Prediction markets are often described as a clean use case for onchain finance.

Clear outcomes. Binary payoffs. Transparent settlement.

And yet, most prediction markets today still rely on liquidity and market-making designs that were already showing cracks a decade ago.

This is not a problem of adoption but rather a problem of structure.

Billions of dollars in notional exposure pass through prediction markets each year. And nearly all of it is intermediated through mechanisms that assume:

• static liquidity,
• passive participants,
• and benign market conditions.

Those assumptions no longer hold.

Being outdated is not the core issue. The deeper problem is that these designs actively degrade price quality, leak value, and discourage serious liquidity providers.

Let’s break down why.

The Core Mechanical Issue: Prediction Markets Are Volatile by Design

At their core, prediction markets compress uncertainty into narrow price bands.

A contract trading at 0.52 implies a 52% probability. A move to 0.60 is not “small.” It is a material repricing of belief.

Now combine that with:

• binary settlement
• fixed expiry
• and asymmetric information flow.
  
  

Instead of prices adjusting gradually, markets reprice in sharp bursts as volatility arrives in concentrated intervals.

Liquidity is needed most precisely at moments of uncertainty, not during calm consensus.

• When new information arrives in prediction markets, prices tend to reorient abruptly rather than adjust incrementally, because information itself usually arrives in discrete moments such as data releases, legal decisions, public statements, or vote count updates. Participants who interpret that information faster or with better context act immediately, and if sufficient liquidity is not present at that moment, price moves rapidly across levels as standing orders are consumed, undermining the market’s ability to absorb new information in a controlled manner.
  
  
• When sentiment shifts quickly, the market is no longer reacting solely to underlying facts but to expectations about how others will respond to those facts, which creates one-sided flow. In the absence of adequate liquidity, relatively small imbalances begin to compound, spreads widen, and prices overshoot what a more stable environment would support, turning what should be a measured adjustment into a discontinuous move that is difficult to trade around and easy to regret.
  
  
• As resolution approaches, the nature of risk becomes increasingly asymmetric, with less time available to correct mispricing and no future volatility to dilute errors. Inventory management becomes harder, as market makers can no longer rely on time diversification or gradual rebalancing, and each trade begins to resemble a binary exposure rather than managed risk. This is precisely when participants most need reliable liquidity to enter or exit positions responsibly, and precisely when many liquidity providers are forced to reduce size or step back altogether.

If liquidity thins during these phases, the market does not merely become volatile. It becomes less informative. Prices stop reflecting aggregated belief and start reflecting whoever was willing to cross the spread first.

A prediction market that cannot sustain liquidity during information shocks and near resolution is not failing because of lack of interest. It is failing because its market-making assumptions are misaligned with how uncertainty actually resolves.

Unfortunately, this is exactly when most existing market-making models perform the worst.

Problem #1: Passive Liquidity Breaks Under Information Shock

Most prediction markets rely on passive liquidity provisioning models inspired by early AMMs or static order books.

These models are built on assumptions that work in slower, more continuous markets, where liquidity can sit idle until it is traded, prices adjust gradually as beliefs shift, and adverse selection can be managed through spread and size alone. In prediction markets, those assumptions break down almost immediately.

Information tends to arrive in discrete, credible bursts that resolve uncertainty rather than refine it, and when that happens price no longer progresses gradually through the order book but instead repositions abruptly as standing liquidity is consumed. Orders that appeared safe moments earlier are taken in quick succession, and whatever liquidity remains either widens aggressively or withdraws entirely.

At that point, market makers are no longer earning compensation for providing liquidity but are instead absorbing information risk without a corresponding return, as the mechanics reward participants who act first and penalize those who were simply present in the book.

The outcome is consistent across venues: depth thins, repricing occurs in sharp steps rather than controlled adjustments, and participants who arrive slightly later face poor execution not because their view is incorrect, but because liquidity retreated before the market could stabilize.

This is not a problem of participation or interest, but a structural mismatch between how prediction markets resolve information and how their liquidity models are designed to operate.

