How do we solve the liquidity problem with prediction markets?

Prediction markets don't suffer from a shortage of opinions. The long tail markets actually suffer from a shortage of liquidity that's deep enough to absorb order flow, tight enough to give fair prices, and fast enough to reflect new information. Remember, in prediction markets, the traders are the supply and the demand is measured by how valuable the information is that you glean from the odds. If the odds aren't accurate due to lack of liquidity, markets don't forecast but rather just serve as a pretty interface to show the stranded orders which leads to a poor user experience. Our view at Agara is simple: solve liquidity as a system problem, not a feature. That means unifying fragmented order flow, pairing the right market mechanic with the right venues, rewarding market makers, and putting the right markets in front of the right users at the right time.

In a simplistic version, there are four main characters in a prediction market. There's you, who wants to express a view right now. There's a counterparty who wants the other side, but not right now, maybe later, who knows. There's a market maker whose entire job is standing between you two and absorbing the timing mismatch and the inventory risk providing two sided liquidity. And there's the venue that sets the rules, runs the market and pays the market maker for doing this service.

Remove any one of these four and price discovery slows. Spreads widen. The product stops feeling like trading and starts feeling like waiting. Waiting on unfilled orders is just bad UX.

Why is liquidity scarce?

Binary contracts make the maker's job hard in a specific way. In a stock, a bad headline can hurt, but you usually find something above zero tomorrow. The company still exists. It still makes widgets or loses money or whatever. In a binary prediction market, one side goes to literally zero at resolution. "Will this happen by Tuesday?" Either it happens or it doesn't. One side gets a dollar. The other side gets nothing. You can hedge this delta most of the time. Market makers know how to do this. But the left tail is awkward. When the thing actually happens (or doesn't), the valuation snaps to zero and your hedge breaks at the worst possible moment, which is to say the moment when you most need it to work. This characteristic pushes pros to quote wide or sit out long-tail markets. As a result, "always-on" liquidity collapses outside the headline contracts unless you pay for it (called subsidies or liquidity rewards). Moreover, in prediction markets there is always a risk of adversarial or toxic flow where you might be trading against an insider which further disincentivizes the MMs to participate in certain markets prone to this.

Semantic fragmentation of markets often make this worse, and not in subtle ways. Venues split the same state of the world into separate markets because the words are different. "Will Candidate A win?" and "Will Candidate B lose?" are the same event. But they sit in different pools anyway, with different order books, and sometimes different prices because (and I cannot stress this enough) the platform treats one event as two events. Multi-outcome events get even worse. "Who will win the election?" gets sliced into a swarm of binaries: "Will Candidate A win?" "Will Candidate B win?" "Will Candidate C win?" These are mutually exclusive outcomes of one event. The platform treats them as separate, so prices drift apart and liquidity gets fragmented. On a bad day the implied probabilities don't even add to 100%.

Resolution issues further tax liquidity. Handwritten rules miss edge cases. If the resolution policy is vague, or if the dispute process takes days (or weeks!), capital gets stuck when users most want to recycle it. Market makers charge for that uncertainty. The charge shows up in the spread. Regular users notice differently: if cashouts lag, habits never form. The app becomes a place you read headlines instead of a place you trade on them.

Discovery is off too. Most venues shove U.S. politics, U.S. sports and a few global headlines to everyone. If you live in Ho Chi Minh City and care about a local vote, or you just want today's La Liga team update in Spanish with reliable sources, you will not see it. Liquidity follows attention. Attention follows relevance. The news feed is part of the exchange.

Finally, the instruments don't even match retail's expectations. Sportsbooks do 70% of their volumes on parlays because they compress multi step narratives into lottery-like payoffs. Prediction markets force you into one fully collateralized line with a capped payoff and no leverage. People want narrative bundles and the chance to one click express their view. The gap is both product and pricing: if you want retail dollars, you need payout shapes they already love, without blowing up risk.

You can fix all of the above. The solution is treating liquidity as the product instead of the thing that happens after you have a product.

