Author: Owen, GO2MARS's WEB3 Research
This article will provide an in-depth analysis of the prediction market ecosystem after the World Cup, offer a deep assessment of its current market landscape and industry trends, and analyze the main arbitrage profit strategies.
I. Introduction
2026 is a year of rapid expansion for prediction markets. The World Cup brought in a large number of new users, and the industry's driving force has expanded from a small number of elections and events to normalized trading covering the entire year.
Since entering October, three main threads in the industry have received concentrated attention:
• Kalshi's compliance process continues to advance, and the jurisdictional dispute between the federal government and the states has entered the Supreme Court stage;
• Polymarket's token issuance process has been repeatedly delayed. The founder's statements around Token 2049 did not provide a clear timetable, and the market prices its token issuance within the year at only 5% (as shown in the figure below);
• The founder of Predict.fun stated that it will issue a token within a few weeks, and the project is regarded as one of the most closely watched airdrops of the year.
Figure: Market expectations for Polymarket's token issuance timing
Recently, the most concentrated discussion has actually been about Opinion. Opinion is a prediction market on the BNB Chain, led by YZi Labs, and was once regarded as the core project within the Binance ecosystem benchmarked against Polymarket, attracting a large number of points participants and retail users before its March TGE.
The trend after the TGE exceeded participants' expectations. The total airdrop accounted for 23.5%, but only 3.5% (of total supply) was unlocked at the TGE, with the rest released linearly over 7 months; the project team also cut and cleared some points before and after the TGE, and many participants received less than their input costs. Since then, the token price has continued to decline, approaching historical lows in September, down more than 90% from its historical high.
Figure: OPN token price trend (March to October 2026)
On October 4, three leading KOLs publicly criticized the project team; the next day, the three were removed from the group, and their shares in the KOL round were simultaneously reclaimed by the project team (as shown in the figure below). The community reacted coldly, because when the airdrops of points participants were significantly cut half a year earlier, no one spoke out publicly either.
Behind the heated discussion of the incident is the booming industry narrative—prediction markets are undoubtedly the most closely watched track in 2026, and their popularity has not faded with the end of the World Cup. Their form is extending from a single betting platform to an information aggregation institution and financial infrastructure.
Based on this, this article will provide an in-depth analysis of the prediction market ecosystem after the World Cup, offer a deep assessment of its current market landscape and industry trends, and conduct an in-depth analysis of the main arbitrage profit strategies.
II. Market Status: Demand Continues After the World Cup, a Duopoly Has Basically Formed
2.1 User Demand: No Decline After the World Cup
Previously, there was a judgment in the market that prediction markets were only a short-term trend driven by the World Cup, and demand would quickly fall back after the event ended. Data from the two months after the World Cup do not support this judgment: trading volume fell slightly in July, then hit a historical high in September (as shown in the figure below).
The changes from August to September can be broken down incrementally. According to DeFiRate statistics, nominal trading volume in August was about $50.55 billion, rising to about $77.65 billion in September, an increase of about $27.1 billion, or 54%.
Table: Nominal trading volume of major platforms from July to September 2026 ($100 million)
The above figures are calculated on a nominal trading volume basis, that is, each contract is counted at a face value of $1, which differs from the funds actually paid by users. If calculated by actual funds paid, September was about $20.78 billion, only 23% higher than August, less than half of the nominal increase.
Breaking down the above increment item by item shows that the $27.1 billion amount actually mainly came from three items, and the largest share came from "Parlay."
Kalshi parlays: about $17.8 billion, accounting for 66% of the increment. A parlay packages multiple games into one contract. For example, betting on three teams to win at the same time pays out $1 only if all are correct, and goes to zero if one is wrong. Such contracts are difficult to win, and their unit price is often only a few cents to a dozen cents. One dollar can buy multiple contracts, which are recorded as several dollars in trading volume by face value. According to DefiLlama statistics, parlays as a share of Kalshi's nominal trading volume rose from 1.7% in November 2025 to 46.5% in August 2026, and approached 60% in mid-September. Therefore, Kalshi's nominal trading volume in September grew 52%, while actual funds paid grew only 21%.
New entry points: about $6.4 billion, accounting for 24% of the increment. Polymarket's September trading volume increased by $4.6 billion, with all the increment coming from the U.S. site. The U.S. site's monthly trading volume more than doubled, while the international site declined slightly. New platforms such as Novig and Robinhood's Rothera contributed about $1.8 billion in total. These platforms directly target U.S. users, and some are embedded in brokerage or betting apps familiar to users, with almost no barrier to entry.
