By Two lenses on one premise
The primitive can be complex. Using it shouldn't be. I wanted to explore that simple premise across two lenses.
Complexity belongs in the primitive. The math and stuff can be as expressive as it needs to be, while the thing the trader actually touches stays simple.
Same underlying, new interactions. Prediction markets are making a lot of previously untradeable things tradeable. But even the things that are already tradeable have massive design space left in them. There's latent demand sitting behind the trading surface itself.
The World Cup was a test of both
We ran a campaign app on our mechanism using fantasy points as the underlying. They're a clean numerical representation of player performance that millions of people already think in. The interesting thing was what you could do with that number once you had a market primitive capable of expressing different beliefs around it. There's a full case study here.
You could say:
"The outcome is above A."
"I'm not sure, but I think it's probably between B and C."
"The number is going to be around D, and the closer it goes to D, the more I want to get paid."
"X is going to outscore Y."
Historically these have tended to show up in different products, different markets, or different ways of interacting with the market. In our case they were four interactions on one market: dynamic over/unders, ranges, precision and duels.
How do they work though? With a dynamic over/under you're not stuck with the line you're given. You slide the line until the odds match your belief.
Figure 1. Dynamic over/under. The line is yours to move, and the price follows it.
Ranges let you say "somewhere in here" and get paid for being right about the zone.
Figure 2. Ranges. Pick the zone, get paid for being right about it.
We weren't the only ones building on it
We also weren't the only ones running surfaces on these markets. Polysights and Trepa plugged the same markets into their own apps, and their users traded through a variation of interface. If the core and the surface are genuinely separate layers, other people can build the surface.
Precision for Trepa users meant a trading modality they're more familiar with.
Figure 3. Precision on Trepa. Name a call, choose your confidence, and the payoff sharpens as the band tightens.
Messi against Kane, who scores more fantasy points? Under the hood is a market on the difference between two distributions, which nothing built on a single over/under can express. The question the user answers is the simplest one in sport.
Figure 4. The duel ticket. Two players, one number, and the same engine pricing every shape on that axis.
Same underlying market. Completely different trading experiences with varying levels of cognitive tax.
The lesson: don't simplify the thing being traded, instead simplify the way people express what they think about it.
Where this goes in sport
Sports are a great battleground for this because the resolution is well defined and the events are already understood. There's very little ambiguity about what actually happened. But not all sports markets are created equal. If you ask how many goals a player is going to score, you quickly see that a single numerical market is pointless. You want to represent a whole distribution of possible outcomes, while the trader doesn't necessarily want to think about the distribution. They just want to express a view for a price they like. This is where the product layer does the work.
I am obsessed with the NFL so will use it as an example of where this could go sports wise.
You can already trade underlying player stats via props on sportsbooks and some prediction markets, though not many. There's nothing particularly novel about being able to trade Justin Jefferson receiving yards. The question is what happens when you give people completely different ways to interact with that same market.
Everyone thinks Justin Jefferson is going to get around 80 yards? Cool, I think he's going over 91. I'll take the 2.5x for that. Different views should have different payoffs, with the user mapping their own risk and payoff requirements whilst providing information.
And the reason this works particularly well in the NFL is that all of it is already part of the ecosystem. There are tens of millions of fantasy players. Commentators and podcasters talk about lines constantly. A huge amount of the statistical analysis produced around sports is already driven by betting markets and odds. Fantasy products, prediction columns, analysts, rankings, they're all already operating in this world. We're just adding another way to interact with something people already care about.
And the distribution that forms becomes useful information in its own right. Instead of just hearing that an expert thinks Jefferson is going to have a good game, you can see the market's actual range of outcomes. Fantasy products and analysts can take that data and put it in their models. The trading experience creates something useful beyond the trade itself. Don't just make it degen fun, try get a positive externality out of it.
The same logic past sport
Take crypto spot and RWAs. Already tradeable, endless instruments built around them. 100x leverage anyone? But nearly every instrument is a variation on the same view, direction plus leverage.
The moment your belief has any shape to it, "ETH finishes the year between 5k and 6k", or "I want to get paid more the further above 10k it goes", you're in options territory, and that surface has had retail flooding in over recent years, especially at 0DTE. A dynamic over/under or a range or a precision ticket on a price is the same belief with the cognitive tax removed. The demand for those views already exists in people's heads, it just has limited options of where to go.
Asset prices are on-chain, the mechanism is on-chain, so give people new ways to trade that don't require them to be "a trader" to contribute.
Once the primitive is generalised, the underlyings stop being limited to things that already have a market. Anything that resolves to a number can trade. Compute is a great example here, and our friends at Ito Markets are doing very cool work on this.
Here the thing being priced would be the GPU-hour. A distributional market on a benchmark GPU-hour rate lets a builder hedge their largest cost, lets a speculator express a view on the AI cycle at whatever shape they hold it, and the curve that forms feeds every desk pricing against it. Previously unpriced, becoming priced in real time. The same goes for anything people already argue about in numbers. Mortgage rates, inflation prints, freight, gas fees.
Complex core, simple surface. Build new things, make them easy to use, and don't dumb down what underpins them. The markets that already exist still have plenty of new ways in. Let's go find them.