The Market Structure of Prediction & Event Markets, 2024–2029.
The rise of information finance — and its first liquid instrument.
A two-year reckoning, and a three-year wager.
In twenty-four months a fringe curiosity priced its way into the financial plumbing. Who built the rails, who is fighting over them, and where the whole thing is pointed next.
It never is. The sportsbooks spent a decade and a fortune assembling state-by-state licenses, convinced the patchwork itself was the moat. The exchanges assumed event contracts were a novelty act. The regulators assumed they could simply decline to engage. Each was proven wrong on roughly the same timetable, and for the same reason.
A prediction market is a deceptively boring thing: a contract that pays a dollar if something happens and nothing if it doesn't. Price it between zero and a hundred cents and you have read a probability straight off an order book. That mechanism is older than the people now getting rich from it. What changed in the last two years was not the instrument. It was who was allowed to list it, who decided to clear it, and which regulator was willing to plant a flag.
This paper does two things. It walks the last two years — the regulatory inversion, the capital flood, the venue scramble, the arrival of the incumbents — participant by participant, because market structure is a story about who occupies which seat. Then it carries the picture three years forward, with the usual caveat that anyone forecasting this industry in 2024 was wrong, and probably modestly so.
Between early 2024 and mid-2026, U.S. prediction markets went from a single contested venue trading election contracts to a multi-venue, vertically integrating sector that cleared roughly $44 billion of global volume in 2025 and $29.8 billion in April 2026 alone — with the New York Stock Exchange's parent, the country's largest derivatives exchange, both major sportsbooks, and a Coatue-led $22 billion Kalshi round all inside the stack. The decisive variable was never product. It was jurisdiction: the moment the CFTC chose to treat event contracts as its exclusive province, the federal derivatives wrapper became a way to route around fifty separate state gambling regimes with a single license. Institutions remain on the data more than in the book — estimated under 5% of open interest — but the sell side has begun to stir. The fight that remains — federal preemption versus state police power, now across 19 federal lawsuits with a circuit split forming — is the single largest determinant of the structure that exists in 2029.
Sources: RiskSmart Intelligence global-regulation tracking (Dune Analytics / DeFi Rate); WSJ / CoinDesk / PYMNTS on funding; Congressional Research Service; Holland & Knight, Norton Rose Fulbright; CNBC; Reuters. Figures are point-in-time and move weekly. (An alternative Kalshi+Polymarket-only compilation puts 2025 throughput nearer $63.5B; the $44B figure is the broader sector-tracked number used across this series.)
The transformation has a clean arc: a regulator stops fighting, capital notices, the product finds its real demand in sports, the venue licenses become trophies, and the incumbents arrive last — as they always do.
In the first half of 2024, the sector was essentially one disputed exchange and a crypto platform that claimed to block U.S. users. Kalshi was litigating the CFTC for the right to list election contracts at all — a fight it won in the D.C. Circuit, the foundational federal precedent that election contracts are not “gaming.” The Commission, under its prior leadership, went the other way: in June 2024 it proposed a rule branding political and sports event contracts “contrary to the public interest.” The rule was never finalized. That non-finalization became one of the most consequential events in the sector's history.
The one early sign of where this was heading came in April 2024, when Susquehanna International Group — one of the largest options market makers in the world, an entire firm built around pricing probabilities — became the first major financial institution to publicly make markets on Kalshi. The mechanics were familiar. The spreads were absurdly wide. There was nobody else there.
The 2024 election cycle proved the demand; the change in administration changed the posture. On January 29, 2026, CFTC Chairman Michael Selig withdrew the Biden-era proposed ban, vacated a 2025 staff advisory, and directed new “clear standards” rulemaking — calling the prior posture “a frolic into merit regulation.” By February the Commission had replaced posture with doctrine: event contracts are swaps, swaps are federal, and the states are intruding.
If a sports event contract is a “swap” under the Commodity Exchange Act, it is federally regulated and — the CFTC argues — shielded from state gambling law by preemption. That single classification lets a platform offer what looks, smells, and pays out like a sports bet in states that never legalized sports betting, under one federal license instead of forty separate state ones. The entire 2025–26 scramble is downstream of that sentence.
Once the legal ambiguity began to clear, the money arrived at a pace that embarrassed every prior estimate.
The shift from retail experimentation toward institutional relevance is now a topic on the volatility podcasts the buyside actually listens to. On Dean Curnutt's Alpha Exchange (May 11, 2026), the conversation turned to prediction markets tied to elections, Fed policy, GDP, inflation, and geopolitical outcomes — framed explicitly as moving from retail experimentation toward institutional relevance, with Kalshi as the reference venue and the discussion's sentiment on the category running net positive.
