As Polymarket and Others Step into the 'Background': The Second Half of Prediction Markets Beyond Exchanges?

Probability is no longer exclusive to a single platform; the aggregation, derivatives, and smart execution infrastructure surrounding it are taking shape.

Written by: Farmer Frank

Prediction markets are undergoing an interesting change lately.

From Binance and Coinbase to Interactive Brokers (IBKR) and Robinhood, leading players are shifting their focus, no longer fixated on replicating a Polymarket.

Everyone is looking for ways to gradually push Polymarket and Kalshi into the 'background', transforming them into underlying capabilities that other financial products can directly utilize.

This is quite similar to today's stock trading, where users buy TSLA through Futu, Tiger, or Robinhood, and most users do not care which Market Maker their orders ultimately flow to or which clearing system they go through.

The future of prediction markets may be similar.

The front end could be a brokerage, wallet, news app, or even an AI agent, with aggregators and routers in between, while the actual providers of markets, liquidity, and settlement—Polymarket and Kalshi—are increasingly resembling hidden financial infrastructure.

Looking back at DeFi, from AMM to aggregators, derivatives, professional market-making, and smart execution, this path has already been traversed.

Prediction markets may also be entering a similar second half.

1. Prediction Markets Are Starting to Move More Towards the 'Background'

In recent years, prediction markets have proven that the 'uncertainty' in the real world can indeed become a tradable asset with a price and liquidity.

From the U.S. presidential election to the World Cup and other sports events, as well as crypto, macroeconomics, and entertainment culture, many issues that could only be discussed have been compressed into Event Contracts that can be bought and sold by platforms like Polymarket and Kalshi.

This is a crucial step, and as this layer of demand is gradually validated, the strategies of leading players have begun to change significantly.

In April this year, Binance integrated Predict.fun, providing market capabilities through third-party prediction market infrastructure while offering a front-end entry for users to trade probability events directly in the app; two months later, it further opened the Prediction Markets API, allowing quantitative strategies, trading bots, and third-party products to directly access market data and trading capabilities.

Coinbase is following a similar route, incorporating Prediction Markets into its 'Everything Exchange', but the initial market liquidity all comes from Kalshi, and it has explicitly stated that it will support more Prediction Market venues in the future.

On the traditional finance side, IBKR has taken it a step further.

In May this year, IBKR directly integrated three Prediction Markets—Kalshi, CME Group, and ForecastEx—into a unified interface, allowing users to search for events, compare prices and liquidity across different exchanges, and complete transactions without needing to open separate accounts.

Robinhood has also begun to extend into trading and clearing, collaborating with Susquehanna to operate Rothera, which has taken over the CFTC-registered trading and clearing infrastructure of the former MIAXdx/LedgerX, and has started routing some World Cup and professional baseball Event Contracts to this affiliated exchange since June.

The actions of these leading players point to the same thing: prediction markets are transitioning from a 'destination' that users need to actively visit into a financial capability that can be called upon by other products.

In the past, if we wanted to trade the upcoming U.S. midterm elections, we might have needed to first open Polymarket and then find the corresponding market, but in the future, it could be an event card appearing on the Binance homepage or a real-time probability bouncing next to a financial news item on Robinhood.

Users may not even realize they are trading in prediction markets, and whether the orders ultimately come from Predict.fun, Kalshi, CME, or are routed to multiple markets is less important.

This also means that Prediction Markets are becoming increasingly invisible, while Prediction Assets are becoming more significant.

Once this step is reached, the truly interesting question for the industry changes—how should the liquidity scattered across Polymarket, Kalshi, Predict.fun, and even more markets be organized?

For instance, regarding the U.S. midterm elections, where are the best prices? Where is the deepest liquidity? Is there a probability divergence between two markets?

These questions are very similar to those in early DeFi. After Uniswap proved that tokens could be traded on-chain, the market did not stop at 'recreating ten Uniswaps'; what truly emerged were liquidity aggregators like 1inch, smart execution networks like CoW Swap, derivatives infrastructures like Hyperliquid, and the market makers and quantitative trading systems that grew around them.

Prediction markets are also gradually reaching a similar stage.

