Prediction Markets Aren't Casinos, but 90% of Volume Comes From Sports
Kalshi traded four years of compliance for 90% market share, but its revenue structure looks more like a sportsbook's; what's genuinely underrated are the quiet markets like FDA approvals.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Comply for four years first, then open your first market
Kalshi's two founders wrote "regulatory first" into the company's principles and spent four years getting regulatory approval before launching their first market. Mansoor calls this the anti-Silicon Valley model: everyone else ships the product and patches in compliance later, while they went to the federal government first and asked, "how should this be regulated?" The cost was that for the first four years they wanted to quit almost every day, held together by "the cost of regret is too high." The payoff came in 2024, when a federal appeals court ruled Kalshi could list election contracts, landing before the presidential election. Mansoor attributes their 90% market share directly to that path: institutions would rather plug into a platform inside the system, because changing the rules from outside the system is harder.
— Tarek MansourPrediction markets force you to shut up, not to take a position
Robin Hanson lays out three mechanisms that make prediction markets more accurate than polls or expert interviews. First, people being interviewed have an incentive to be interesting, dramatic, and to tell the questioner what they want to hear, whereas the market only rewards being right. Second, the market gives you an incentive to shut up about topics you don't understand — you only bet on the handful of markets where you genuinely have an edge, while interviewees answer confidently on things they don't know. Third, the market doesn't require you to have a direct opinion on a topic; it invites you to find mispricings and errors in other people's pricing, and pays you to correct them. He closes with the old poker-table line: sit down, look for the sucker, and if you can't find him, the sucker is you — so don't play that market.
— Robin HansonInternal prediction markets die on the project owner's excuse
Eli Lilly ran an internal prediction market in the early 2000s that successfully predicted which drug compounds would pass Phase III trials; Google ran one too, called Google Profit. Both were considered effective, and neither survived. Hanson's explanation: the project owner wants to know whether they'll deliver on time, but what they want even more is a good excuse if they fail. And the best excuse is precisely "everything was going fine and then at the last minute something nobody could have foreseen happened." A market that keeps accurately predicting "you will be late" voids that excuse in advance, so nobody wants it to exist.
— Robin HansonDomain expertise is not the key skill for forecasting the future
Mansoor cites Philip Tetlock's Superforecasting research: put experts in a given domain — geopolitics, say — up against a group of non-experts forecasting the same set of events, and the non-experts win. His explanation is that domain experts become too doctrinaire over time, while people who can step back and look at a problem coolly are better calibrated and less biased. Mansoor says this is a fairly heretical claim in politics, academia, and financial analysis — it means that if you want a more accurate answer, you should look for a diverse group of participants who pride themselves on self-calibration, critical reading, and filtering information, not an authority in some field. He treats this as Kalshi's underlying logic.
— Tarek MansourOne contract takes a month to write, for grammar and apostrophes
Nicole Kagan writes the contracts and self-certifies them with the CFTC. The most complex category is "what someone will say in a given speech," and they recently rewrote the rules because a pile of grammar questions came up: what if there's an apostrophe? If they misspeak or mispronounce it, does it still count as the same word? What if it appears inside a proper noun? What if the livestream cuts out midway and the video is uploaded somewhere else afterward? Those rules took over a month to write, ran seven or eight pages, and are extremely detailed. When Bloomberg reported on them, it guessed they were built for automated processing and must be for AI; Kagan says that's not it at all — they wrote them so users could understand, and never considered that "something a human can understand is probably something a computer can understand too."
— Nicole KaganWhen you can't call a winner, settle at the last fair price
Kalshi's contracts are usually binary: pay $1 if the event happens, $0 if it doesn't, with every position fully collateralized — if one side stakes 30 cents, someone must stake 70. But some situations can't be judged yes or no. Cardi B was clearly dancing during her seven-second Super Bowl halftime cameo, but she had no mic and there was no way to confirm whether she was singing, so the team said "we are epistemologically uncertain" and settled at the last fair price. Another case is 50-50: when multiple people win simultaneously, each settles at 50 cents. Kagan says they have the right to do this when it's "genuinely unclear" whether an event meets the criteria, and they've already revised the halftime show contract — the new rules require an audible lead or backing vocal, and under them Cardi B did not perform.
