The silence at the center of the order book
You pull up a wholesale prediction market on an event that genuinely matters, something with real money behind it, and you notice it. The bid-ask spread on the headline contract is razor-thin, three or four cents on a dollar-denominated binary. Professional liquidity providers are clearly present, posting size, refreshing quotes. Scroll down to the related contracts, the ones that would let you express a more precise view, and the order book is empty. Not thin. Empty. No market maker is touching those contracts at any price they'd be willing to defend.
That silence is structural, not accidental. It follows directly from how wholesale prediction markets are built, what their liquidity providers need to survive, and which categories of risk those providers have quietly decided are unquotable on any terms they can accept.
What a wholesale prediction market actually is
A wholesale prediction market sits above the retail layer. Where a consumer-facing platform might let a hobbyist put twenty dollars on an election, a wholesale venue is designed for institutional participants: proprietary trading firms, hedge funds, and professional market makers who think in portfolios rather than positions. Contracts are larger, margin requirements are higher, and the counterparties on the other side of your trade are assumed to be informed.
The liquidity provider in this environment is not a bookmaker taking directional risk. The classic LP model is closer to a specialist on a stock exchange: post both a bid and an offer, collect the spread, hedge the resulting inventory through correlated instruments or offsetting positions in related contracts, and repeat. Profit comes from the flow, not from having the right view on the underlying event. This matters enormously, because it means the LP's entire business model depends on two things: the ability to hedge, and a reliable estimate of how much adverse selection they'll face from better-informed counterparties.
Strip away either of those, and the LP goes home.
The machinery of hedging, and where it breaks
Take a concrete example. Imagine a wholesale market offering binary contracts on whether a major central bank's next policy decision will be a rate increase, a hold, or a cut. An LP posting quotes on those three contracts can delta-hedge against interest rate futures, options on government bonds, and currency forwards. The underlying is liquid, the correlation is tight, and the LP's inventory risk on a given contract is offset by instruments that trade continuously, in size, on exchanges around the world.
Now imagine the same market offers a contract on whether a specific regulatory agency will approve a particular merger within ninety days. Completely different problem. There is no liquid financial instrument whose price tracks the agency's internal deliberations. You cannot short a futures contract on bureaucratic process. The LP can look at base rates for merger approvals in similar sectors, but base rates are not hedges. If a large informed trader comes in and buys a thousand contracts at the LP's offer, the LP is now short a position they cannot lay off anywhere. The only way to reduce that exposure is to move their quote aggressively away from the market, which is another way of saying they stop quoting.
This is the first structural blind spot: events whose resolution is driven by a small number of discrete, unobservable decisions made by individuals inside closed institutions. Regulatory outcomes, judicial rulings, internal corporate votes. No hedge instrument exists for these because no tradeable proxy for the information exists.
The adverse selection trap
The second mechanism is subtler and, in some ways, more damaging to the LP's willingness to quote.
In any prediction market, some participants know more than the LP. That's the point of the market. The LP accepts some adverse selection as a cost of doing business, and the spread compensates for it. Standard market microstructure theory, going back to work by Lawrence Glosten and Paul Milgrom in the 1980s, models this explicitly: the spread is partly compensation for the risk of trading against someone who already knows the answer.
The LP can tolerate this when adverse selection is distributed and bounded. Roughly one in ten traders with a modest informational edge is a manageable tax on the book. What the LP cannot tolerate is a contract where adverse selection is concentrated and catastrophic, where the population of likely counterparties is dominated by the small group of people who actually know the outcome. Think of it less like fishing in a stocked lake and more like fishing in a tank where the trout can read.
Consider a contract on whether a private company will announce a specific acquisition target within sixty days. The informed traders on that contract are, almost by definition, the bankers, lawyers, and executives who are party to the deal. Everyone else is guessing. An LP posting a quote is essentially offering a free option to insiders. The moment the LP's quote is live, it is a target. The spread required to compensate for that adverse selection risk would be so wide that no trader without inside information would ever transact, which means the only flow the LP attracts is the worst possible flow. The contract is structurally unquotable, full stop.
Two traders, call them Sophia and Marcus, both make markets on wholesale prediction venues at competing prop firms. They will quote the same central bank contract all day, grinding out fractions of a cent on thousands of contracts, a business that likely clears low seven figures annually across the book before costs. Ask either of them to quote the corporate acquisition contract, and they say the same thing: the spread they'd need to post would function less as a market and more as a warning sign.
When the resolution mechanism itself is the problem
A third category of unquotable risk concerns events where the resolution criteria are ambiguous, contested, or subject to retroactive interpretation.
Wholesale LPs are not just pricing the probability of an outcome. They are pricing the probability of that outcome as defined by the contract's resolution rules, adjudicated by whoever controls the oracle or settlement process. If those two things might diverge, the LP faces a risk that has nothing to do with the underlying event at all.
Imagine a contract that resolves YES if a country's GDP growth exceeds two percent in a given fiscal year. Straightforward enough, until you ask: which statistical release? The preliminary estimate, the first revision, or the final revision published eighteen months later? What happens if the national statistics office changes its methodology mid-year? What if there's a dispute between the market's resolution committee and the published data? The LP now has to price not just economic outcomes but governance risk inside the market itself. That is a second-order uncertainty layered on top of the first, and most LPs decline the contract rather than try to model it.
This is why the most liquid wholesale prediction markets specify resolution criteria with a precision that looks almost neurotic. The pedantry is the product. Every ambiguity in the contract spec is a risk the LP will charge for, or walk away from.
The contracts that survive, and why
Look at what actually trades in wholesale prediction markets with genuine depth and the pattern is consistent. Liquid contracts share three properties.
First, the resolution event is exogenous and publicly observable: an election result, a central bank rate decision, a recorded sports outcome. No small group controls the information. Second, correlated hedging instruments exist in adjacent financial markets. Third, the resolution criteria are mechanical and unambiguous, with no interpretation required.
When all three conditions hold, LPs can post tight spreads, manage inventory efficiently, and absorb significant informed flow without existential risk to their book. When any one of the three is missing, the spread widens dramatically. When two or three are missing, the contract sits unquoted.
Found a contract you care about with a two-cent spread and real size on both sides? You are looking at something that satisfies all three. The market is doing its job. The spread on a contract where only one condition holds might be fifteen or twenty cents on a binary. That is not inefficiency. That is the LP honestly pricing the risk they cannot hedge away.
The deeper consequence
Here is the uncomfortable part, and it is worth sitting with. Wholesale prediction markets are most liquid, and therefore most informationally efficient, on precisely the questions where other information sources are also plentiful: major elections, scheduled economic releases, high-profile sporting events. The questions where a prediction market's aggregation function would be most valuable, where the information is genuinely dispersed and hard to synthesize, are almost always the questions where LPs refuse to quote.
This is not a flaw that better technology will fix. It is a direct consequence of the economic logic of liquidity provision. An LP who quotes unhedgeable, adverse-selection-dominated contracts will lose money to informed traders until they stop. The market structure does not fail to aggregate difficult information because of a design oversight. It fails because no rational intermediary will stand in the middle of a trade where one side knows the answer and the other does not, not for any spread a genuine buyer would accept.
The silence in the order book is not ignorance. It is the market telling you, with considerable precision, exactly where the limits of what markets can know begin.