That is the tell: prediction markets have migrated from the gambling press into cross-asset volatility conversations among risk professionals. The retail plumbing already exists; the institutional analytical layer — mapping event probabilities into pricing, risk, and reset workflows — is the part that still doesn't exist at scale. That gap is the opportunity.
By spring 2026 the sector had four things it lacked two years earlier: a federal regulator actively defending it in court, an incumbent exchange complex (ICE, CME, now Cboe and Nasdaq) inside the tent, a retail distribution layer reaching tens of millions of accounts, and a structural enemy — the states, the sports leagues, the tribes, federal prosecutors, and a growing bipartisan bloc in Congress — large enough to put the whole thing at risk. That tension is the market structure. Everything below is who sits where inside it.
Before the history, the question every institutional reader is entitled to ask: why should an exchange, a bank, an asset manager, or a risk desk care about a market that still trades mostly Super Bowl contracts?
Because what is actually being built here is a new financial function, and it has a name now — information finance: the business of turning information into a priced, tradeable, hedgeable instrument. Susquehanna and DRW already staff desks under that banner; what the category lacks is not participants but a definition. Prediction markets are simply its first liquid instrument. Seen that way, the contract is quietly becoming four things at once, only one of which is entertainment:
That is the lens for everything that follows. The history matters because it explains how a fringe instrument acquired federal infrastructure; the infrastructure is why the instrument now lands on the desks of people who have never placed a bet in their lives.
The fair challenge: we already have futures, options, CDS, fed funds futures, election ETFs, and structured products. What does a binary event contract measure that those don't?
The distinction is the whole pitch for information finance: options give you volatility; information finance gives you the probability itself. Explicit probability, cleanly, on events the existing toolkit prices only indirectly or not at all. An option encodes a probability distribution over a continuous price, entangled with volatility, time, and rates — you can back out an implied probability, but you are inferring it, not reading it. A prediction contract is the probability: a price of 38 cents is a 38% chance, full stop. And it extends to exogenous, non-financial events — a specific election outcome, a rate decision on a named date, a geopolitical trigger — that no clean listed instrument covers.
| Instrument | What it actually measures |
|---|---|
| Polls / surveys | Stated opinion, no money at risk, slow to update |
| Futures | Directional economic exposure to a continuous price |
| Options | Volatility and a distribution — probability only by inference |
| CDS | Default probability, but only for credit events |
| Prediction / event contracts | Explicit probability of a discrete, named outcome — including non-financial ones |
That is the intellectual foundation for institutional interest. Not that event contracts replace the derivatives complex — they fill a specific hole in it: a clean, tradeable, centrally-cleared probability on the discrete events that move portfolios but resist the continuous-payoff instruments built for prices.
Before the history, the endpoint. An information finance desk is a machine for transformation: information, client demand, and capital flow in; the desk prices, risks, makes markets, and settles; liquidity, structured products, signal, and captured spread flow out. Each particle is a unit of flow, colored by what it is. Hover a node for its role; filter by stream to trace one path. This is the institutional shape everything else in the paper is moving toward.
Illustrative architecture; particles are a visual abstraction of flow, not measured volume. The five outflows map one-to-one to the dealer-economics fee pools: market-making spread, structured products, data/signal, financing & cross-margin, and captured P&L.
Everything above is described in prose; here it is as flow. Read left to right — who is trading, how they reach the book, who is on the other side. Today it is almost one color: retail, through direct accounts and brokerage apps, matched by market makers. Press play, and watch the ribbons that barely exist today — FCM connectivity, RFQ and block, dealer liquidity — widen year by year. Toggle Notional $ to see the second truth: even as retail's share falls, the book grows, so retail flow keeps rising while the pie expands around it.
Flow shares are illustrative and directional, calibrated to the structure in this brief rather than to reported volumes. The notional path ($44B in 2025, anchored to the ~$29.8B-per-month April 2026 run-rate and scaled forward) is an illustrative trajectory, not a forecast.
Market structure is seating. A prediction-market trade now passes through a layered stack — a venue to list it, a clearinghouse to guarantee it, a broker to access it, a market maker to price it, demand to take it, and vendors to wire it together. The defining move of 2025–26 was firms racing to own more than one layer at once.
The vertical stack a single event-contract trade passes through, with representative occupants. The 2025–26 contest is firms racing to own more than one adjacent layer — ICE buying into clearing-and-venue, CME renting venue-and-clearing to sportsbooks, Robinhood building its own venue (Rothera) rather than renting Kalshi's.
The clearest gauge of the rush isn't the valuations — it's the queue at the regulator. Since early 2025 the CFTC has taken in 17 applications for new designated contract markets; roughly seven were approved and about ten are still pending as of March 2026, with review now running up to six months. And the line keeps lengthening — Smarkets (the UK betting exchange), Bullish (the NYSE-listed digital-asset platform, filing for a clearinghouse alongside), and DimeTrades all filed fresh DCM applications in the weeks after that count. A federal exchange license — once a rarity held by a handful of incumbents — is now something dozens of firms are queuing for, because it is the single key that unlocks fifty states at once.