— Nicole KaganNinety percent of volume comes from sports, and that's the most uncomfortable fact
Dubner puts it directly: roughly 90% of Kalshi's volume is concentrated in a category barely mentioned until now — sports, the business that makes DraftKings and FanDuel nervous. Mansoor's response is that sports' share is falling fast and other categories are growing faster, but he concedes that "10% is also a lot, don't forget we're very large." He insists on distinguishing gambling from trading: all derivatives markets are mostly speculative, speculation provides liquidity, and the key question isn't whether speculation exists but what the business model is — one model hunts for losers and bans winners, the other pursues price discovery and wants smart people to show up. He also concedes the second model carries risk, saying there is such a thing as "responsible gambling" and "responsible trading," and that platforms have a responsibility to monitor.
— Tarek MansourSeventy to eighty percent of active users never place a trade
Mansoor says one of his favorite numbers is this: 70% to 80% of active users don't trade. They just log in every day, keep it next to X and a couple of newspapers, and use it to calibrate or check information. Kalshi's business model is a neutral platform, like the NYSE or Nasdaq, taking an average 1% on trades and not caring who wins or loses. He also says the information environment is getting more polarized and the signal-to-noise ratio is falling, because social media algorithms reward clickbait: write a nuanced, restrained, multi-paragraph take on X and you get four likes; write something wildly outrageous and you get tens of thousands. Prediction markets have exactly the opposite incentive — the nuanced, boring, well-calibrated judgment is the one that makes money.
— Tarek MansourIn their own words · checked verbatim
What Luan and I decided is we're going to abide by a core principle in the company, which is regulatory first. And we spent four years getting regulated before we launched a single market.
Tarek Mansour8:16
There's this old saying in poker, when you sit down to a poker table, look around and find the fool. That's who you'll be making your money off of. If you don't see the fool, it's you. Walk away.
Robin Hanson15:36
If you run a project, you want to know if you'll make the deadline, but you more want to have a good excuse if you fail.
Robin Hanson20:48
It kind of means that domain expertise is not one of the most important things for being a good predictor of the future. That's a fairly heretical thing to say in some circles.
Tarek Mansour22:58
We wrote them for user clarity and didn't even consider the fact that if it's clear to a person, it's probably clear to a computer too.
Nicole Kagan33:28
If you write a well-nuanced, well-measured, multi-paragraph take about something, you get four likes on X. If you write something extreme and off the rails, you get tens of thousands of likes.
Tarek Mansour45:00
They tell you we don't trade in financial markets because it's gambling. And it's like, what do you mean by that? It's like, well, the game is rigged against us. There is no shot for us to beat the Wall Street hedge funds, et cetera. Whereas here, I can do research.
Tarek Mansour46:02
Figures
| Kalshi monthly volume (when Kagan joined) | about $300 million | 24:14 |
| Kalshi monthly volume (current) | about $14 billion | 24:14 |
| Kalshi platform fee | 1% on average | 44:00 |
| Share of Kalshi active users who don't trade | 70% to 80% | 44:00 |
| Sports share of Kalshi volume | about 90% | 46:02 |
| Kalshi total weekly volume | about $4 billion | 45:00 |
| Iowa Electronic Markets trader cap | $500 per person | 12:27 |
| Share of times the Iowa market beat the polls | 74% | 12:27 |
| Length of the new halftime show contract rules | seven or eight pages | 33:28 |
| Time to write the new halftime show contract rules | over a month | 33:28 |
Glossary
- prediction market
- A market where people bet real money on whether an event will occur, with prices reflecting probabilities.
- CFTC
- The federal agency that regulates U.S. derivatives and prediction markets.
- self-certification
- An exchange submits contract rules to the CFTC, and they can go live if the regulator doesn't intervene.
- MNPI
- Material non-public information; those who hold it are barred from trading on it.
- KYC
- Know Your Customer, the anti-fraud process by which financial institutions verify user identity.
- decision market
- A prediction-market variant proposed by Robin Hanson that uses prices to directly guide organizational decisions.
How to listen
Founders and investors interested in prediction markets, regulatory arbitrage, and information-aggregation mechanisms; anyone who wants to understand Kalshi's business model and compliance path.
The opening history — the Greek oracle, the Andean potato farmers, and the Iowa Electronic Markets — can be fast-forwarded.