New-season single games: about $2.9 billion, accounting for 11% of the increment. In early September, the NFL and U.S. college football seasons began at the same time, and Kalshi's ordinary single-game contract trading volume rebounded accordingly. Most parlay bets were on football, and the most active events on new platforms were also sports events. The new season formed the underlying backdrop of this round of growth.
Figure 5 Breakdown of nominal trading volume increment from August to September 2026
Overall, the main impact left by the World Cup was user education and channel penetration, and new users did not churn after the event ended. Apart from parlays, newly opened entry points on various platforms (Hyperliquid, Robinhood, and Predict.fun within the Binance ecosystem) and the new season are all continuing to bring incremental growth. The industry's driving force has expanded from single major events such as elections and the World Cup to pan-sports events throughout the year, and cyclicality has clearly shortened.
2.2 Industry Landscape: The Duopoly Has Basically Taken Shape, with Kalshi in the Lead
According to DefiLlama statistics on the past 7-day trading volume, the top ten platforms totaled $3.70 billion, of which Kalshi accounted for 61.9%, and Polymarket (international site and U.S. site) accounted for 32.7%. The two together accounted for 94.6%, while the remaining 8 accounted for 5.4%.
Figure Share of prediction market platforms' past 7-day trading volume (top ten)
The relative size of the two depends on the statistical basis. By DefiLlama's actual trading basis, Kalshi is about 1.9 times Polymarket; by the nominal basis including parlay face value, Kalshi's September trading volume ($60.69 billion) was about 4.5 times Polymarket's ($13.40 billion).
The category structures of the two platforms underwent convergence and then re-divergence (as shown in the figures below):
Figure: Kalshi and Polymarket category structure: convergence before and after the World Cup (January, March, May, July 2026)
Figure: Kalshi and Polymarket category structure: re-divergence in September
Data source: public platform data, compiled by Go2Mars Research Institute; category share calculated by nominal trading volume
Before the World Cup, Kalshi's sports share was between 77% and 89%; Polymarket's sports share was 43% to 57%, with crypto and politics each accounting for 15% to 26%, with the three categories standing side by side.
During the World Cup, Polymarket concentrated toward sports, with the sports share rising to 82% in July, close to Kalshi's 78%, and the two platforms' categories converged.
After the World Cup, divergence appeared again. Kalshi broadly promoted parlays, with sports single games accounting for 22% in September, parlays and others accounting for 61%, and sports-related categories totaling more than 80%; Polymarket split into two segments, the international site and the U.S. site. The U.S. site was 98% sports single games (and had begun parlay promotion), while the international site's sports single games accounted for only 46%, with the rest being crypto, politics, and other categories, and it had also begun parlay pilots.
Accordingly, the development directions of the two major platforms are already clear: Kalshi concentrates on sports events, pays almost no attention to political contracts, and sports parlays occupy most of its share; Polymarket's U.S. site is almost entirely sports single games, while the international site retains broader categories such as crypto and politics.
2.3 Emerging Platforms: Limited Scale, Varying MechanismsIn addition to the two leading platforms, current new entrants include Binance-affiliated Predict.fun and Opinion, Hyperliquid's HIP-4 event markets, and Rothera, a joint venture between Robinhood and Susquehanna.
Table: Overview of Major New Entrants in Prediction Markets
In terms of scale, Predict.fun and Opinion, as the third- and fourth-ranked markets, together account for less than 4% of cumulative trading volume, a clear gap from the two giants. However, emerging prediction market platforms of various kinds are more diverse in product design, with innovation concentrated in two areas:
Mechanism innovation. For example, introducing perpetual contracts or even further introducing leverage, allowing event contracts to share margin with perpetual contracts (HIP-4 adopts this design); another example is allowing users to create their own markets, giving rise to prediction markets oriented toward social networks (SocialFi+Prediction).
Theme expansion. Represented by Predict.fun, its themes lean more toward entertainment and social hot topics, attracting new users through diversified markets, and it can even list real-time event predictions on hot topics such as "Justin Sun and Jing Tian."
Based on the above analysis, the current industry landscape has three notable characteristics.