Fortune is a typical early case; as of August this year, it has completed multiple rounds of financing, including Seed and Pre-A, raising over $4 million for upgrading Fortune Agent, integrating more prediction markets, and expanding liquidity and related infrastructure.

What it aims to tap into is 'The Liquidity Infrastructure for Prediction Markets'; not long ago, it further integrated liquidity from Polymarket, along with the previously supported Predict.fun, into a unified entry for Fortune Markets.

Users can now view event markets from different liquidity sources in one entry, comparing trading volumes, liquidity, and implied probabilities without having to switch back and forth between different Prediction Markets.

This is just an early form, but it already reveals what Fortune aims to do—gradually abstract the originally fragmented Prediction Markets into a unified liquidity layer.

2. Not Just 'The Prediction Market Version of 1inch', What Else Can Be Done?

Of course, if we simply understand new players like Fortune as 'the Prediction Market version of 1inch', it would be somewhat simplistic.

Prediction markets may not fully replicate the development path of DeFi; new players like Fortune are primarily trying to shift their perspective from Prediction Market to Prediction Asset.

It may seem like just a word apart, but the underlying logic is vastly different: Market focuses on 'where to trade', while Asset concerns 'what is actually being traded and what can be built around it'.

Just like BTC does not belong to Binance, Tesla is not only tradable on Robinhood, U.S. stocks are not limited to spot trading, and crypto is not confined to spot trading; once Prediction Assets can be sought, priced, combined, and traded across different markets, the infrastructure above the market truly has a chance to form.

Taking Fortune as an example, the system that new players want to build can be roughly divided into several interconnected layers.

1. Gather Dispersed Prediction Assets Together

As we all know, today's prediction markets are still highly fragmented.

The same type of event may exist simultaneously on Polymarket, Predict.fun, and other platforms, with different platforms having their own independent market structures, liquidity, and pricing.

As early as April this year, Fortune's native Prediction Market was launched, and by August, it further integrated Polymarket CLOB v2, along with the previously supported Predict.fun, into a unified entry for Fortune Markets.

Users can now directly browse event markets from different liquidity sources in Fortune Markets, comparing their liquidity, trading volumes, and implied probabilities before establishing positions.

The new trading process has also added features like Outcome Selection, Position Preview, and Portfolio, gradually linking market discovery, position establishment, and subsequent management together.

In the past, if a user wanted to trade the same type of political, sports, or crypto event, they might have needed to enter several platforms separately to search and then compare prices and depths themselves; now, Fortune Markets compresses this process into 'discovering events → comparing different platforms → choosing prices and liquidity → establishing positions → unified management.'

In theory, the aggregation of prediction markets is far more complex than a typical DEX Aggregator.

1 ETH remains the same ETH on Uniswap and Curve, but two seemingly identical prediction markets can become completely different assets due to slight differences in deadlines, event definitions, judgment criteria, or settlement rules.

From this perspective, what Fortune is primarily doing is not creating more markets, but gradually transforming Prediction Assets that are originally scattered across different platforms into a more searchable, comparable, and tradable asset pool.

2. Moving from "Buy YES / NO" to More Complete Financial Products

Once the assets are connected, the next question naturally changes.

The most common trading method in today's prediction markets is still to bet on a favorable outcome by buying YES; if not favorable, buy NO, and then wait for the event to settle.

This is very similar to the early days of Crypto when there was only Spot trading.

However, if Prediction Assets eventually develop into a sufficiently large asset class, trading demand theoretically will not remain solely in binary betting. When the underlying asset size is large enough, markets typically continue to grow leverage, options, combinations, hedges, and structured products.

This is also why Fortune has incorporated Prediction Derivatives into its overall product direction; it aims to transform Event Contracts from a binary contract that "waits for the final answer" into Prediction Assets that can be combined, managed, and strategically used.

This step is actually crucial.

Because the true maturity of an asset class is often not measured by how lively the spot market is, but by whether it can form a sufficiently rich financial structure around it.

BTC gradually formed today's complete trading system after moving from spot to Perpetual, Options, and structured products; the stock market also has futures, options, ETFs, and various combination tools.

What Fortune is betting on now is that Prediction Assets will also undergo a similar financialization process.

