Market making in prediction markets: a practitioner's guide

What a prediction market maker actually does

A market maker in a prediction market does the same thing a market maker does anywhere: it stands ready to buy and to sell, posting a bid below its estimate of fair value and an offer above it. Someone who wants to express a view can trade immediately instead of waiting for a natural counterparty to arrive.

What is different is the instrument. An event contract pays a fixed amount (usually $1, or 100¢) if a defined outcome occurs and nothing otherwise. Its fair price is, to a first approximation, the probability of the outcome. That gives event markets three properties that shape everything about how they are made:

• Bounded prices. A contract cannot trade below 0 or above 100. Volatility is not constant; it is mechanically linked to where the price sits.
• Terminal certainty. Every contract resolves to exactly 0 or 100. Inventory you hold at resolution is not marked to a model; it is marked to the world.
• Jumps. A large share of the information arrives in discrete moments, such as a data release, a court ruling or a final whistle, rather than as a continuous drift.

A good market maker is paid, through the spread, to absorb the risk those properties create. A bad one is paid nothing and absorbs it anyway.

Step one: a fair value you are prepared to defend

Every quote starts from a number. For some contracts that number comes from a structural model: a nowcast of next month's inflation print, a weather ensemble, a poll aggregate. For others it comes largely from the market itself, adjusted for what the market cannot see. In practice it is a blend, and the weights are learned from how each source has performed on resolved contracts.

The quality of this number determines whether market making is a business or a donation. If your fair value is systematically worse than the average informed participant's, every trade against you is slightly more likely to be a trade you should not have done. This is adverse selection, and in event markets it is concentrated: the participants most likely to trade with you right after a release are the ones who read it fastest.

Step two: setting the spread

The spread compensates the market maker for three costs: adverse selection, inventory risk and fees. All three vary with price in event markets.

A useful starting point is the standard deviation of the contract's terminal payoff. For a contract trading at probability p, the payoff is Bernoulli with variance p(1−p). That quantity is largest at 50¢ and shrinks toward zero at the tails, which is why spreads in liquid event markets tend to be widest around the middle and tightest near certainty.

Here a is a floor (tick size, operational cost), σ_info(t) scales with how much information is expected to arrive before you can re-quote, fee(p) is the venue's fee at that price, and skew(q) shifts both sides according to current inventory q.

The information term is what separates event-market making from making a market in a stock. Ten minutes before a scheduled inflation release, the expected information arrival over the next quote lifetime is enormous; ten minutes after, it collapses. A professional market maker's spread breathes with the calendar.

Step three: inventory and skew

When a market maker's bid is hit, it becomes long the outcome. If it does nothing, a run of one-sided flow can leave it with a large position on a single binary result, which is exactly the risk the business is meant to avoid. The standard response is skew: move both the bid and the offer against your position so that the next trade is more likely to reduce it.

A simple, robust skew is linear in inventory and scaled by the contract's payoff variance:

Event markets add a twist: inventory is rarely independent across contracts. A position in "Fed cuts in December" is correlated with positions in "two-year yield below X", "CPI below Y" and any contract on the number of cuts this year. Mature desks therefore manage inventory at the level of themes, aggregating exposure across contracts that share a driver, and skew each contract by its contribution to theme risk rather than by its own position alone.

Where a cheaper hedge exists in a correlated contract, it is often better to buy the hedge than to skew aggressively. Over-skewing protects inventory at the cost of quote quality, and venues notice.

Step four: surviving jump risk

Most losses in event-market making come from a small number of moments. A scheduled release prints far from consensus, or a surprise announcement lands, and a stale quote is lifted before it can be pulled. Three tools matter:

1. Calendars. Every scheduled information event is known in advance. Quotes on affected contracts widen and shrink in size on a pre-set schedule into the release, then re-tighten as soon as the new fair value is computed.
2. Speed on primary sources. For scheduled releases, the winner parses the primary source (the statistical agency, the central bank, the official results feed) directly. Every millisecond of delay is a window in which someone faster can trade against your old price.
3. Size discipline. Displayed size near a jump should be small enough that the worst plausible outcome is a bad day, not a bad year.

Unscheduled news is different. There, being fast matters less than being correctly sceptical: early reports are often wrong, and the right reaction to an unverified headline is usually a wider spread, not a new mid.

Fees, rebates and incentive programmes

Venue fee design materially changes how markets are made. Some regulated event exchanges have published fee schedules that scale with p(1−p), so that a trade near 50¢ costs more than one near the tails, with separate treatment for resting (maker) orders. Several on-chain venues have charged no trading fees on most markets, sometimes paired with liquidity rewards for resting near the midpoint.

For a market maker, the relevant quantity is the all-in cost of a round trip, including fees on both sides, any rebates and the cost of capital locked as collateral until resolution. That last item is easy to forget. A contract that resolves in eleven months ties up collateral for eleven months, and a 1¢ edge on a 50¢ contract is a very different annualised return depending on how long you wait to be paid.

Fee schedules change frequently. Always work from the venue's current rulebook rather than a summary.

Where the edge comes from

Market making in prediction markets is not one trade with a large edge. It is millions of small trades, each with a slightly positive expectation, whose errors are largely uncorrelated. The edge comes from:

• Earning the spread from participants who value immediacy more than a cent or two of price.
• Better fair values than the average counterparty, especially in categories where the market's information is slow or noisy.
• Better risk management, so that the inevitable bad outcomes are survivable and the book stays live when others pull.
• Breadth, because a desk that quotes ten thousand contracts sees far more of the order flow, and learns faster, than one that quotes a hundred.

Liquidity is not something you provide when it is easy. It is something you are paid to provide when it is hard.

That is also why venues value professional liquidity. A market with a two-cent spread at 3 a.m. on a Sunday is a better market than one with a ten-cent spread, and the price it shows is a better forecast.

Frequently asked questions

Related: Pricing event contracts · Cross-venue arbitrage · Venue programme