First, prediction markets no longer rely on short-term speculation around special events and are taking shape as a new type of exchange, while the World Cup's greatest role was completing user education and popularization.
Second, the industry landscape is basically solidified, with Kalshi and Polymarket together accounting for more than 90% of volume, and Kalshi clearly leading.
Third, both platforms' themes are mainly sports, but their paths differ: Kalshi relies on sports parlays, while Polymarket's international site retains more diverse content such as crypto and politics.
III. Evolution of the Industry Narrative: From Gambling to Financial Infrastructure
The development direction of prediction markets is to become a new type of exchange, and their narrative has undergone three layers of evolution: first seen as gambling, then as an information aggregation tool, and ultimately pointing toward hedging tools and underlying financial infrastructure.
3.1 Three Layers of Narrative: Gambling, Information Aggregation, and Hedging Tools
The first layer is gambling. The underlying trading of prediction markets is close to sports betting; in the current market, about 80% of markets are related to sports events, and the main difference from traditional sports betting lies in licensing. Licensed channels in the United States bring this demand into a futures framework, which has triggered a jurisdictional dispute between the federal government and the states (
The second layer is information aggregation. Prices correspond to probabilities; anyone can see the market's pricing of an event and can also bet based on their own judgment, with information advantages flowing from institutions and insiders to ordinary participants willing to bear risk. Supporters often compare it with polls, arguing that prices formed with real money are faster than questionnaires and less susceptible to rhetoric. The value of this layer is most obvious in contracts on elections, macroeconomic data, and technology events, and it is also the main reason platforms seek public opinion and regulatory support.
The third layer is the vision the industry has always emphasized, namely becoming underlying financial infrastructure. Event contracts are structurally binary options and can be split and combined to form hedging tools. Compared with ordinary options, event contracts have a single payout structure, small per-contract amounts, and all-day trading, making them suitable for hedging policy and event risks that traditional derivatives do not cover. When market depth is sufficient, they are expected to form a new type of hedging platform; combined with perpetual contracts, capital efficiency can be further improved.
Table: Comparison of the Three Layers of Prediction Market Narratives
3.2 Validation of Institutional Demand: Pilot Programs for Client Risk Hedging
At present, B-end institutional clients have already piloted risk hedging. On August 19, 2026, broker Cantor Fitzgerald announced that it would open Kalshi event contracts to about 3,000 institutional clients, including hedge funds and family offices, with Cantor offering them more precise event contracts as an asset hedging service. This pilot involves two important issues:
First, why are prediction markets suitable for risk hedging? Event contracts can correspond to a specific single risk. Take holding Apple stock as an example: to hedge downside risk, one need not hedge the entire North American technology sector, but only buy contracts on iPhone sales falling short of expectations. The same logic can be applied to the AI industry chain, such as Nvidia chip sales or OpenAI customer numbers. Institutions no longer need to short ETFs or a basket of tech stocks to indirectly hedge through market Beta; they can directly target Alpha risk at the individual stock level.
Second, can Kalshi handle such trades? That is, are liquidity and depth sufficient? To this end, Cantor brought in market maker Susquehanna Predictions to provide institutional-scale quotes and liquidity. Block trades are negotiated and executed at a single price outside the Kalshi platform, and institutions can also propose customized contracts for specific risks.
Figure: Cantor's Path to Bringing Prediction Markets to Institutional Clients
Similar hedging trades also occurred on Polymarket. On June 2, 2026, Polymarket completed its first on-chain institutional block trade, between FalconX and AneraLabs, involving a six-figure USD hedge linked to Nvidia H100 compute prices, settled using the Ornn Compute Price Index (OCPI). We analyzed this in detail in our previous article "When Compute Itself Becomes a Tradable Asset — From Compute Futures to Compute Dollars."
Figure: Compute Price Hedging Block Trade on Polymarket
From this, three judgments can be drawn:
First, B-end institutions' hedging demand for event contracts has been validated. Both cases—Cantor with Kalshi and FalconX with Polymarket—have been validated.
Second, at the current stage, order book depth is insufficient to handle large hedges, and institutional trades are mainly completed through off-exchange negotiated block trades.
Third, at the current stage, the platform's role is contract design, clearing, and compliance, while liquidity mainly comes from professional market makers. For prediction market platforms, deeply cooperating with market makers and exploring the establishment of instant liquidity for special assets is the main direction for attracting B-end institutional clients.
