Heretic Research: Prediction Market's Next Growth Area Shifts to Backend Infrastructure

Author: Heretic Research, an independent research institution focused on the intersection of Crypto and AI

From settlement friction to cross-platform infrastructure, seeking the next wave of industry alpha

This is the full version of the third research report from Heretic Research. The study covers 282,191 settled prediction markets as of July 15, 2026, and addresses the question: As platform competition has consolidated around traffic, licenses, and liquidity, where will the next wave of industry growth emerge?

The conclusion lies at the outcome layer—the second profit pool formed after standardizing and enabling cross-platform reuse of four capabilities: rule comparison, evidence verification, result confirmation, and payment triggering. All data in this article are point-in-time figures; sources and calculation methodologies are noted within the text.

TL;DR

• The front-end of prediction markets is becoming increasingly crowded; the next wave of incremental growth may come from backend infrastructure. Competition among platforms has already centered on traffic, licenses, and liquidity; however, each new market still requires rules to be explained, evidence to be verified, outcomes to be confirmed, and payments to be settled. These repetitive capabilities are now ripe for standardization and servicing multiple platforms.
• Uncertainty in rules and settlement is creating alpha. HR’s regression model shows that markets with 0.487% of disputes account for 8.64% of trading volume, and persistent spreads remain across similar contracts on different platforms; however, shallow order books and execution costs limit capital allocation. A more scalable opportunity lies in building a reusable product that automates rule comparison, evidence validation, outcome confirmation, and payment triggering.
• The results layer has entered commercialization. Current observable annual revenue is approximately $15 million to $37 million; if covering the entire existing market, this corresponds to approximately $64 million to $161 million; when monetization reaches the level of mature projects, the industry's annual revenue ceiling is approximately $456 million.
• Industry opportunities are taking shape, and the result layer remains in the optimal phase for early research and positioning. Platform equity has already priced in high front-end growth expectations, and HYPE and ICE still have limited capacity to absorb value from the result layer; what truly deserves tracking is which early projects are beginning to gain cross-platform calls and converting those calls into attributable, recurring revenue. Ideal targets have not yet fully emerged, meaning this segment of industry value has not yet been fully priced.

Those who lack trading discipline earn only modest profits; those who define the rules reap the future.

Introduction | The Overlooked Second Probability in Prediction Markets: Who Decides the Outcome, Who Rewrites the Price

Predictive markets often interpret the price of a Yes contract directly as the probability of the event occurring. This interpretation overlooks the most valuable variable in predictive markets: whether the rules recognize the event after it happens, which ultimately determines whether the contract is paid out.

A contract therefore contains two probabilities:

• The probability of the event occurring
• The probability that facts are recognized as Yes according to the rules

When rules are clear and evidence is consistent, the two align; when there are discrepancies in deadlines, information sources, evidence standards, or final rulings, they diverge. This difference is not a technical bug—on the contrary, it alters cash flow and can itself be considered an asset.

Therefore, this article will focus on this “invisible hand” of prediction markets that has yet to be recognized by consensus, and deconstruct it in depth using three definitions: the authority to interpret determines whether facts are acknowledged by rules, rule differentials represent the trading alpha introduced when this acknowledgment bias enters the price, and the outcome layer encapsulates rule comparison, evidence verification, result confirmation, and payment triggering as a cross-platform interface.

These three form a causal chain: the right of interpretation is the cause, the rule gap is the effect, and the outcome layer is the commercialized form of the right of interpretation.

Therefore, who decides whether facts can be recognized by the rules of prediction markets is the overlooked second probability, manifesting on the trading side as rule discrepancies and driving the encapsulation of rule comparison, evidence verification, result confirmation, and payment triggering into shared interfaces on the industry side.

The same rule splits profit attribution into two areas. The one attributed to the outcome layer is the second profit pool mentioned in the title of this report—it represents profit distribution measured by industry口径, and no single entity has yet disclosed it.

Based on comprehensive data evidence and industry research, this report identifies an underpriced growth opportunity: a results layer capable of cross-platform replication and generating sustained revenue will become the primary industry driver for the next stage of the market.

01｜0.487% of market entries sparked controversy, accounting for 8.64% of trading volume: large capital behind low probability events

Measured by the number of disputed cases, disputes are a low-frequency operational issue. Therefore, this metric systematically underestimates the economic value of rule differentials: although small in proportion, the economic value of rule differentials is determined by the amount of capital exposed to disputes.

As of July 15, 2026, this report includes 282,191 samples of dispute rates and regression analyses from markets that closed, had a final notional volume of at least $1,000, and retained a UMA status trail. UMA is the on-chain dispute resolution mechanism used by Polymarket. Within this sample set, markets that entered dispute accounted for only 0.487% of total markets but represented 8.64% of the final notional volume.

The rule differences are concentrated in the most expensive markets.

