Diving deeper into why market making in binary outcome markets breaks the rules of liquidity itself
The Metaphor: Playing Poker Against God
In every market, there is a balance between knowledge and risk. Market makers step into that tension, quoting prices between buyers and sellers, earning small profits for providing liquidity while absorbing losses when the other side knows more. The business works because, over time, information tends to even out.
Prediction markets are the exception.
Imagine playing poker against someone who can see your cards. You might win a few small hands, but in every high-stakes round your opponent knows when to go all in. Eventually your stack is gone. That is what it feels like to make markets in binary prediction markets, where the outcome is a simple yes or no and information is asymmetric by design.
In most markets, probability and uncertainty protect liquidity providers. In prediction markets, those protections collapse. Once the outcome is known, half of your inventory will go to zero with mathematical certainty.
The Case Study: Eric Adams and the $250,000 Certainty
Consider the real-world case that traders now call the Adams gap.
A market opened on whether New York City Mayor Eric Adams would drop out of the race before the end of the month. It traded around 25 percent. To a casual participant, that seemed plausible: political rumors, investigations, and pressure all suggested a real chance.
But anyone inside Adams’s campaign, or close to City Hall, knew he had no intention of dropping out. If one of them took a one-million-dollar position on the “no dropout” side, the market maker who quoted the 25 percent price would lose $250,000 as the contract settled at 100 percent.
There is no hedge for that. You cannot short “Adams drops out” on Kalshi while buying “Adams stays in” on Polymarket and expect a clean offset. The correlation is broken by timing, latency, and venue structure. Once perfect information enters, your exposure is absolute.
That single trade reveals the paradox: the more a prediction market reflects real-world events, the more it invites informed traders who have an advantage no model can neutralize. Market makers end up subsidizing truth itself.
The Core Problem: One Side Always Goes to Zero
Traditional market makers thrive because inventory retains some value. In equities, if you are long Apple and short Microsoft, both may move against you, but neither goes to zero. You can hedge, rebalance, and survive.
Binary markets do not work that way. Every position eventually resolves into a 0 or a 1. The losing side becomes worthless, and the winning side captures all value. A liquidity provider who quotes both sides is guaranteed to hold inventory that will become valueless.
If the market were perfectly random, you could balance those outcomes over time. But prediction markets are not random. They attract traders who know more than you: insiders, campaign staff, or people with early access to private polling. In that environment, every quote is a risk that your counterparty sees your cards.
This is not a matter of bad luck or poor risk management. It is structural. The math itself ensures that, without vast uninformed flow to offset informed trades, liquidity provision becomes a negative expected-value business.
In prediction markets, liquidity is not rewarded. It is punished. The more truth a market reflects, the worse it becomes for those who make it possible.
Why Traditional Market-Making Models Fail
Most automated market-making systems were designed for continuous price curves, not binary resolution. When a token or asset trades on both sides of a range, an AMM can offset losses on one side with gains on the other. In prediction markets, one side always disappears.
AMMs like Polymarket’s early constant-product pools face a built-in flaw: if half of the inventory will go to zero, the pool’s liquidity dries up as traders move toward the winning outcome. The protocol cannot replenish itself fast enough.
Subsidies can delay the problem but not solve it. A $10 million liquidity incentive can bootstrap trading for a short period, yet as soon as insiders exploit the information gap, the losses exceed the subsidy. The model works for experimental markets, not for trillion-dollar ones.
Frequent batch auctions, or FBAs, can mitigate latency risk by matching orders in short intervals instead of continuously. This reduces sniping but not the underlying asymmetry. If one participant has perfect information, no auction cadence can protect the uninformed side.
The result is a liquidity death spiral. Market makers lose, spreads widen, retail participants leave, and even informed traders eventually lose counterparties to take the other side of their bets.
The Death Spiral
The majority of prediction markets have some form of gap risk: the sudden repricing of an event when new information becomes public. In crypto markets, makers can step away during known risk windows, such as major economic announcements. In prediction markets, there is no safe window. The information edge can appear at any moment.
If you are quoting a market on whether a candidate will resign, you cannot predict when the news will break. When it does, you are the first to pay for it. Your quotes fill instantly, and your inventory becomes worthless.
This dynamic discourages liquidity provision altogether. Without liquidity, spreads widen, traders receive poor fills, and overall activity declines. That loss of depth makes the next information shock even more severe, further driving away market makers.
This is why most prediction markets remain thin and volatile despite interest from professional traders. It is not a question of infrastructure or user interface. It is a question of math.
Why It Matters
Prediction markets promise a pure reflection of collective intelligence. They are supposed to aggregate dispersed information into accurate forecasts. For that vision to work, markets must be liquid enough for people to enter and exit easily.
When liquidity itself becomes unprofitable, the system fails. The very participants who should stabilize prices are the first to disappear. Without them, prices become jumpy and unreliable, destroying the signal that prediction markets are meant to produce.
The irony is that as prediction markets become more accurate, they become less sustainable. Perfect information kills liquidity.
Emerging Solutions
A few projects are experimenting with ways to soften the asymmetry. Some, like Gondor Lending and Multiverse Finance, are exploring collateralized lending and synthetic exposure that allow liquidity providers to diversify risk across multiple markets. Others are testing adaptive fee curves that rise dynamically when volatility spikes, compensating makers for potential loss.
These approaches miss the core problem. Diversification assumes independent risks, but prediction market information clusters naturally. Political insiders have edge across multiple related markets. Corporate executives know about earnings, M&A, and leadership simultaneously. When one market gaps on new information, correlated markets gap together. A diversified portfolio doesn't reduce exposure to informed traders—it expands the attack surface.
The underlying issue remains: binary outcomes create finality that no existing model can hedge. When an event resolves, one side goes to zero with mathematical certainty. No fee structure, liquidity incentive, or portfolio strategy changes that.
What may be needed is not better execution of current models, but entirely new primitives. Markets that fragment risk differently, resolve more granularly, or restructure how information flows through the system. The infrastructure to support such innovation doesn't exist yet—but that's the direction the industry must move if prediction markets are to scale beyond niche volumes.
The Path Forward
Prediction markets sit at the edge of what finance and computation can model. They turn human uncertainty into tradable data, and in doing so expose the limits of every existing liquidity mechanism.
If we want markets that can reflect truth in real time, we need systems that move at the same speed as information. That means new primitives, new infrastructure, and architectures that treat liquidity as a living process rather than a static pool. Prediction markets are not broken. They are simply ahead of their time.
Linera is building the foundation for that future: a blockchain architecture designed for real-time applications where new market structures can be tested at internet speed.
About Linera
Linera is the first blockchain optimized for real-time applications and interoperability at internet scale. Founded by a former Meta/Novi research engineer, Linera introduces microchains—lightweight, parallel chains that enable direct user interaction. Backed by top-tier investors, Linera is live on testnet and preparing for mainnet launch.