3.3 Capital and Regulation: Rising Valuations and the Compliance Game
The evolution of the narrative and institutional pilots have driven continued capital inflows, but one point must be faced squarely: the industry's progress is still constrained by regulatory compliance.
Capital pricing. The two major platforms have been continuously marked up in valuation since 2025. Kalshi's valuation rose from USD 2 billion in June 2025 to USD 22 billion in May 2026, while Polymarket rose from USD 1.2 billion to about USD 15 billion in April 2026. According to media reports, both are negotiating a new round of financing, with valuations of about USD 40 billion and more than USD 20 billion, respectively.
Figure: Valuation Progress of Kalshi and Polymarket
By comparison, it can be seen that Kalshi's valuation increase is larger and its financing pace is denser, reflecting that Kalshi is more recognized and favored by the capital market. On the other hand, the latest financing round has been slow to reach an agreement, which is actually constrained by the current regulatory compliance stance.
The main regulatory variable is the jurisdictional dispute between the federal government and the states, and there is currently no consensus. The starting point of the dispute is who should regulate Kalshi's sports contracts:
Kalshi is a CFTC-registered contract exchange and argues that sports contracts are swaps under the federal Commodity Exchange Act, that federal law preempts state law, and that state gambling laws have no jurisdiction;
The states argue that such contracts are essentially sports betting and must obtain state licenses. The two sides have therefore litigated in multiple states, with cases in both federal and state courts.
Table: Kalshi U.S. Regulatory Compliance Progress Table
On the federal court track, the courts' judgments have diverged. The same federal legal question has led to opposite conclusions in different circuits, giving the Supreme Court a reason to intervene. New Jersey, as the losing party in the Third Circuit, filed a petition for certiorari with the Supreme Court on September 2, and 39 states subsequently jointly supported accepting the case. Kalshi's deadline to respond is November 9, and whether the case will ultimately be accepted will only be determined after that.
The state court track is also advancing in parallel, mainly through lawsuits filed against Kalshi. States such as Massachusetts and Connecticut have directly sued Kalshi under state law, with cases heard by their own state courts, and Kalshi has raised federal preemption as a defense. The outcomes of such cases will also affect the situation the Supreme Court will face in the future.
At the federal administrative level, the CFTC clearly stands on the side of prediction market platforms. In April 2026, the CFTC sued Arizona, Connecticut, and Illinois, arguing that it is the sole regulator of prediction markets and that state government restrictions should be preempted by federal law.In summary, there is still no definitive conclusion at the U.S. federal administrative and legislative levels, and before a nationwide uniform law is formed, the legality of Kalshi's operations across states will continue to vary.
IV. Technical Operation: Price Formation and Settlement Mechanism
The technical layer of prediction markets can be broken down into three links: how shares are created (collateralization and minting), how prices are formed (matching), and how outcomes are confirmed and paid out (oracles and settlement).
4.1 Technical Underpinning: Taking the Lifecycle of an Event Contract as an Example
Prediction market contracts are binary event contracts: if the event occurs, each Yes share pays out $1; if the event does not occur, it pays out 0. From creation to payout, a contract goes through five links: collateralized minting, matching and trading, event occurrence, oracle adjudication, and redemption.
Figure: The Lifecycle of an Event Contract
At the underlying architecture level, mainstream platforms have three routes (as shown in the table below). Kalshi is closest to a traditional exchange, with matching, clearing, and adjudication all completed by licensed entities, and users deposit funds in U.S. dollars, at the cost of access being affected by regulation and state law. Polymarket adopts a hybrid architecture: matching is completed off-chain to ensure speed, while trades and fund settlement are on-chain, and positions are recorded as ERC-1155 tokens. Hyperliquid places event contracts directly into its own trading engine, sharing an account with perpetual contracts and spot.
Table: Comparison Table of Technical Routes of Major Prediction Market Platforms
4.2 Price Formation Mechanism
Price Is Probability, Anchored by the Complete Set
The price of each contract ranges from $0 to $1, corresponding to the market-implied probability, with 25 cents meaning a 25% probability. Prices are constrained by the complete-set mechanism: $1 of collateral can mint 1 Yes share and 1 No share, and conversely, a set of Yes and No can be merged and redeemed for $1. If the sum of the prices of Yes and No is below $1, buying both and merging them yields a profit; if above $1, the reverse operation applies. Arbitrage behavior anchors the sum of the two in a binary market at approximately $1.