Our POV (and what we're building)

One event, one pool, many faces

We will map different wordings that mean the same thing to a single canonical event ID and resolution criteria. Liquidity lives in one place. The Vietnamese card can say "Ông Trump có thắng không?", the English card can say "Will Biden lose?", and a Philippines card can have a market about whether the next president will be a Democrat or a Republican. Different faces. Same pool. One trade updates every alias. This is how multi-outcome markets were always supposed to work. So yes, we will aggregate Polymarket, Kalshi, and any venue that already hosts the canonical line. If the best market lives there, we route flow. If a market does not exist, we create it while checking for semantic similarity (more on that below), but we will not create five lookalikes that split order flow. The mandate is grow demand and deepen one pool, not spawn clones that compete with each other.

Localization is the wrapper that grows your addressable market. Aggregation is the back end that prevents you from fracturing your depth. Do both and liquidity compounds instead of getting shaved off by synonyms and translation artifacts.

Hybrid architecture: CLOBs and LS-LMSR

Most volume in serious markets still runs on order books because that's where pro market makers can quote tight and manage inventory within a familiar architecture. So we pair a CLOB for headline markets with an LS-LMSR pool for long-tail and multi-outcome markets that would otherwise die. Big elections, major crypto markets, core sports markets all will live on a CLOB. We will run a standard liquidity rewards program to incentivize two sided liquidity with strict controls around quote quality obligations and position limits. For the long-tail markets, we use a LS-LMSR which adapts its liquidity parameter as activity grows. Early trades move prices enough to reveal signal and depth scales automatically as the market earns attention (and thereby liquidity). With this AMM you can always provide liquidity for the long tail markets 24×7 and as the platform, it has the added benefit of the loss being bounded at a market level.

Retail friendly trading mechanics

Retail loves leverage. Short term binary markets already exhibit huge implied leverage as expiry approaches (around 470x on average). Today PMs avoid leverage due to extreme volatility. While this is true for the long tail markets, the popular ones are surprisingly stable and with mitigations like leverage decay (reduce leverage around resolution time and around the edge of the bands <10% and >95%), collateral insurance (a protective fee that's refunded on profitable closes, consumed only on losses or liquidations) and protections around fluctuations (a small reserve to ride out mean-reverting spikes) we can easily provide about 4–5x leverage for the liquid markets.

Parlays are another mechanic which are already quite popular in sports but we will be bringing it every market in Agara. People love them because they compress a narrative into a one click lottery ticket. You're not betting on whether the Lakers win. You're betting on whether the Lakers win and LeBron scores 30 and it goes to overtime. That's a story. Stories are fun. We will let users buy simple story bundles ("Trump wins and markets rally") but price every bundle off the same canonical distribution so the depth and liquidity stays in one place. While we are experimenting with our in house parlay pricing engine, we also plan to use a third party for pricing parlays similar to Kalshi.

We want to provide all these mechanics without making it seem like a Bloomberg terminal and in an extremely retail friendly, mobile first UI.

AI for creation, relevance, and resolution

Markets that pay out quickly and resolve cleanly attract both sides of the book. So we score potential markets by virality, non-manipulability, resolvability, topic liquidity and time to payout. AI agents watch trusted feeds in multiple languages, propose questions with clear contractable rules, exhaustive edge cases and citations, and push them to the right users based on history and location. Time to market creation is super important because news is ephemeral and liquidity follows attention. What is trending this week rarely trends the next week too. When it is time to settle, the same pipeline resolves first pass, escalates only when needed, and keeps an audit trail. LLMs are particularly good at closing loopholes in abstract rules; paired with a dispute backstop, they cut both ambiguity and time-to-cash. We are also working on an AI market creation pre-alpha prototype which you can find here.

About Agara

Agara is the local-first AI powered prediction venue built on stablecoin rails with on chain liquidity and retail friendly trading mechanics and UX.

Agara is the place where knowledge turns into capital.

All Posts