However, this also raises another question: If in the future a user faces not just a few markets, but hundreds or thousands of Prediction Assets, or even different combinations and strategies, will people still have the ability to complete all research and execution themselves?

This is where the Fortune AI Agent should truly come into play.

3. Directly Integrating AI into the Trading Chain

Realistically speaking, prediction markets may be one of the most easily understood financial scenarios for AI Agents.

Because there is a very clear transmission chain: changes in the real world → new information emerges → event probabilities change → markets reprice → trading opportunities arise.

However, traditional trading requires manually completing the entire process, from reading news, scrolling through social media, analyzing market sentiment, to judging whether this news will change the probability of a certain event, and finally finding the corresponding market, comparing prices, determining positions, and executing trades.

What Fortune Agent aims to compress is precisely this chain.

The currently launched Trading Agent is designed as a Multi-Agent System that continuously seeks opportunities from three types of signals: News, Sentiment, and Arbitrage, and further verifies conditions. After users connect their wallets, they can set the amount for each trade, risk level, and whether to enable Automated Trading.

If we delve deeper into Fortune's subsequent descriptions of the Agent, including 24/7 Market Intelligence, Structured Decision-making, Risk Management, and Execution, then when combined with the earlier Fortune Markets, the entire product logic is completely connected.

For example, if a macro event suddenly presents new policy signals.

The Agent first captures the information and assesses whether it might change the true probability of a certain event; subsequently, Fortune Markets can simultaneously provide relevant Prediction Assets, prices, and liquidity from different venues; if market quotes have not fully reflected the new information, the Agent further seeks more suitable trading opportunities and execution paths.

At this point, what Fortune aims to do is no longer just an "AI predictor"; it is closer to connecting the information layer, asset layer, liquidity layer, and execution layer together.

And on the outermost layer, there is an Incentive Layer responsible for cold-starting this network. Fortune has currently established an incentive system centered around F Points, including Daily Check-in and invitation mechanisms; users can earn 10% of the F earnings of other participants they invite.

These may seem like common Points, NFTs, and Referral plays in traditional Web3 projects, but within the entire product architecture, they actually address a very real problem:

Where does the early liquidity, trading users, and ecological participants of a new Prediction Asset network come from?

More user participation brings more trading and liquidity; deeper liquidity improves transaction experience, further attracting new users and strategies; as Agents and more financial products join, trading frequency and strategy complexity may also increase.

Thus, when we string Fortune together from start to finish, what it is truly trying to build is not an isolated function, but a relatively complete chain:

• Prediction Markets provide underlying event assets;
• Fortune Markets connect assets with liquidity;
• Prediction Derivatives expand the financial expression of assets;
• Fortune Agent handles information processing and trade execution;
• Incentive Layer provides early growth momentum for the entire network;

Therefore, if we must give Fortune a positioning, it is neither just a Prediction Market nor just a "1inch for prediction markets."

More accurately, it aims to be a trading and execution infrastructure centered around Prediction Assets.

3. When "Probability" Truly Becomes an Asset

Of course, whether Fortune can truly realize this roadmap is still difficult to conclude at this point.

It remains an early-stage project.

Polymarket's liquidity access and Agent have already shown actual products, but unified order routing, mature Prediction Derivatives, and sufficiently deep cross-market liquidity networks have yet to reach their final form.

However, if we pull back the perspective to the entire industry, the direction Fortune is betting on is not isolated. Apex has already begun integrating Kalshi's Event Contracts into brokerage infrastructure via API; Paradigm is also developing a Prediction Market Terminal for professional traders and researching internal market-making and Prediction Market Index.

Prediction markets are increasingly resembling a real financial market, and a true financial market will certainly not consist solely of exchanges.

At least three layers of changes are worth observing next.

1. From Bet to Portfolio, Highly Financialized

Today, many people first encounter Prediction Markets by understanding them as "I bet whether something will happen."

But for more mature traders, they can actually create a series of asset combinations that express complete viewpoints.

For example, when a trader judges that U.S. inflation is rising again and the Fed is turning hawkish, they may not only trade on "whether the next FOMC will raise interest rates" but can simultaneously establish positions around different events like "Will the Fed maintain higher rates?", "Will BTC break a certain price by the end of the year?", and "Will the U.S. avoid recession?".