In multi-outcome events, Polymarket sets mutually exclusive outcomes as a negative risk (neg risk) combination: exactly one outcome is Yes and the rest are No, and the sum of the Yes prices of each outcome is likewise anchored near $1. The price summation arbitrage in Section 5.1 is the recovery of deviations from this anchor.
Order Book: Execution Price and Displayed Price
The order book is the price discovery method adopted by mainstream platforms. Taking Polymarket as an example, the price displayed on the page is the midpoint between the best bid and best ask; when the bid-ask spread exceeds $0.10, it switches to displaying the latest traded price. In actual execution, the buyer trades at the ask price and the seller trades at the bid price. When two complementary orders are matched (for example, a Yes bid at 0.60 and a No bid at 0.40), $1 of collateral mints a complete set, and the Yes and No are delivered to the two buyers respectively.
Regarding fees, after Polymarket V2, fees are set by the protocol at the time of matching according to the market. The taker fee varies with price, with the formula fee = C × feeRate × p × (1 − p), where p is the execution price, C is the number of shares traded, and feeRate is the fee coefficient for each market; makers are not charged. The fee is therefore highest near 50 cents and decreases toward both 0 and 1.
Figure: Taker Fee as a Function of Contract Price (Illustrative)
Data source: Polymarket official documentation (V2 migration notes), Go2Mars Research Institute estimates
AMM: Another Route
The order book relies on market makers to continuously quote, and newly opened or unpopular markets often have no one posting orders. AMM replaces quotes with liquidity pools and formulas, allowing anyone to trade at any time. Early prediction market projects (such as Gnosis's Omen) used constant product AMM, but such mechanisms are not naturally suited to binary event contracts: the price of an event contract changes with probability and time remaining, and liquidity providers are almost certain to lose before expiry, with the value extracted by arbitrageurs (LVR) especially large when the price is close to 0 or 1.
Paradigm's pm-AMM proposed in November 2024 addresses this problem: assuming event outcomes follow Brownian motion (such as point differentials), it keeps the expected LVR as a constant proportion of pool value, concentrates liquidity near 50% probability, and gradually withdraws it as expiry approaches. The scope of that research is events such as matches where information continuously arrives, and it is not suitable for one-off events, and the analysis does not account for fees. At present, mainstream platforms still mainly use order books, with AMM used more for cold-starting long-tail markets.
Table: Comparison of Three Types of Pricing Mechanisms in Prediction Markets
Collateral and Capital Efficiency
Prediction markets use full collateralization: the price paid to buy Yes is the maximum loss, selling is equivalent to buying No, and there is no insufficient margin or liquidation. The benefit is settlement safety and low counterparty risk; the cost is capital occupation. For long-term contracts, the capital invested cannot be recovered until the event settles, and the price is depressed by the cost of capital, which is also the reason for the low annualized returns of long-term contracts in Part V. Platforms therefore compensate for the occupation with rewards: Polymarket pays holding rewards of up to about 4% on some long-term markets, and Kalshi also pays floating interest on account funds and positions.
Institutional capital is more sensitive to capital occupation. On September 22, 2026, Kalshi Klear applied to the CFTC to adopt risk-based margin for eligible event contracts: collateral is calculated based on expected short-term risk and no longer equals the maximum possible loss. The application is only for institutions (futures commission merchants, clearing members, etc.), limited to long-term economic, financial, political, and commercial contracts, while sports, culture, and others remain fully collateralized. It is currently in a 45-day review period and has not yet taken effect. Hyperliquid takes a different route: event contracts and perpetual contracts share the same margin account, but event contracts themselves have no leverage and no liquidation.
4.3 Oracles and Settlement
Settlement adjudication determines what the contract ultimately pays out, and it is the most disputed link in the prediction market technology chain.
Table: Comparison Table of Three Settlement Adjudication Methods
Polymarket's process is as shown in the figure below: the proposer stakes about 750 pUSD to submit an outcome, entering a 2-hour challenge period; if there is no challenge, it settles; if there is a challenge, a new proposal is made, and if challenged again, it is handed over to UMA token holders for a vote, with 1 to 2 days of debate and about 2 days of voting, totaling 4 to 6 days.