1.1 The higher the trading volume, the more concentrated the dispute risk.

Layered by final notional trading volume, the sample dispute rate increases monotonically across tiers, rising from 0.26% in the lowest tier to 30.14% in the highest tier.

Accordingly, this report employs regression modeling, and the results show that, after controlling for factors such as market year, category, and duration, trading volume remains significantly positively correlated with disputes: for every tenfold increase in market trading volume, the relative likelihood of a market entering dispute increases by approximately 2.27 times. This indicates that disputes are clearly concentrated in high-value markets. However, since trading volume may include transactions occurring after disputes have arisen, this result only demonstrates correlation and cannot determine whether high trading volume causes disputes or whether disputes generate more trading. Regardless of the direction, it has become a recurring cost for large sums of capital.

1.2 The right of interpretation for high-value transactions is moved from the backend to the frontend.

When a single event determines only thousands of dollars, ambiguous rules can be temporarily handled by operations staff; when a single event determines tens of millions of dollars, ambiguity in rules simultaneously affects fund disbursements, market-making positions, and platform reputation.

High-value trades therefore require four things to be visible before placing an order: whether the rules can be compared, whether the evidence can be verified, whether the outcome can be confirmed, and whether the payment can be triggered. Missing any one of these, market makers charge for uncertainty through wider spreads, smaller positions, or higher capital occupancy. These costs are ultimately passed on to traders, manifesting as higher execution costs, lower executable sizes, and poorer capital efficiency.

Therefore, the dispute rate of 0.487% is significant because the amount of funds locked up far exceeds its proportion in terms of transaction volume. When rule discrepancies are concentrated in the most expensive markets, the real question to ask is who holds the power to alter payment outcomes.

02 | Centralization of Interpretation Rights: A Minority of Wallets Hold Final Voting Power on Disputes

The original terms of a prediction market cannot cover all possible facts. Where the terms fall short, outcome verification shifts from rule enforcement to interpretation: whoever has the final say can rewrite cash flows. This is the first source of rule weakness: the allocation of interpretive authority—who ultimately decides within a platform; the more centralized this authority, the harder it becomes to predict the likelihood of rule adherence.

2.1 Strategy Case: An event occurring does not equate to contract acknowledgment

According to Polymarket's contract terms, the market should settle as Yes if the Strategy sells Bitcoin before 11:59 PM EDT on May 31, 2026. The Strategy later disclosed that it sold 32 BTC between May 26 and 31.

The market ultimately settled as No. The key dispute is not “whether it was sold,” but “whether the sale had to occur before the deadline or had to be publicly confirmed before the deadline.” The original rule was written based on the event occurrence time, while the supplemental clarification added on June 1st incorporated the public confirmation time into the judgment. Galaxy Research’s review of the rule text and supplemental clarification showed that traders were faced with two different standards for rule acknowledgment.

This is not “dispute occurred, so price fluctuated,” but a recognizable rule interpretation shock: the disclosure of facts pushed Yes up, and a supplemental clarification brought it back down. The facts didn’t change—what changed was the probability of those facts being recognized under the rules. Interpretation rights directly determine cash flow. This time, it was the supplemental clarification and the subsequent voting process that rewrote the cash flow.

2.2 Open participation has not eliminated centralization of power

This voting process is open to everyone. The dispute on Polymarket has entered the UMA vote, and any token holder can theoretically participate, but voting power is distributed according to token holdings.

A Bloomberg survey of on-chain voting found that among more than 6,400 addresses that participated in contentious votes over three years, just nine wallets accounted for about half of the voting power. Another investigation by The Wall Street Journal, as reported by the media, found that in over 300 contentious cases, at least one voter held a related market position.

No proof of manipulation is required—the conflict of interest itself constitutes a risk: a small number of wallets may hold relevant market positions while also participating in determining the final settlement outcome. When trading interests and outcome adjudication are concentrated in the same group of participants, the market must account for this potential influence in pricing. To assess the magnitude of this impact, one must first examine the costs and barriers to entering the dispute resolution process.

2.3 Large positions increase the economic incentive to engage in dispute resolution procedures

The threshold for initiating a dispute is low. According to Polymarket’s dispute documentation, disputants typically only need to match the proposer’s collateral, commonly $750. For those holding large positions, this cost may be far smaller than the funds at stake in the dispute outcome.

Suppose a trader holds a $100,000 position in Yes. If the market is proposed to settle as No, this position could be reduced to nearly zero. In this case, even if the trader believes the probability of a successful dispute is only 10%, spending $750 to initiate a dispute may still be a reasonable choice.

Ignore research, gas, and other additional costs. If the dispute is successful, the trader recovers their $100,000 position and receives half of the losing party’s margin, which is $375; if the dispute fails, they lose their own $750 deposit. Their expected profit is approximately:

10% × (100,000 + 375) − 90% × 750 = $9,362.50

In this example, even if the probability of success is low, the gap between the potential recovery amount and the margin is still sufficient to create a strong incentive to initiate a dispute. The larger the position, the lower the minimum probability of success a trader is willing to accept to initiate a dispute.

