Can You Trust Prediction Markets?

Prediction markets want you to think that they're the future of betting, but during the Super Bowl one market was resolved in a way that left everyone furious. People who thought they were right were mad, and there were also other people who saw it a different way, and they were also mad. You might have been one of them, and Kalshi ended up paying both sides. But the weird part is that this was almost destined to happen, and it exposes a much bigger problem. In fact, it exposes five problems that prediction markets need to solve if they want people to trust them. I'll get to that mess in a minute, but if you bet it, you already know what I'm talking about.

And if you didn't, it was a disaster. First though, we need to talk about some fundamental issues with prediction markets and how they affect every bet you make or contract you buy on these sites. The first problem is the most annoying one to users, but it's also the one that's the most fixable, the fees. By design, a prediction market doesn't have any skin in the game. They just run the exchange. In order to be profitable, they take a commission on the trading. Commissions which can sometimes be worse than a traditional sports book. You see, when you bet minus 110 at a sports book on something which has 50/50 true odds, you're essentially paying a 4 and 1/2% fee.

That's what the minus 110 translates to. On Kalshi, you might pay 1.75% in fees. At Polymarket, up to 2 and 1/2%. Now you're saying, "But Jack, that's still less than a sports book." Well, yes, but when you consider there's a market maker on the other side who also wants to make money, it squeezes down the potential for profit. On a true 50/50 play, someone might offer 52 cents on the side you want. That's like a minus 108 bet, and it beats most sportsbooks.

But after fees, you're paying about 52.9 cents on Kalshi. And now you're up to minus 112, and that doesn't beat a lot of sportsbooks. Fees can also be confusing. Sometimes they are folded in if you view in American odds, but not in native implied probability pricing. And sometimes fees aren't included in the displayed price at all. It all depends on the exchange and the market. Fees and their transparency is a problem that needs to be fixed, but it's not even the biggest part of this problem because exchanges blur the lines with market makers.

Now, the CFTC doesn't allow prediction markets to be their own market makers. So, companies like Kalshi have created different entities to market make on their platform. The CFTC allows Kalshi Trading to operate as a market maker on Kalshi. They just mandate that the two are independent entities walled off from each other. Now, I have no reason to doubt that Kalshi and Kalshi Trading comply with those rules. Same for any other exchange that has a trading arm that is walled off. The problem is the fees. You see, Kalshi Trading pays Kalshi's fee structure, but those fees go right back into the coffers that fund Kalshi Trading. Therefore, the market maker effectively operates without fees.

This gives them an edge over everyone else who attempts to market make with fees. They can offer a better price than their competition among makers, and that's bad if you want to try to compete with them as a market maker. Fees are friction, and the more friction, the less efficient the market can operate. Wall Street has always had the same problem. What solved it was competition. Once discount brokers got into the mix, competition kept lowering the floor for commissions in stock market trades. The same should help prediction markets. Competition will keep commissions in check.

But not all of these problems have simple solutions. Take for instance tick sizes. A tick mark is the smallest increment that a market price can move. When you convert American odds to the implied probability pricing that's used by exchanges, you lose the ability to price probabilities between whole cents. For instance, 54 cents equals a 54% probability. But what is 54% probability expressed in American odds? Well, it's minus 117. Stepping down a tick mark to 53 cents is minus 113. Stepping up to 55 cents is minus 122. To a sports bettor, those are some big steps. You can see here how prediction market tick sizes match up against American odds.

It's not as simple though as just making tick marks run to tenths or hundredths of a cent because that only encourages more penny jumping. That's when one market maker offers a price just a tick better than the competition. You should already be familiar with this concept if you ever stayed home sick from school and watched The Price Is Right. $1,200 brings me to Timothy. 1201 1201 Many sharp bettors have made compelling cases for solving this problem. Some suggest breaking down into smaller tick sizes only on the fringes, but the best solution still remains to be found. And that's a smaller problem than this next one on our list. Slippage.

Which if you're not a Wall Street walk, just means buying contracts at a price you want but getting filled at a price you didn't agree to. Let's say that you want a thousand contracts on the Dallas Cowboys at 55 cents. You neglected to check the order book and there's only 500 contracts available at 55 cents, but there are 2,000 available at 57 cents. You'll get your first 500 filled at 55 cents and the other 500 filled at 57 cents. That's slippage. You didn't get the price you wanted. You got filled across what was available. Recently, DraftKings co-founder Matt Kalish exploded onto Twitter with some of the problems that he has with prediction markets.

Now, if you get past the pot calling the kettle a predatory operator, one of the points he made was about slippage. He showed a bet slip where Kalish was showing 93 to 1 on Brooks Koepka to win the PGA Championship after day one, but if you wanted to bet a thousand dollars on it, your entire bet would be filled at a number that totaled 38 to 1. His point is that the recreational user of a prediction market wouldn't know to check the order book to see how much is available at each price. And he's right, but it's also easy to avoid.