Cards synthesize RiskSmart Intelligence tracking and public reporting (Reuters, CNBC, WSJ via CoinDesk/PYMNTS, Finance Magnates, DeFi Rate, CRS) as of late May 2026. Roles are analytical, not legal designations.
Banks do not enter markets because they are interesting. They enter because there is revenue, and the revenue case has now appeared in pieces. The event-contract stack offers a dealer at least five identifiable fee pools, several already live:
| Revenue line | Status & who is moving |
|---|---|
| RFQ / block spreads | The institutional-size channel; the natural home for Tradeweb- and MarketAxess-style protocols as blocks arrive |
| Structured products & event-linked notes | Live — Marex issued the first $10M prediction-market note (7% if Nvidia stays #1) |
| Prime brokerage & financing | Margin, leverage, and cross-margin against futures — the FCM-adjacent opening |
| Market making | SIG, Jump and DRW already run it; a bank principal desk is the open question |
| Data monetization | Selling the probability feed and the surveillance / resolution data as a product |
This is where Goldman (confirmed exploring), Morgan Stanley (on Kalshi's cap table), Marex (already issuing), and eventually BNP, Jefferies and BofA become part of the story — not as enthusiasts, but as desks chasing identifiable fee pools. The sell side enters when the spreadsheet works, and the spreadsheet is starting to.
A market structure is defined as much by its unresolved questions as by its participants. These are the five that will decide which 2029 actually arrives.
Those are the existential ones. The five fault lines below are the structural pressures that determine which 2029 arrives short of outright death:
Risk management here is not a configuration change to the options stack; it is a different methodology, because the payoff is binary and the question shifts from “how wide is the spread” to “how accurate is the price.” Two features of these markets are, properly handled, profit centers rather than hazards.
The spread benchmark. Execution cost stratifies cleanly by liquidity, and a desk needs to price the tier it is actually trading: roughly 1–2¢ on flagship high-profile contracts (major elections, marquee sports), 3–5¢ on moderate markets (a specific Fed decision, mid-cap crypto levels), and 10¢+ on thin or stale books where the midpoint is an unreliable signal. The trap is confusing quoted with effective spread: in a thin book a large order sweeps levels, so the cost of size runs well above the screen. This is the same block-versus-CLOB problem from the scale exhibit, priced.
The favorite-longshot bias. Prediction markets inherit a systematic, well-documented mispricing from a century of betting-market data: longshots (around 5¢) are chronically overpriced and heavy favorites (around 95¢) chronically underpriced, because retail overpays for tail lottery tickets. For a calibrated desk this is not noise — it is harvestable edge: sell the overpriced tails, hold the mathematically superior side, and treat calibration accuracy itself as the product. It is the clearest example of why information finance rewards a pricing discipline the retail order book does not have.
Forecasting this sector is a humbling exercise; the people who tried in 2024 undershot reality. So this is framed as three states of the world rather than a point estimate — with the honest admission that the probabilities below are judgment, not measurement.
Before the scenarios, a discipline worth stating plainly — separating what is observed from what is reasoned from what is wagered:
| Bucket | Example from this paper |
|---|---|
| Observable | $29.8B traded in April 2026; 17 DCM applications filed since early 2025; ICE's $2B commitment |
| Inference | Sports likely remains the dominant category through 2028; institutional share stays sub-10% until depth arrives |
| Scenario | A full-stack market could exceed $2T of annual volume by 2030 — a conditional path, not a forecast |
The scenarios below are the third bucket. They are deliberately framed as states of the world, with probabilities that are judgment, not measurement.
Preemption broadly holds but messily. The Supreme Court eventually clarifies that CFTC-registered event contracts are federal instruments, while carving room for state consumer-protection and integrity overlays. Congress passes a narrow framework — the most sensitive categories restricted, the rest legitimized. The sector consolidates into a handful of vertically integrated stacks.
Preemption is affirmed cleanly and early. A favorable framework passes. Depth arrives, the liquidity ceiling lifts, and event contracts become a normal line item in institutional risk management. The “truth machine” thesis is realized: prediction prices are a primary macro signal across the street.
The Supreme Court declines preemption, or Congress legislates a hard line, or the insider-trading and integrity scandals poison the politics. Sports contracts get pushed back into the state gambling perimeter. The federal-wrapper advantage collapses, the sector re-fragments, valuations reset hard, and the incumbents quietly write down their positions.