Therefore, the expected return formula for the dispute is:

Disputed expected return = q × (V + 0.5B) − (1−q) × B − C

Here, V is the position cash flow recoverable after a successful dispute, B is the dispute margin, C is the cost of research, gas, and opportunity cost, and q is the estimated probability that the favorable interpretation will be accepted.

Thus, the break-even condition for initiating a dispute is:

q > (B + C) ÷ (V + 1.5B)

This means that the larger the position value, the lower the minimum success probability required to initiate a dispute. $750 cannot determine the final voting outcome, but it is sufficient to allow a favorable interpretation to enter the cash flow determination process.

03 | The rule difference has already been priced in; strategy capacity is limited by order book depth

Previously, we identified who holds the right of interpretation. Next, we will see that each platform’s interpretation standards differ. This is the second source of rule disparity: differences in interpretation standards. The same second probability renders the former unpredictable and the latter unequal across platforms.

The rule differential has effectively become a tradable price discrepancy. Predictive market traders must not only assess whether an event will occur, but also verify whether two seemingly identical contracts use the same deadline, information source, and settlement criteria, and whether they pay out under the same conditions. Only when the settlement terms are identical can the price difference constitute arbitrage; if the rules differ, traders are effectively betting on whether the market has mispriced these rule differences.

3.1 Cross-platform pricing evidence: semantically equivalent, yet prices still differ

A 2026 cross-platform study covering ten major platforms and over 100,000 events, after manual verification of natural language descriptions, outcome confirmation semantics, and time ranges, found that some events were listed simultaneously on multiple platforms, and yet persistent, executable price discrepancies of 2%–4% remained between semantically equivalent markets. The study defined this phenomenon as “semantic non-fungibility”: without unified event identification and outcome confirmation standards, prices cannot fully converge.

A set of observations closer to the current order book showed the same direction. From July 1 to July 7, 2026, in the BTC year-end $120,000 contract with trading on both sides, Kalshi's Yes price remained approximately 2.08 percentage points higher than Polymarket's throughout the week. The Polymarket–Kalshi matched-question observations turned the concept of “rules and platform structure entering price” into a visible spread.

The market hasn't priced in the second probability; it simply hasn't yet developed a unified pricing language. For traders, this is precisely the source of Alpha.

3.2 Earn the spread under the same settlement condition, earn the judgment difference under different settlement conditions

A classic arbitrage strategy in prediction markets is cross-platform arbitrage. However, cross-platform trading first requires determining whether two seemingly identical contracts will actually pay out under the same circumstances.

Suppose Platform A and Platform B both have a market on whether Bitcoin will exceed $120,000 by year-end. Platform A’s Yes price is $0.54, and Platform B’s No price is $0.43, totaling only $0.97 on both sides.

If two contracts use identical expiration times, price sources, and settlement criteria, then regardless of the final outcome, one side will pay $1. After deducting fees, slippage, and funding costs, as long as the total cost remains below $1, traders can lock in the spread.

However, if Platform A uses the index price at 23:59 UTC on December 31, while Platform B uses the close price in Eastern Time, the two contracts may result in different settlements. In this case, buying both at $0.97 does not equate to risk-free arbitrage; traders are essentially betting on whether the market has underestimated the likelihood of different outcomes arising from these two sets of rules.

Therefore, cross-platform trading can be divided into two categories:

• Settlement terms are identical: traders profit from price discrepancies; as long as all costs are lower than the final payment amount, profits are locked in.
• Settlement conditions vary: traders profit based on their judgment of the rules, and their returns depend on whether their understanding of the cutoff time, information sources, and settlement criteria is more accurate than the market's.

For type two trades, what truly needs to be estimated is not whether the event will occur, but the probability that this contract will ultimately be ruled as Yes by the rules.

For example, if a Yes contract is priced at $0.54, and the combined fees, slippage, and funding cost amount to $0.015, then the trader must believe there is at least a 55.5% probability it will settle as Yes for the trade to have a positive expected value.

The break-even rule acknowledges that probability = market price + total cost per contract.

If a trader estimates the probability of a final settlement of "Yes" at 60%, which is above the breakeven line of 55.5%, the expected profit per contract is approximately $0.045; however, if the estimated probability is only 53%, the trade is not worthwhile, even if the trader believes the event is likely to occur.

Therefore, even if two contracts appear to be based on the same prediction event, they cannot necessarily be arbitrated directly. Traders must first determine whether they use identical settlement conditions; if the conditions are the same, profit comes from the price spread; if the conditions differ, profit comes from understanding the rules better.

3.3 The spread is genuinely arbitrageable, but the strategy's capacity is limited by order book depth.

The previous section explained how to identify and construct a arbitrage opportunity, but for a strategy to accommodate large volumes of capital, two more practical questions must be answered: Can these price discrepancies actually be realized, and how much capital can be deployed per opportunity?

