I've been running a prediction market in Europe for 15 years. The press and the public are getting prediction markets badly wrong. So here are 10 truths from someone who's been doing this since before it was cool.
The core issue: the margin is the problem.
A sportsbook builds its margin into every price. That makes it a bad deal. It also makes it a bad forecast. A price with a fat margin baked in doesn't tell you the chance of anything, and nobody is competing to correct it. The house sets the number.
A prediction market with low transaction fees is the opposite. The price is set by people competing against each other, and the cheaper it is to trade, the closer that price gets to the truth.
So prediction markets have two roles in the world: destroy the sportsbook margin, and help humans prognosticate the future. It's really one job. Kill the margin and you get the forecast for free.
1) Elections for show, sports for dough
In golf you drive for show and putt for dough. Prediction markets are the same: election markets for show, sports for dough.
All the energy, interest and ultimately trading volume is in sports. Trading on events is very cultural. People prefer to bet on things they understand, culturally and rules-wise. Everybody knows how an NFL game works.
Elections and other political markets, with a few exceptions, aren't regular; the rules change, and the parameters of the bet can change. There is real prognostication value in election markets. They'll just never be the economic engine.
2) Most people don't want to trade
Bids and offers are confusing to most people. People know and like simple sportsbook interfaces.
The category is blowing up in the US because it's the first way a lot of people can bet on sports at all, in states without legal sportsbooks and for 18- to 21-year-olds. Not because there's an unmet need to trade the US midterms.
That's the demand. The question is who serves it best. The customer wants a sportsbook experience, but they deserve exchange prices. Those two things aren't in conflict. The exchange is the engine, not the interface.
3) Transaction fees will trend toward exchange levels
The fee model at one of the leading US prediction markets works out to 7% of your expected winnings, charged on every trade, win or lose. The established rate on European exchanges is 2% of net winnings, and only winners pay it. The US customer is paying 3.5x more, even worse if the customer decides to trade out and will pay fees a second time.
On a $1,000 stake:
| Price | 7% of expected winnings, always paid | 2% of net winnings, expected cost |
|---|---|---|
| 10% (longshot) | $63.00 | $18.00 |
| 50% (coin flip) | $35.00 | $10.00 |
| 90% (favourite) | $7.00 | $2.00 |
It's priced around the same as a sportsbook margin. And a fat fee doesn't just cost customers money. It makes the price worse. You can't claim to be the truth machine and charge sportsbook-adjacent rates. This will get competed away.
4) Paying market makers is not sustainable
In a zero-sum game, having price makers is of paramount importance. We know. We've built a market-making business for over a decade and run one of the world's largest institutional event market makers.
But there's a difference between paying people to show up and having people who are good at pricing. Paying market makers with kickbacks and incentives can work in the short term. It always ends badly because the incentives become misaligned: the venue wants depth, the maker wants the subsidy.
Market making is a skill, not a subsidy. The liquidity that lasts comes from people who make money pricing well. There is a graveyard of prediction markets that have tried this approach.
5) Paying brokers for distribution is not sustainable
Routing volume through FCM brokers and paying them a cut of every contract is a temporary lever for buying distribution. It works until the broker realises it can keep the whole fee.
That's already happening. Robinhood launched its own exchange this summer and moved its World Cup contracts onto it. Rented distribution leaves, and the biggest brokers will become exchanges themselves.
6) Liquidity will pool to a few winners
Exchange dynamics dictate that you won't have 10+ exchanges spread out. Liquidity attracts liquidity. There will be only a few winners left standing.
In the UK, where this is 20 years old, there are 4 exchanges, down from 50. The US now has more than a dozen licensed venues. Most of them won't exist in five years.
7) A price is a probability, not a prediction
If you recall Who Wants to Be a Millionaire, the most valuable lifeline was always Ask the Audience. There really is wisdom in the crowd, and there's no better mechanism to aggregate it than a prediction market.
But markets aren't "right" or "wrong". There's a popular narrative that the markets got this election right or that referendum wrong. A price is a snapshot of the chance of something happening. If something is priced at 20%, it should happen 1 time in 5. If there's a 20% chance of rain tomorrow and it rains, the forecast wasn't wrong.
So judge a market over hundreds of events, not one. Do that and markets beat the other ways humans guess the future: polls, experts, or guessing. With one condition. The market has to be fair, cheap to trade, have a diverse set of traders and enough volume. Take any of those away and the price is just a number.
8) The venue is responsible for the market
People cheat on their taxes and at golf. It would be naive to think they wouldn't try to cheat in prediction markets. Cheating comes in all forms (insider trading, bribes), but with most of the recent insider trading cases, the blame belongs to the venue, not the asset class.
The venue is responsible for the integrity of the market. If a market is open to manipulation, it shouldn't be offered. In my 15 years I've seen all manner of cheating. It happens. But it's well understood by responsible operators; it's heavily regulated internally and externally, and it's policed.
Regulation hasn't caught up. States are trying to regulate intent. The federal government regulates the transaction. It's impossible to regulate intent.
9) Taste matters
The landscape of possible events is theoretically infinite, so it's possible to construct markets that monetise human suffering. Markets on the death of the leader of Iran should not exist. Somebody chose to list that.
Event design is also very complicated. Vague rules and disputed settlements are a design failure, not bad luck. What you choose to list, and how you write it, tells people what kind of venue you are.
10) The role of institutions will be small
Prediction markets are and will remain small compared to traditional finance. The ephemeral nature and relative illiquidity of the markets don't map well to the deep equity and derivative markets. Institutions won't find a rich hunting ground for new alpha. They'll find a high-margin (relative to equities) environment but very low volume.
So who wins?
The winner looks like a sportsbook on the front and an exchange underneath. Simple to use, exchange prices, liquidity from people who are good at pricing, not people paid to show up.
Yes, that's what we built at Smarkets. That's why I believe it.
Prediction markets will eventually eat the margin of sportsbooks. The ones that pay for distribution and pay market makers to trade will die off. A few will be left standing.