Either you check the order book to make sure that you can get filled at the price you want before you bet or you submit your play as a limit order where you get filled at what's available at that price and the remainder is offered with you as the market maker until somebody wants the other side of your number. Even better though, prediction markets could create a slippage catcher that wouldn't allow slippage to exceed X number of cents or just keep it simple and just make all orders limit orders by default. These are problems that are growing pains for prediction markets and prediction market users.

The more people that get used to these platforms, the more likely it is that these issues smooth themselves out. But a problem that will require more thought is one you probably don't even know is a problem. But before we do that, check out this new market that I just set up on Kalshi. Here's what you do. Buy low, click the buttons, and then go back and trade on your inside information. Anyway, like I said, there's a bigger problem beyond tick marks and limit orders. There's been a popular evolution in prediction markets. Kalshi calls them combos. Polymarket calls them combinatorial outcome contracts, but US bettors understand them as parlays.

In an exchange, someone has to make the market and someone has to take the market. Market making parleys might seem like something that's hard to pull off. And that's where LendingTree comes in. If you don't remember LendingTree, it was a service that let you compare loan rates with multiple lenders. It was the early aughts and internet services were still novel and exciting. So was Stone Cold Steve Austin, the iPod, floor tracksuits. It was a better time. Anyway, LendingTree's pitch to borrowers was that lenders could compete for their loans and offer better rates. And their pitch to lenders was that it was a good source of profitable leads. They had qualified borrowers coming to them.

Parlays on prediction markets are like the LendingTree system. It's called RFQ or request for quote. The better signals which markets they'd like to parlay, and then the various market makers, we'll call them providers, get that information and create a quote. Among the information that's sent out to providers during the RFQ process is a creator ID. Now, this is a unique identifier that gets transmitted anytime the user submits a request for quote. Providers can use that ID to build a profile that tells them how successful this user is in their parlays. They could also identify users that are just looking for pricing mistakes in parlays or identify users that are bad bettors.

Offer quotes to the last group and ignore the first group and you've probably got a pretty profitable business. If you're a sharp bettor hunting for edges, you might not want yourself known. But what makes this hard to eliminate is that there are valid reasons why a provider would need to know their counterparty. For instance, a requester who is not looking to actually bet, but rather discover the pricing algorithm of a provider. They could reverse engineer correlation in a same game parlay and then penny jump other providers when the next request comes in. Of course, profiling your counterparty is nothing new. Sportsbooks have always done it.

However, exchanges market themselves as being different than the sportsbook experience, more fair, a better playing field for aspirational bettors. Asymmetric information in a market that strives for efficient pricing is never a recipe for success. It remains to be seen how prediction markets will handle this one. Though even if prediction markets solve all these problems, there's still another issue which almost killed daily fantasy sports. We'll talk about that in a second. But first, remember the Super Bowl?

Here's why it's relevant to this entire conversation. Before the halftime show of Super Bowl 60, Kalshi opened a market on whether Cardi B would be a performer during Bad Bunny's halftime show. Sharp groups often get information about the halftime show, and this year was no different. But what was weird is that there were groups on opposite sides of the question. While Bad Bunny rapped through the show, one shot showed Cardi B dancing in front of the set.

But was she a performer according to Cahlsee's own market rules? Now, those rules state that to satisfy the yes, a performer appears on stage and contributes to a musical performance by singing, rapping, or playing a musical instrument. Did a couple of seconds of background dancing count as a performance? Now, anyone betting the yes was probably counting on the two of them doing I Like It on stage. This wasn't that. Anyone betting on no had their own arguments. The only way she gets on stage is if they do I Like It. He performed that song on at halftime in 2020 with Shakira, so he's not going to repeat the song.

But even Cahlsee couldn't decide. They settled the market reverting to where it had been trading before the performance, paying 26 cents to the yes and 74 cents to the no buyers. Now, if prediction markets can't write clear rules, will recreational users ever buy into the system? And would you? But there's one last problem that's driving people crazy. And it's the marketing. In 2015, Americans were fed up with the proliferation of advertising related to daily fantasy sports. You see, there were two big DFS companies, one with a green logo and one with a blue logo. They claimed their product was legal in all 50 states by way of a federal loophole.

The public sentiment grew more negative because they couldn't get away from the ads and the marketing. In 2018, state-regulated sports betting came about, and those same two companies quickly dominated the space as well. But again, Americans didn't like it always being in their face. Now, we have prediction markets with two major prediction market operators, one green, one blue. The ads and product placement are everywhere again. The engagement farming on social media is out of control. If these prediction markets want to earn the public trust and be a viable way that people assess risk, they need to grow up and be the adults in the room.