First, the spread is not a theoretical opportunity on paper—it has already formed a strategic market. On-chain research of settled markets from April 2024 to April 2025 shows that intra-market rebalancing and cross-market portfolio trading generated approximately $40 million in realized arbitrage profits. This demonstrates that mispricings in prediction markets can be systematically identified and ultimately converted into actual returns.

Second, although spreads exist, the actual executable volume is extremely limited. A study covering 173 NBA games reconstructed over 75 million order book snapshots and identified 290 arbitrage opportunities. The median gross spread yield per opportunity was only about 1.01%; and 76.9% of these were constrained by order book depth, with an average executable volume of just 14.8 contracts per constrained opportunity. Opportunities for single-market anomalies were even fewer, with a median duration of only 3.6 seconds.

This means that if the gross yield of an arbitrage opportunity is approximately 1%, you could theoretically earn about $1 for every $100 invested; however, if only about 15 orders are available on the order book, each costing close to $1, then the actual capital you can deploy is limited to around $15, resulting in a gross profit of only about $0.15. Although the yield appears attractive, the amount of capital you can deploy is very limited, and the opportunity may disappear within seconds.

Therefore, arbitrage trading is a business that relies on research and execution capabilities, not an infinitely scalable market opportunity. Contract research, semantic matching, real-order-book identification, and multi-leg execution determine whether traders can seize opportunities; order book depth and opportunity duration determine how much capital this strategy can accommodate.

Traders earn one-time price discrepancies that must be repeatedly sought out and executed each time. What may truly scale with the overall growth of market trading volume is not personally trading these frictions, but rather infrastructure that charges ongoing fees to reduce regulatory uncertainty, improve outcome confirmation, and minimize manual discretion.

04 | The bigger alpha isn't in the order book: Turn interpretation into a product

Readers here may wonder why order books in prediction markets are generally thin. A key reason is uncertainty in settlement rules. Market makers cannot accurately determine how the platform will ultimately classify a given outcome, so they widen bid-ask spreads and reduce quote sizes to compensate for potential risk.

This uncertainty has given rise to two distinct business models. Traders seek and exploit mispricings caused by rule discrepancies, but their profits are limited by the order book depth of individual markets and the duration of opportunities; meanwhile, the outcome layer aims to reduce uncertainty around rules and settlement by turning rule comparison, evidence verification, result confirmation, and payment triggering into reusable services, charging multiple markets and platforms for access.

Both stem from the same market friction: traders profit from the existence of friction, while layers profit by reducing friction.

To achieve this, the results layer must first break down “how results are confirmed.” The difficulty of confirming different events varies: facts such as prices, times, and on-chain states can be directly read by machines; events with dispersed evidence but relatively clear rules can be assisted by AI in retrieval and organization; and events with ambiguous rules, large amounts involved, or conflicting evidence still require human judgment as the final decision.

The productization of the result layer begins with this division of labor. Each standardized confirmation task adds another capability that the platform can procure and invoke.

4.1 Deterministic outcomes are being standardized

The outcome layer first productizes the most clear and easily verifiable events, such as the price of Bitcoin at a specific time, whether a blockchain address has completed a transfer, or whether a certain state has occurred before a deadline. The common characteristic of these outcomes is that, as long as the data source, retrieval time, and judgment criteria are clearly defined in advance, the system can directly retrieve the data, determine Yes or No, and trigger payment.

For example, with the market question “Will Bitcoin be above a certain price at a specified time?”, the market only needs to predefine which price source to use, at which exact moment to read it, and what condition constitutes a “Yes.” Once the deadline is reached, the system automatically retrieves the data and settles the outcome without requiring any further interpretation of the rules.

Hyperliquid's HIP-4 is turning this process into a standardized product. It integrates binary outcome contracts into a unified trading system, enabling rule setting, trading, result verification, and payments to all occur within the same infrastructure, while allowing the entire template to be reused by the next set of events. This standardized product has already been validated by demand; Polymarket, powered by Chainlink, has累计 achieved over $3.4 billion in trading volume for short-term crypto markets.

Regulated platforms are also adopting the same approach. Certain gold, crude oil, and agricultural markets on Kalshi use external price data provided by Pyth as the basis for outcomes, demonstrating that result confirmation can transition from internal platform processes to a service provided by professional data vendors and reused across multiple markets.

Deterministic results are important not only because they are easier to automate, but because they first demonstrated that result confirmation could be broken down into a standard interface and become a foundational product of the results layer.

4.2 The value of AI is reducing evidence costs

Deterministic outcomes can directly retrieve data, but many prediction markets face a different challenge: the rules are clearly defined, yet relevant evidence is scattered across announcements, news articles, regulatory documents, or various information sources, sometimes even conflicting with one another. At this point, what truly consumes time is not making the final judgment, but locating the materials, filtering out valid evidence, and clearly organizing the basis for the judgment.

