Kalshi and the dunk machine
Why do prediction markets keep turning probabilities into ragebait?
As voters headed to the polls in Wisconsin’s Democratic gubernatorial primary on August 11, Kalshi, the prediction market, posted the following tweet:
Rival firm Polymarket gave her similar odds. Later that day, Hong narrowly lost to her opponent David Crowley.
This is funny. But the fact that the prediction was wrong is not the point: even perfectly calibrated 95% predictions will be wrong 5% of the time. While the prediction markets didn’t seem to pick up any additional signal beyond her double-digit polling advantage, this is not what the post is about. Prediction markets position themselves as truth machines, converting financial incentives into unbiased probabilities. And then their social media accounts go out and post things like…this.
It’s not that the tweet isn’t true. Hong really is a self-described Democratic Socialist who has, in the past, shown remarkable hostility toward a number of holidays. But with its sneering tone, the tweet may as well read “here is an annoying left-wing politician saying annoying left-wing things, and look, Democrats are going to vote for her anyway.” The probability becomes the dunk.
This was not an aberration. Both Kalshi and its rival Polymarket routinely turn their markets into material for right-coded political dunks. Here is a tiny sample:
- Kalshi: BREAKING: Tim Walz is 30x more likely to be charged with a federal crime than to be President
- Kalshi: BREAKING: Trump is more likely to become the next US President than AOC - despite being constitutionally ineligible
- Polymarket: BREAKING: Kamala Harris eying 2028 Presidential Campaign. Trump has higher odds getting elected for a 3rd term (5% chance) than Kamala winning (4% chance).
- Polymarket: BREAKING: 75% chance Mamdani doesn’t open a single socialist, city-owned grocery store. Do New Yorkers deserve free groceries?
- Polymarket: BREAKING: Democratic Socialist Francesca Hong now says Thanksgiving is her “favorite holiday” — after previously calling for its cancellation.
- Polymarket: NEW: Gavin Newsom’s final budget proposal as governor seeks $33,000 in taxpayer funds for an official portrait of himself. (This one is tenuously attached to the market for Newsom becoming the Democratic presidential nominee for 2028)
Do these posts advance Kalshi’s mission of “promoting informed decision-making” and “[contributing] to a better-informed public”? Do they help Polymarket fulfill its promise that prediction markets will “fight back against bullshit, clickbait, and propaganda”?
No. Prediction markets are supposed to reward people for being right; Twitter rewards people for getting attention. A trader that repeatedly turns 95% probabilities into certainties will eventually be punished for it. But a social media account that turns 95% probabilities into partisan dunks will be rewarded every time, even (perhaps especially) when it blows up in their face.
As I have alluded to before, I find prediction markets theoretically interesting but perverse in practice. I’ve followed the idea for years, largely through statistician Andrew Gelman’s long-running coverage of the topic. With the rise of Polymarket and Kalshi, prediction markets have finally found a mainstream audience, one heavily weighted toward young men, crypto people, gamblers, and political obsessives. It is no surprise that their social media skews right. But if things were different and they were posting “BREAKING: 95% CHANCE MAGA LUNATIC WINS PRIMARY”, the fundamental problem would be the same: the marketing strategy undermines the mission.