Figure: Polymarket's UMA Optimistic Oracle Settlement Process
This mechanism has exposed two types of problems:
The first is proposal quality: under the old OOV2, inexperienced proposers often proposed early, losing their deposit and also delaying market settlement. In August 2025, UMA passed UMIP-189, and Polymarket migrated to MOOV2, where only whitelisted addresses can propose, with an initial list of 37 addresses.
The second is concentration of voting power: in March 2025, a large UMA holder used voting power to influence the adjudication of a certain market, which Polymarket called an unprecedented governance attack.
On the other hand, for short-cycle crypto price markets, Polymarket switched to Chainlink price feeds for automatic settlement, no longer relying on manual proposals.
Hyperliquid, meanwhile, starting May 25, 2026, has validators publish the results of off-chain events with the help of automated news feed software, with the entire process closed within the Hyperliquid L1 and not relying on UMA, Chainlink, or Pyth.
4.4 Expansion Direction: The Convergence of Event Contracts and Perpetual Contracts
Perpetual contracts are derivatives with no expiry date that anchor to the spot price through funding rates, and are the most liquid products in the crypto market. Event contracts have an expiry date, and trading is concentrated around matches and event nodes. At present, three convergence paths have emerged between the two types of products:
Table: Three Convergence Paths of Event Contracts and Perpetual Contracts
There are two main reasons for the convergence. For platforms, perpetual contracts trade continuously and do not depend on the match cycle, which can compensate for the pulse-like nature of event contract trading; for traders, outcome contracts and perpetual contracts share margin, allowing event contracts to hedge the event risk faced by perpetual positions.
One point worth noting is that, at a deeper level, it is the perpetuation of event contracts themselves. Because event contract prices are bounded and converge to 0 or 1 at expiry, there is no continuous spot price to anchor to, and funding rates are difficult to apply directly. Therefore, the industry is currently exploring using probability indices as the underlying for perpetual contracts, or combining margin with perpetual positions within the same account. This product form is the most cutting-edge exploration and development direction of prediction markets.
V. Yield StrategiesPrediction markets have become a frontier area of research in arbitrage and quantitative trading. In recent years, several papers have studied mispricing and arbitrage opportunities within them, and the fields of finance and asset pricing have also begun to discuss these strategies.
When evaluating returns, long-term holding is used as the benchmark. For some long-term markets on Polymarket, the maximum holding reward is about 4% annualized; meanwhile, buying No in the "Second Coming of Jesus market" and holding to maturity equates to about 5.4% annualized. Such returns carry almost no risk of misjudgment, so: a strategy's annualized return must at least exceed this range for the capital tied up to be worthwhile.
Table: Representative arbitrage return strategies on prediction market platforms
5.1 Price aggregation arbitrage
Among all outcomes of an event, only one will pay out 1 dollar, and the rest go to zero. Therefore, by buying one share of every outcome, no matter which one occurs, 1 dollar can be recovered. If the total cost of buying them is less than 1 dollar, the difference is profit independent of the outcome.
For example, suppose an event has three mutually exclusive and exhaustive outcomes, and the Yes ask prices are 30, 35, and 32 cents respectively, totaling 97 cents. Buying all three means one will ultimately pay out 1 dollar, for a gross profit of 3 cents, about 3.1%. Such price gaps mostly appear in multi-outcome events, because the order books of each outcome are independent of one another and prices easily deviate. In binary markets, complementary orders for Yes and No are directly matched by the matching engine, so the spread is difficult to sustain.
Similar opportunities also exist between logically related markets, that is, simultaneously watching several markets that are mutually exclusive or logically ordered in sequence. For example: for a team to win the championship, it must first win the league title, so the price of the former should not be higher than that of the latter, and when an inversion appears, there is profit to be made.
Figure: Conceptual diagram of price aggregation arbitrage
5.2 Two-sided market making
Market making is placing orders simultaneously on both the bid and ask sides. For example, if the best bid for Yes is 49 cents and the best ask is 51 cents, a market maker places a buy order at 49 cents and a sell order at 51 cents, and if one share is filled on each side, they earn a 2-cent spread. These 2 cents come from the taker: the taker pays for immediate execution, paying 51 cents when buying and receiving only 49 cents when selling, and the spread is the price they pay for immediate execution.
The spread simultaneously compensates the market maker for two types of risk they bear.
First, inventory risk: when only one side is filled, a one-sided position remains in hand, and if the price moves in an unfavorable direction, there will be a loss.