AI is best suited to handle this part of the work. It can accelerate information retrieval, synthesize data from different sources, identify conflicting evidence, and generate a package of evidence for final judgment. However, whether it can directly replace the final decision-making depends on real-world performance testing.

A study of UMA dispute markets shows that AI achieves an 89.58% agreement rate in post-dispute review, but only a 33.88% recall rate in pre-dispute identification. The former indicates that, with complete information, AI often reproduces the final ruling, but does not mean it has identified an independent “true answer” apart from the original judgment; the latter reveals that AI still cannot reliably detect most high-risk markets in advance.

Therefore, AI is currently better suited to be placed in the middle of the decision-making process, rather than at the final stage. What AI truly reduces is the cost of evidence processing before reaching a judgment.

Therefore, a more practical division of labor is: machines handle reading clear data, AI retrieves, filters, and organizes evidence, and humans make final judgments on markets with ambiguous rules, large amounts, or conflicting evidence.

Once this division of labor is standardized, result confirmation is no longer just an internal operational process within a single platform. Rule comparison, evidence verification, result confirmation, and payment triggering can be broken down into independent interfaces, allowing multiple platforms, market makers, institutions, and agents to reuse them.

This is precisely where the result layer begins: transforming rules, evidence, and settlement capabilities that were previously siloed within individual platforms into infrastructure that can be procured across platforms. The next chapter will answer how these capabilities generate revenue once they begin to be repeatedly invoked.

05 | A monetization path for interpretation rights has emerged: the results layer is forming a second profit pool

When rule comparison, evidence verification, result confirmation, and payment triggering become reusable interfaces across platforms, their revenue no longer comes solely from final settlement, but also from repeated invocation of these four capabilities throughout the transaction chain. At this stage, interpretation rights are no longer just a governance power—they begin to generate a second profit pool.

But for the result layer to generate independent revenue, two conditions must be met: the same set of capabilities must be reusable across different markets rather than rebuilt for each event; and platforms, institutions, or developers must be willing to pay continuously for these calls.

HIP-4 first proved the first condition, confirming that it can not only be automated but also replicated across different types of markets.

5.1 HIP-4 has achieved product closure for asset flows, but trading volume remains unstable.

As previously established in HIP-4, deterministic outcomes such as prices can be machine-read and automatically settled. Its further business significance is that the same set of transaction and confirmation processes need not serve only one event.

When a new outcome market launches, the platform does not need to rebuild the entire system. Market creation, rule encoding, order distribution, liquidity integration, result confirmation, and fund settlement can all leverage existing processes:

Event is written as a rule → Rule generates resulting assets → Market continues trading → Machine confirms outcome → Funds are automatically paid → Template is reused by the next batch of events

This reuse capability has been validated in actual trading. Since its launch, HIP-4 has generated approximately $277 million in cumulative trading volume, with about $182 million traded in the past 30 days. This demonstrates that users are genuinely willing to trade assets confirmed and settled automatically by machines, rather than merely accepting such products in theory.

The trade structure also shows that the same infrastructure can serve different types of demand. In this batch of trades, the sports market contributed approximately 82%, but after the surge in trading activity from major events subsided, trading activity primarily returned to the crypto assets market. The sports market provides event-driven trading peaks, while crypto assets offer higher-frequency, more stable daily demand.

This enables the result layer to exhibit growth characteristics of software and data infrastructure: new markets no longer need to build settlement processes from scratch; the four established capabilities of the result layer can be replicated across events. The more trading categories and the higher the number of calls, the greater the marginal value of the standardized interface. However, whether call demand can be converted into independent revenue depends on who is willing to pay and under what pricing model.

5.2 Invocation, Workflow, and Data Authorization: Migration of Three Revenue Streams and Paying Parties

The results layer can be charged along three pathways: result invocation and payment triggering, institutional data and audit workflows, and result data authorization and distribution. Each pathway corresponds to different buyers, purchase motivations, and charging units.

The importance of the three types of revenue differs: institutional contracts can enhance revenue stability, data subscriptions and licensing can increase the proportion of recurring revenue and improve valuation quality, while growth in scale still primarily relies on result calls and settlement fees.

From the four suppliers already identified, payers are gradually expanding from on-chain protocols to regulated platforms and institutional budgets:

• Chainlink: There is a high-frequency demand for the proof result interface.

Polymarket, powered by Chainlink, has seen cumulative trading volume exceeding $3.4 billion in short-term crypto markets. This data demonstrates that when outcomes can be automatically confirmed by standardized data, the same interface can be repeatedly utilized across a large number of markets.

• Azuro: Prove that infrastructure calls can be converted into real revenue.

Azuro has generated approximately $4.5 million in cumulative protocol revenue from $530 million in cumulative trading volume. It provides not just the end result, but also infrastructure such as market creation, data provision, and settlement. This demonstrates that the backend capabilities of prediction markets can serve not only as internal platform costs, but can also generate independent revenue through protocol fees and revenue sharing.