Every branch hinges on one node: whether a sports event contract is, finally and durably, a federal swap. Get that right and the rest of the forecast is detail. Get it wrong and no amount of capital, liquidity, or incumbency matters. The market is, fittingly, pricing the answer in real time — which is the most prediction-market thing about the entire situation.
Every market-structure shift is a transfer. The federal wrapper expands the pie for some participants and quietly erodes the ground under others. Where the value moves:
| Winner | Why |
|---|---|
| Exchanges & clearers | The federal wrapper expands addressable market to fifty states under one license; clearing is the layer hardest to disintermediate |
| Market makers | A brand-new volatility and probability surface with wide early spreads and cross-venue arbitrage |
| Brokerages | A high-engagement product that monetizes distribution they already own |
| Data vendors | A new signal stream to normalize, distribute, and sell across fragmented venues |
| Risk & infrastructure vendors | A new asset class that needs instrument representation, margining, and lifecycle tooling |
| Loser | Why |
|---|---|
| State sportsbooks | Federal preemption routes around the state licenses that were their moat |
| State regulators & tax bases | Jurisdiction — and gaming-tax revenue — erodes toward a federal regime |
| Traditional polling | A faster, money-backed probability signal competes with the survey |
| Sports-data monopolies | Pressure on data-licensing economics as a parallel venue prices the same outcomes |
Winner / loser placement is analytical judgment, not a prediction of any specific firm's outcome.
Set aside whether you want to trade a Super Bowl contract. The structural story matters to anyone who builds or runs institutional capital-markets infrastructure, for three concrete reasons.
Binary event contracts are arriving inside CFTC-regulated venues, cleared through DCOs, surfaced on the same terminals as rates and credit, and now wrapped into structured notes. If they cross into real-money hedging — the bull case — execution, risk, P&L, margin, and lifecycle systems will need to represent them as first-class instruments. The firms that can model a discrete-outcome payoff alongside a swap will be ready; the ones treating it as gambling will be late.
Even in the bear case, prediction-market prices remain a fast, skin-in-the-game probability feed for macro and event risk — now Fed-validated as competitive with consensus forecasts. Treated as an input — calibration surface, early-warning indicator, a cross-check against survey-based forecasts — it has value independent of whether any institution ever trades a contract. The signal survives the venue.
The sportsbooks built a state-by-state empire and assumed the structure was the moat. A federal wrapper made the moat a liability. The same logic has come for printing, for trading floors, for settlement messaging, and it will come for whatever today's incumbents assume is permanent. The useful posture is not to predict which way prediction markets resolve. It is to keep your own plumbing flexible enough that you don't much care.
The hardest layer to retrofit is risk infrastructure. Event contracts break assumptions traditional risk systems were built on: payoffs are discontinuous and settle to a binary, not a smooth distribution; positions across contracts carry hidden correlation shocks, since one geopolitical trigger can resolve a dozen contracts at once; and the underlyings include political and geopolitical events with no historical volatility series to lean on. Margining a binary, stress-testing a book of them, and representing them alongside swaps and options in a single risk picture is not a configuration change — it is new methodology. The firms that treat event contracts as a first-class instrument rather than a novelty will be the ones ready when the institutional flow arrives.
Two years ago this was a courtroom curiosity. Three years from now it is either a normal corner of the derivatives market or a cautionary tale about jurisdiction. The contracts, naturally, are already trading on which.
Volume decomposes into five categories — the retail engine (sports, mentions) and the emerging macro book (elections, economic, geopolitical) — each with its own growth. Move the four drivers and the curve re-prices live. When every driver clears ~55, a super-additive “stack” synergy engages: the unlocks stop adding and start compounding.
Anchors: $44B in 2025, ~$340B in 2026. 2025–26 are fixed; the levers drive 2027–2030. Growth rates, splits, sensitivities and the synergy term are illustrative and editable, not a forecast.
A multi-trillion-dollar number sounds enormous until you put it on the same ruler as the rest of the system. Keep flow (annual volume) separate from stock (notional outstanding), and remember today's volume is tiny retail tickets while the institutional flow arriving is a handful of very large blocks.
Sources: AGA (US sports-betting handle ~$148B, 2024); CoinGlass/CoinGecko (crypto-derivatives ~$85.7T, 2025); BIS (OTC notional $846T, June 2025); ISDA. PM figures illustrative; flow and stock are different metrics.
The projector draws one curve per scenario; reality is a distribution. This runs thousands of Monte Carlo paths to 2030 in parallel — each a different roll of the dice — and renders them as a probability cloud. The honest output isn't “$1 trillion” but “a 47% chance of clearing a trillion under these assumptions.” It runs live in your browser.
Illustrative Monte Carlo. Each path applies the projector's logic plus a per-year stochastic shock whose volatility grows with horizon. Growth rates, sensitivities and the synergy term are editable assumptions, not a forecast.