market-distorting whale sits in the green circle of the Venn diagram, although he is likely exposed to different electoral information than US voters, or perhaps the orange diagram if he’s a projecting Trump maxi; or both.

1 reply

0 recast

0 reaction

Thomas pfp

@aviationdoctor.eth

By projecting, I mean that nothing stops a political maxi (or group of maxis) to signal artificial support for a candidate through a massive bet that moves the market’s needle. Which is why survey reporting is illegal during the ballot casting period in some countries (incl. France) because the results alone can sway undecided voters. Entry or exit polls also have uncertainty & a step resolution, but arguably to a lesser degree because they specifically sample only people who are eligible to vote, and sample each one exactly once. And the question being asked is who they’ll vote for, not who they believe will get elected (otherwise smaller independent candidates, for example, would never show up in polls, because everybody understands they stand no chance of being elected despite that particular person choosing to vote for them). So those voting polls are a different survey instrument, measuring only an intent. Everything else they don’t measure shows as confounding noise in the Polymarket instrument.

1 reply

0 recast

0 reaction

artlu 🎩 pfp

in classic EMH (strong form, semi-strong form, weak form) all of this can be arbitraged. The existence of a large gap falls under "limits of arbitrage". Arguably crypto has fewer limits--reg, rep, lev, shorting--but appears to exhibit larger, less arb-able gaps. The one limit crypto has more of is Uncertainty, although smart contracts and PM mechanisms are supposed to eliminate that? Any other explanations? If I'm a pooled investment vehicle like a hedge fund, why wouldn't I hunt these dumb, rich, opinionated non-American whales' bets?

0 reply

0 recast

0 reaction