• Pyth: Demonstrates that regulated platforms also purchase external data outputs.

Kalshi uses Pyth-provided data in certain gold, crude oil, and agricultural markets. Unlike on-chain protocols that charge per call, such collaborations resemble exchanges purchasing external market data and settlement inputs, with payment logic shifting from on-chain traffic to compliance, reliability, and responsibility allocation.

• ICE: Demonstrates that prediction market data can be integrated into institutional information products.

ICE integrates Polymarket’s probability and sentiment data into real-time data streams and historical databases for distribution to institutional clients. By this stage, what is being sold is no longer just the final outcome of a market, but the data itself continuously generated by prediction markets. The pricing model has accordingly shifted to data subscriptions, licensing, and revenue sharing.

These four cases demonstrate that the result layer can generate revenue in different ways: Chainlink proves demand for oracle calls, Azuro proves protocol fees, Pyth proves that regulated platforms are willing to purchase external inputs, and ICE proves that result data can enter institutional distribution systems.

Combined with HIP-4 from the previous section, a relatively complete commercialization chain for the result layer has emerged:

The product is traded → The API is called repeatedly → The platform acquires external capabilities → The result data enters the institutional workflow

Market structure also determines the importance of different revenue streams. In June 2026, Kalshi, Polymarket US, and Polymarket International combined generated $44.8 billion in volume, corresponding to a simple annualized run rate of $537.6 billion. Of this, regulated markets accounted for 77.1%, while on-chain markets accounted for 22.9%.

The two types of markets have different requirements for the outcome layer. On-chain markets typically delegate outcome confirmation to oracles, dispute resolution mechanisms, and smart contract payments, making them better suited for usage-based pricing such as calls, settlement counts, or protocol fees. Regulated platforms, however, retain final settlement responsibility but may still procure external data, auditing tools, and institutional distribution services, making them better suited for annual contracts, API subscriptions, and data licensing.

Since regulated markets already account for the majority of trading volume, future revenue growth for the result layer is more likely to come from compliant data, audit tools, and institutional workflows, rather than simply replicating on-chain dispute fee models. However, this assessment hinges on one critical precondition: these capabilities must be offered externally by independent providers. If exchanges choose to build and internally use these capabilities over the long term, they become internal platform costs and cannot be estimated as external result layer revenue based on total market trading volume.

5.3 From a $10 million foundation to a second profit pool worth hundreds of millions of dollars

As previously established, there is genuine demand at the outcome layer, and revenue models such as protocol splits, external data procurement, and institutional licensing have emerged. However, the industry currently does not separately disclose "outcome layer revenue," making it impossible to obtain a complete market size.

Therefore, this section employs a three-step estimation:

1. Estimate the current observable revenue layer based on actual protocol income generated;
2. Assuming the same fee-generating capacity covers the existing prediction markets, estimate the potential revenue expansion opportunity;
3. Refer to projects with proven business models to estimate the upper limit of revenue in a more mature state.

These three layers are not three independent predictions, but rather three stages along the same commercialization path. The further along you go, the higher the potential revenue—and the stronger the assumptions required.

Current observable commercial foundation: $15 million – $37 million

As of July 28, 2026, the 33 on-chain prediction market protocols listed on DefiLlama generated a combined revenue of approximately $146 million over the past year and about $30.26 million in the last 30 days.

However, not all of this revenue belongs to the outcome layer—it also includes revenue from the platform, liquidity, and other protocol layers. To estimate the portion that goes to the outcome layer, refer to Azuro’s actual revenue distribution rules: 10% of pool profits are allocated to Data Providers.

Calculate based on this ratio:

• The revenue for the past year corresponds to a result layer revenue of approximately $14.61 million;
• Annualized based on revenue over the past 30 days, equivalent to approximately $36.82 million.

After rounding, the current revenue layer supported by public income and actual distribution rules is approximately $15 million to $37 million.

This range does not represent the full revenue of the entire industry, but only the portion that can currently be confirmed from public data. It reflects commercial revenue that has already begun, not future projections.

Cover existing market: $64 million—$161 million

Based on the June trading volume mentioned above, the on-chain market corresponds to approximately $123.1 billion in annualized trading volume. Dividing the observed layer revenue by the on-chain trading volume yields the current observed layer monetization rate:

$15 million–$37 million ÷ $123.1 billion = 1.19–2.99 bp

The 1.19–2.99 bp here can be understood as approximately $1.19–$2.99 in revenue layer income generated per $10,000 in predicted market trading volume.

If this conversion rate applies not only to on-chain markets but also to the entire current prediction market, the industry's annual revenue would be approximately $64 million to $161 million.

This scenario does not assume that the prediction market continues to grow at a high rate, but only that existing trading volume increasingly uses external outcome layer services. However, it still depends on two conditions: outcome verification, data provision, and payment triggering must cover existing transaction volumes; and regulated platforms must consistently outsource data, auditing, and outcome workflows to external providers, establishing observable contracts, invoices, and renewals.