Second, adverse selection: traders with information will specifically hit the market maker's quotes, at which point the quotes are already stale, and the market maker is effectively taking orders at old prices.
At the same time, platforms often provide additional incentives to makers. Polymarket does not charge makers and returns 15% to 25% of taker fees to market makers by category; Kalshi also does not charge makers on most contracts. The rebated funds likewise come from takers, with the aim of encouraging order placement and deepening the book, while the above risks are still borne by the market makers themselves.
5.3 End-of-market sweeping
End-of-market sweeping is buying when the outcome is basically determined and the price has risen to 95 to 99 cents, then holding to payout to earn the remaining few cents. Taking buying Yes at 97 cents as an example: at maturity it pays out 1 dollar, and after deducting about 0.2 cents in fees, a single trade nets about 2.8 cents, for a net return of about 2.9%.
A single-trade return of only 2.9%, but because the holding period is short and capital is recovered quickly, it can be reinvested repeatedly. Simplifying the single-trade return to a one-day operation and annualizing it by the number of times it can be repeated within a year gives the annualized return, calculated as 2.9% x 365 / remaining days. If 30 days remain, the annualized return is about 35%; with 90 days remaining it is about 11.7%, with 180 days about 5.8%, and with 365 days only 2.9%. The 35% premise is that contracts with 30 days remaining and a price of 97 cents can be found throughout the year, which is a theoretical upper limit.
Figure: Annualized returns of end-of-market sweeping and long-term holding returns within the platform
The chart above also marks two long-term holding benchmarks: the maximum holding reward for some long-term markets on Polymarket is about 4% annualized, and buying No in the "Second Coming of Jesus" market and holding to maturity is about 5.4% annualized.
The farther from payout, the lower the annualized return of end-of-market sweeping: with about half a year remaining, it is comparable to these two benchmarks, so this strategy is more suitable for contracts with short remaining time.
The risk of end-of-market sweeping is concentrated in the tail: one misjudgment loses 97 cents, while each win earns only about 2.8 cents, requiring about 35 wins to recover. In addition, once a Polymarket result enters UMA dispute, funds must be tied up for an additional 4 to 6 days, and the result may also be reversed.
5.4 Points arbitrage: the case of Predict.fun
Points arbitrage earns the expected token airdrop, not the trading itself. Predict.fun distributes Predict Points (PP) to users every week. The official documentation states that most points go to makers who provide liquidity: orders must meet a minimum size, the bid-ask spread must be below a threshold, and two-sided quoting receives the highest reward; trading and holding also earn points. The market expects the points to be redeemed for a token airdrop, and the official token issuance arrangements have not yet been announced.
The cost of points varies by role. Fees are paid by takers, with the official base rate at 2%, about 2% when the price is no higher than 50 cents, and lower as the price rises; makers are not charged, and some markets can also receive a rebate of 25% to 30% of taker fees (trial through the end of October). The maker's cost is therefore mainly inventory risk and adverse selection, while the taker's cost is the fee.
How much a single point is worth depends on the total number of points, the airdrop share, and the valuation after token issuance. According to an Odaily estimate, cumulative points are about 430 million, expected to increase to 460 million to 500 million at token issuance.
Figure: Estimate of cumulative Predict.fun points distribution
There is a corresponding market on Polymarket for the fully diluted valuation (FDV) one day after token issuance, and the chart below shows its implied probability distribution (including the probability of whether a token is issued): the probability of FDV above 800 million dollars is about 50%, and above 1 billion dollars about 38%. The official airdrop share has not been announced, and Odaily assumes 5% to 20%. Based on an FDV of 800 million dollars and 500 million points, each point is worth about 0.08 to 0.32 dollars.
Figure: Implied probability of Predict.fun FDV one day after token issuance
About 10 million points are distributed each week, corresponding to a value of about 800,000 to 3.2 million dollars; DefiLlama shows that Predict.fun's fees over the past 7 days were about 1.46 million dollars. When the airdrop share is 5%, weekly fees are about 1.8 times the value of points; at 10%, about 0.9 times; at 20%, about 0.5 times. Only when the airdrop share exceeds about 9% is the value of points higher than fees. This comparison pits platform-wide fees against the value of all points, while takers receive only a small portion of the points, so the actual return is lower than shown in the chart.