These individually are Event Contracts, but when combined, they can express a complete macro Thesis of Higher for Longer.

If this stage truly emerges, Portfolio Management, Correlation, Hedging, and Risk Management will naturally follow, and at that time, if projects like Fortune can truly complement the Derivatives and combination layers, their value will no longer just be helping users "open fewer web pages."

2. From Single Market Trading to Cross-Market Aggregated Execution

This is easy to understand; as the market grows, the importance of a single platform may actually decrease.

Crypto ultimately did not form a pattern of "one exchange carrying all liquidity"; Prediction Markets are unlikely to do so either.

Different regulatory systems, user groups, market makers, event categories, and regions will create long-term market segmentation. Market segmentation itself is an opportunity for infrastructure, as arbitrageurs need prices, market makers need order flow, institutions need depth, and ordinary users need the best execution prices. Agents need enough venues to scan and execute.

Therefore, once Prediction Markets truly mature, trading access may become increasingly "invisible." For example, in a brokerage app, one might see a news item: "Fed adjusts interest rate probability next month: 72%"

Next to it, a button allows for immediate purchase. It doesn’t matter whether the underlying orders come from Polymarket, Kalshi, or are split across three markets; what matters is the execution.

This is the real change that API integration can bring.

3. From Human Traders to AI Traders

I firmly believe that prediction markets may become one of the most natural financial application scenarios for AI Agents.

This is because it inherently follows the chain of "information → probability → price → transaction," which is precisely where AI excels.

If Crypto provides AI Agents with a financial system that allows for direct control of assets 24/7 without a bank account, then Prediction Markets further provide a market where "cognition" can be traded directly.

One of AI's core capabilities is processing information, and the core asset of Prediction Markets is the probability formed after information is compressed.

The two are a natural fit.

We can imagine a future where an Agent listens to news, social media, macro data, on-chain data, company announcements, sports events, policy documents, etc., and continuously recalculates its probability model.

Once market prices deviate significantly from the model, it will place orders directly, and at that point, the speed of Prediction Markets will also change.

In the past, Alpha might have come from seeing a news item earlier than others; in the future, it may become: my Agent understands what this news means faster than your Agent; and even further down the line, who can convert cognitive advantages into transactions faster across more venues.

Thus, informational advantages, model advantages, and execution advantages will gradually become one and the same.

This may be the truly interesting aspect of AI × Prediction Markets. However, before all these imaginations can materialize, there are several real-world issues that cannot be overlooked.

• First, liquidity: without sufficiently deep order books, even the most advanced routers and derivatives are meaningless;
• Second, resolution: prediction markets ultimately need a credible, clear, and as uncontroversial as possible event settlement mechanism;
• Next is regulation: whether an Event Contract belongs to derivatives, gambling, or a new financial instrument varies significantly across regions;
• Lastly, there’s the Agent itself: AI can process vast amounts of information, but there remains a considerable gap between "being able to summarize news" and "being able to consistently generate Alpha."

These issues will not automatically disappear just because the market is growing.

Only when these infrastructures are gradually completed can Prediction Assets truly evolve from a novel trading category into a mature asset class.

In conclusion

Objectively speaking, from the 2024 U.S. presidential election to the 2026 World Cup, Polymarket and Kalshi have already completed the most challenging first round of user education for prediction markets:

Many uncertainties in the real world can indeed be traded.

But this feels more like the first half. If we look back at the history of almost all financial markets, we find that proving an asset can be traded has never been the end of the story.

As more participants, assets, and platforms emerge, the factors determining market maturity are often those less glamorous aspects outside the exchanges, such as liquidity, market making, routing, derivatives, risk management, portfolio strategies, and more free and automated execution.

DeFi has already gone through this, and prediction markets are likely following a similar path.

Thus, when looking at new players like Fortune today, what truly deserves attention is how we can trade these increasingly probabilistic assets more efficiently after they become real assets.

It requires deeper liquidity, richer financial instruments, more efficient execution, and AI Agents capable of understanding the real world 24/7 and continuously recalculating probabilities.

And this may very well be the true second half of Prediction Markets transitioning from "markets" to Prediction Assets.