Mature productization ceiling: approximately $456 million

The second layer uses currently low industry monetization rates. To estimate revenue potential as the business model matures, refer to the proven fee structure already implemented by Azuro.

Azuro's cumulative protocol revenue accounts for approximately 84.9 basis points of its cumulative trading volume. Of this, approximately 10% is allocated to Data Providers, resulting in a monetization rate at the outcome layer of approximately:

84.9bp × 10% = 8.49bp

This means that for every $10,000 in trading volume, approximately $8.49 in result layer revenue is generated. This is not a theoretical rate, but a level already achieved by an actively operating project. If the entire prediction market reached this monetization rate, the resulting result layer industry’s annual revenue would be approximately $456 million.

It requires the results layer to cover more markets and achieve a higher revenue share across stages such as rule comparison, evidence validation, result confirmation, and payment triggering. Therefore, $456 million is better suited as a reference upper limit under a mature business model, rather than a near-term forecast.

10–18x is derived from data and financial infrastructure transactions by IHS Markit, Black Knight, and Adenza, used to measure industry revenue with recurring and high visibility.

After the full-year revenue at the result layer surpassed $100 million, the operational value corresponded to $1 billion–$1.8 billion, and the industry transitioned from early product validation to an institutional allocable stage.

Demand, fees, and distribution have already emerged but are scattered across different platforms and providers: HIP-4 generates calls, Azuro structures revenue sharing, and Pyth, Chainlink, and ICE dominate data and distribution interfaces. The industry value at the outcome layer is already measurable; next, the return on investment will depend on who can turn these calls into sustained revenue.

06 | How Do Industry Trends Map to Investable Assets?

We used the first five chapters to derive why the market needs a results layer, how it forms products, and the potential revenue it could generate. Next, it naturally follows to answer another question: even if the results layer has industry value, can investors capture this portion of the returns through existing assets?

6.1 The prediction platform's equity includes capabilities of the result layer, but it is not the most direct investment tool.

The most straightforward investment approach is to hold equity in a prediction market platform. However, the valuation of a prediction platform encompasses more than just the outcome layer—it also includes trading volume, licenses, liquidity, user channels, brand strength, and expectations for growth in the prediction market.

Two primary valuation anchors are currently observable: Kalshi completed a new funding round in May 2026 at a $22 billion valuation. Polymarket had an approximately $8 billion pre-money valuation when it received investment from ICE in October 2025, and ICE subsequently made a $600 million direct investment in March 2026.

This means that while platform equity can provide exposure to predicted market growth, it is not the purest investment instrument at the outcome layer. At current valuations, investors paying for platform equity are primarily paying for transaction growth, licenses, and network effects—not a standalone, independently validated outcome-layer asset.

Therefore, our judgment is clear: we do not recommend chasing higher platform equity at this time.

6.2 Investment Threshold: Cross-Platform Calls and Attributable Revenue

Four industry positions around the outcome layer have emerged. The weights of these four positions differ, ranked from highest to lowest based on the directness of revenue attribution:

• Deterministic data and machine verification — Pyth and Chainlink;
• Subjective outcomes and dispute resolution — UMA, Kleros, and other dispute arbitration mechanisms;
• App distribution and workflow entry points—DFlow and other distribution interfaces, Agent interfaces, and professional rule tools, referred to in this report as Builder apps;
• Institutional data standardization and distribution — ICE, see separately in 6.4.

The first two directly attribute revenue to the ownership layer, the third is realized through growth, and the fourth has already been incorporated into a publicly traded company's financial statements.

In the primary market, prioritize researching two types of projects: those that can be cross-platform invoked without relying on a single exchange, and those that can convert invocation into ARR, traffic fees, or data revenue sharing. The final step in asset formation is revenue attribution. The easiest place to test this criterion is HIP-4, which already has trading volume, fees, and tokens.

6.3 HYPE: You can buy platform growth, but HIP-4 has not yet generated significant fee contributions

HYPE is one of the easiest related assets to trade currently, but it represents the overall growth of the Hyperliquid platform rather than pure revenue layer income. To determine whether HIP-4 has impacted HYPE, the key is not the trading volume itself, but the proportion of fees generated by it relative to Hyperliquid’s total fees.

HIP-4 Fee Contribution = HIP-4 Fees ÷ Hyperliquid Total Fees

Based on a 30-day trading volume of $182 million and a fee rate of 4–7 bps, the monthly fees for HIP-4 are estimated at $72,800–$127,400. During the same period, Hyperliquid’s total fees were approximately $57.5 million, meaning HIP-4’s fees account for only 0.13%–0.22%.

The current share percentage indicates that HIP-4 has validated product demand, but it is not yet sufficient to constitute an independent valuation increment for HYPE. When the fee share is below 1%, it functions more like a product option; only when it consistently exceeds 1% does it begin to emerge as a visible business line; and only when it surpasses 5% might it substantially impact HYPE’s valuation.