Figure: Fees as a proportion of the value of points per dollar of trading volume
But one point that needs to be noted is that all current strategy figures are estimates. The final result depends on the project team's airdrop share and unlock arrangements. Before the final airdrop plan is distributed, no one can guarantee that the anti-farming crackdown and reduction of airdrop shares seen with Opinion will not happen again.
5.5 Cross-platform arbitrage and high-frequency bots
The principle of cross-platform arbitrage is the same as price aggregation arbitrage, except that the two trades are split across two platforms: buy Yes on Polymarket for the same event while buying No on Kalshi, and if the total cost is less than 1 dollar, profit can be locked in. For example, if Yes is quoted at 51 cents and No at 45 cents, totaling 96 cents, the gross profit is 4 cents.
But such price gaps are usually only a few cents, and little remains after deducting fees on both sides. The greater risk lies in settlement rules: the two platforms may have different criteria for judging the same event, and when the rules are inconsistent, both trades may lose at the same time.
A similar approach can also introduce high-frequency bots to capture price gaps between different platforms at the same time, exploiting the time difference between platforms, as well as differences in oracle quotes between platforms, to conduct high-frequency, rapid arbitrage.
6. Trend assessment
Based on the above analysis, we make five judgments about the next stage of prediction markets.
First, after the World Cup, demand normalizes but does not shrink, and subsequent measurement of incremental growth should look at paid capital. After the World Cup ended, demand did not recede, and September volume hit a record high, with the driver expanding from single major events such as elections and the World Cup to pan-sports events throughout the year. However, September nominal volume grew 54%, while paid capital grew only 23%, as parlays magnified the face value. From then on, platform scale should be measured by paid capital or actual volume.Second, the focus of competition has shifted from platforms to gateways. Kalshi and Polymarket together account for 94.6% of trading volume over the past seven days, and the landscape is unlikely to change in the short term. The incremental growth of new entrants mainly comes from embedding into existing accounts and channels: Rothera within Robinhood, HIP-4 within perpetual accounts, and Cantor serving institutional clients. For new entrants, the value of a channel is higher than building a trading venue from scratch.
Third, regulation determines the ceiling of the industry. The federal circuit courts are divided on whether federal law preempts state law. The Third Circuit ruled in favor of Kalshi, while the Ninth and Sixth Circuits ruled against it. Whether the Supreme Court takes up the case will not be determined until after the November 9 response deadline. Before a unified national rule takes shape, the legality of Kalshi's operations across states will continue to diverge, and its fundraising progress will also be affected.
Fourth, demand from institutional B-end clients is already very clear, and prediction markets are expected to take on the role of hedging platforms and financial infrastructure. Cantor opened Kalshi's event contracts to approximately 3,000 institutional clients, and FalconX and Anera Labs completed an on-chain block hedging trade linked to computing power prices on Polymarket. Both cases stem from genuine risk hedging demand. Event contracts can correspond to a single specific risk, and institutions no longer need to hedge indirectly through an entire sector or ETF. Demand has already emerged, and whether it can be absorbed depends on depth and capital efficiency.
Fifth, capital efficiency is the prerequisite for absorbing B-end demand. Event contracts are fully collateralized, tying up a high amount of capital, and order book depth is also insufficient to absorb large hedges. At this stage, this is mainly handled through OTC blocks and professional market makers. The risk margin sought by Kalshi Klear, as well as the integration of prediction markets and perpetual contracts, are both aimed at improving capital efficiency. We judge that event contracts will first mature in blocks and customized contracts, and then return to the order book.
VII. Conclusion
After the World Cup, demand in prediction markets has not receded. User education has already been completed, and the sports season and new gateways continue to bring incremental growth. Kalshi and Polymarket together account for more than 90% of trading volume, and the industry landscape has taken shape.
Once demand is firmly established, the ceiling of the industry is determined by two things:
First, compliance: how the Supreme Court handles Kalshi's case will determine whether licensed platforms can expand sports contracts nationwide.
Second, institutionalization: the cases of Cantor and Kalshi, and FalconX and Polymarket, show that B-end clients' hedging demand has already emerged, and whether that demand can be absorbed and scaled depends on whether order book depth and capital efficiency can keep up.
To observe the progress of these two things, paid-in capital, order book depth, and settlement dispute rates are more reliable than nominal trading volume. Only after these two things are resolved will prediction markets be able to complete the final step of the three-layer narrative, growing from a sports event betting platform into financial infrastructure for risk pricing and becoming an inseparable and important part of the future financial world.