Problem #2: Market Makers Are Trading Against Information, Not Flow

In most markets, market makers manage inventory risk created by client flow.

In prediction markets, they are often trading against information asymmetry.

When a participant trades because they know something material, the market maker is not earning spread. They are subsidizing someone else’s certainty.

Without mechanisms that explicitly price the arrival of new information, limit adverse selection, or dynamically adjust liquidity exposure as conditions change, rational liquidity providers are left with only two viable responses. They either widen spreads aggressively to protect themselves from being picked off, or they step back from the market altogether. In both cases, the result is thinner books and worse execution at exactly the moments liquidity is most needed.

This is why many prediction markets show activity spikes without depth, followed by long periods of stagnation.

Problem #3: Fixed Liquidity Assumptions Ignore Time-to-Resolution Risk

Prediction markets are not perpetual instruments.

Risk increases as resolution approaches.

Yet many prediction market designs treat liquidity as if it were time-neutral, applying the same assumptions whether a market resolves in six months or in six hours. That framing is structurally incorrect.

As time-to-resolution compresses, the nature of risk changes. Information advantage increases because any new data has less time to be absorbed, challenged, or arbitraged away. Volatility rises because repricing becomes more discrete, with fewer opportunities for incremental adjustment. At the same time, inventory risk becomes increasingly difficult to manage, as there is no longer sufficient time to rebalance positions or offset exposure through future flow.

For liquidity providers, this creates a hard constraint. What may be manageable inventory earlier in a market’s life gradually turns into binary exposure as resolution approaches. Without liquidity models that explicitly account for this transition, market makers are forced to reduce size or step back entirely, not because demand disappears, but because the risk profile no longer supports continuous quoting.

This is why time-aware liquidity is not an optimization in prediction markets. It is a requirement dictated by how uncertainty resolves.

Without time-aware liquidity logic, market makers are forced to manually step back at exactly the moment users need them most.

The Consequence: Value Leakage and Fragile Markets

These structural issues create second-order effects:

• Sophisticated participants wait for others to move first.
• Less informed users face worse execution.
• Liquidity becomes episodic rather than continuous
• Market credibility suffers.
  
  

Over time, this discourages institutional participation and reinforces the perception that prediction markets are niche or unreliable.

The problem is not demand but rather the market design itself. 

What Needs to Change: Smarter Market Making

The solution is not more incentives bolted onto the same models.

It requires rethinking market making for prediction markets as a dynamic risk management problem, not a passive liquidity problem.

Liquidity models that respond to the arrival of new information rather than remaining static, pricing behavior that adjusts as resolution approaches instead of treating time as irrelevant, and mechanisms that explicitly compensate market makers for absorbing informational risk rather than penalizing them for providing liquidity when uncertainty is highest.

Market making in prediction markets cannot look like market making in spot assets.

The payoff structure is different.
The risk surface is different.
The behavior of participants is different.

Design needs to reflect that reality.

Where This Is Headed

Prediction markets have proven they can attract attention, volume, and relevance.

Prediction markets have demonstrated that they can attract attention, participation, and relevance around real-world outcomes. What they have not yet demonstrated is that they can consistently produce stable, reliable pricing behavior when participation scales and information arrives unevenly.

That limitation will not be addressed through better marketing, more users, or louder narratives about adoption. Increasing activity without changing structure tends to amplify existing weaknesses rather than resolve them.

The path forward is mechanical, not promotional.

Markets need liquidity models that adjust to changing conditions rather than assuming static behavior. They need incentive structures that compensate liquidity providers for absorbing information risk, not just for being present during quiet periods. And they need market-making designs that are explicitly built for environments where volatility is driven by sudden information arrival rather than continuous flow.

Until those elements are addressed together, prediction markets will continue to function during calm periods and struggle precisely when uncertainty matters most.

Next time, we can go deeper into what effective market making in prediction markets actually looks like, and which design choices separate resilient markets from fragile ones.

Because in markets like these, structure decides everything.

‍