The significance of HIP-4 is that this fee pathway already exists. As the results market grows and trading shifts from event peaks to more stable daily demand, its revenue may gradually flow into protocol fees.

6.4 ICE: Financial transparency, but exposure at the result layer is too small

ICE is currently the most verifiable listed asset related to prediction markets. It holds equity in Polymarket and has integrated Polymarket data into institutional market and historical databases, allowing both the equity value and data distribution value from prediction markets to potentially appear in public financial statements.

As of March 31, 2026, the carrying value of the Polymarket Series D/E preferred shares held by ICE was approximately $2 billion, representing approximately 23% of the issued shares and approximately 14% of the fully diluted ownership. In the first quarter of 2026, ICE recognized a non-cash fair value gain of approximately $389 million due to observable price changes.

ICE's advantage lies in its clear validation, but this exposure remains too small relative to ICE's overall business scale to significantly impact the company's profitability and valuation. Therefore, ICE is better suited as an observational window for whether commercialization outcomes enter institutional budgets and public financial statements, rather than as a highly elastic investment target.

6.5 Early-stage projects and infrastructure protocols: closest to the result layer, but revenue attribution is still unclear

Compared to HYPE and ICE, oracles, dispute arbitration, distribution interfaces, and professional Builders are closer to the outcome layer itself. Pyth and Chainlink provide deterministic data and machine-verified confirmations, UMA and Kleros handle subjective outcomes and disputes, and projects like DFlow connect platforms, agents, and application workflows.

The advantage of such projects is that they have more focused business models and are more likely to directly handle cross-platform calls. However, most projects still face the same issue: product usage does not equate to stable revenue flowing to equity or token holders. The number of partner platforms, supported markets, and call volumes can demonstrate demand, but alone they cannot prove investment value.

Therefore, to determine whether such assets are viable, two metrics must continue to be monitored: first, whether they can be reused across multiple markets beyond a single platform; and second, whether these usages can be converted into ARR, traffic fees, or data royalties, and clearly attributed to the corresponding assets.

The most promising opportunities near the results layer are still concentrated in these early projects, as a complete revenue闭环 has yet to emerge. Whoever first transforms cross-platform interactions into attributable, recurring revenue will likely become one of the first true results-layer assets.

6.6 Current Conclusion: Opportunities are forming, but the ideal asset has not yet emerged

The industrial logic at the outcome layer has gradually taken shape, but investable assets remain in early stages. Platform equity valuations are too high, HYPE has not yet generated significant fee contributions, and ICE’s related exposure is insufficient to deliver high elasticity; early projects truly close to the outcome layer are still awaiting validation of cross-platform invocation and revenue attribution.

This means that the most important thing right now is not to force-select a single "result-layer asset," but to track who first satisfies two conditions: first, providing rules, evidence, and result-verification capabilities to multiple platforms; and second, converting these calls into sustained revenue attributable to equity or tokens.

07 | Conclusion: The earliest alpha often appears before the asset is fully formed

The rapid growth of prediction markets has historically occurred at the frontend, with more platforms entering, more categories launching, and traffic, licenses, and liquidity becoming the core competitive factors.

However, as the market scale expands, the costs associated with rule interpretation, evidence verification, and result confirmation are rising in tandem. Each platform independently building its own processes leads to redundant investments and the creation of incompatible standards. These capabilities, previously confined within individual platforms, are now becoming suitable for standardization and shared use across multiple markets.

Different platforms need to read the same facts, process similar evidence, interpret comparable rules, and bear the same risk of incorrect settlement. Today, these capabilities remain fragmented among oracles, dispute mechanisms, data providers, and platform operations teams. Once standardized, they could become infrastructure collectively procured by multiple markets, much like pricing, clearing, and payments.

This is where the result layer truly matters. Instead of competing for every trader, it serves existing trading demand as the number of markets, trading pairs, and settlement transactions grows. Front-end competition determines where traffic flows, but the result layer becomes a layer that traffic must invoke regardless of which platform it passes through.

The market has not yet established a clear leader, and revenues have not stabilized around any specific class of assets. However, the most valuable stage of industry opportunity often comes not after the landscape has been settled, but when demand begins to emerge across platforms, product boundaries become clearer, and revenue allocation has yet to be fully priced in.

Therefore, what truly needs to be tracked next is who begins serving multiple platforms simultaneously, who turns rules, evidence, and result validation into standardized products, and who first transforms these calls into sustained revenue.

The first phase of prediction markets involves the expansion of access points. In the next phase, value will increasingly flow to infrastructure that does not directly own users but determines how the market completes final payments.

Heretic Research will continue to track the evolution of rule-setting, evidence verification, and outcome validation from within platforms to industry-wide infrastructure, with a focus on cross-platform adoption, fee structures, and revenue allocation. This research will be updated as a clearer industry landscape and investable opportunities emerge.