👋 Welcome to the Predicted — the newsletter covering the business of prediction markets.

Not subscribed yet? You know what to do 👇

Discussed in this piece:

1) Why did Zuckerberg order a prediction market app, code-named Arena

2) The state of the category by the numbers

3) The mess underneath the charts, from a widening regulatory war to Polymarket's fake-winnings scandal

4) Why prediction markets do not work as a business line for Meta, and what the industry gets wrong about the customer

🖋 This piece was written and researched by Pet Berisha and Omar El Safy

Mark Zuckerberg has ordered Meta to build a prediction market.

The New York Times reported on 23 June 2026 that a small team within the company is building a standalone app, code-named Arena, where users predict the outcomes of real-world events such as sports results, elections, and stock moves. People close to the project describe it as a top priority for Zuckerberg himself.

Arena will run on play money. Users get a daily allotment of virtual currency to place their predictions and earn points rather than cash. No real stakes, at least to begin with, although they have not ruled out real money later.

This is not Meta's first attempt at a forecasting game. They ran a prediction app called Forecast from 2020 until they closed shop in 2022.

Should Meta be building a prediction market?

If they want to make it real money, then I think it’s a bad idea.

I have spent the last few weeks talking to founders, investors and operators across prediction markets, and I keep landing in the same place. Gamified and free, this is a fine engagement toy. Paid, as an actual business line, for many businesses, does not work.

To set the scene, let’s look at some of the recent data.

It’s nothing but extraordinary.

Kalshi are in talks to raise at a $40 billion valuation, according to the FT, roughly double the $22 billion mark they raised at in May 2026.

That would value a single event-contract exchange above most fintechs. It comes after Kalshi recorded $17.91 billion in notional volume in May 2026, their ninth consecutive monthly record, up 21% on April.

Polymarket are reportedly raising at around $15 billion. Their volume went the other way, falling to roughly $7.1 billion in May 2026. Across the whole category, May cleared a record $31.2 billion, split about 58% to Kalshi and 28% to Polymarket.

A lot of the recent run-up is World Cup froth. Robinhood-owned Rothera have seen an explosion during the World Cup too, with 97% of event contracts in June 2026 attributed to the most-watched global sporting tournament.

It’s no shock that these numbers have piqued Zuck’s interest.

Behind the numbers, the regulatory situation in the U.S. is beginning to creak, becoming fragmented and complex.

The SEC have now joined the CFTC in scrutinising prediction markets. Concurrently, the state-by-state war is ongoing, the latest of which sees the CFTC suing Kentucky, the ninth state pulled into the jurisdiction battle. Kalshi are in court with Illinois over a new tax. And a House panel have advanced a partial ban.

Every operator is now litigating its way into each new market it wants to serve. That legal overhead is a near-permanent cost at this point that these valuations gloss over, something that a consumer giant like Meta offering event contracts walks straight into, if the product has real money. Of course, this is something Meta are used to, in some respects. They recently, alongside YouTube, lost a landmark ‘addiction trial’ in March 2026.

[But do you know what’s more addictive than social media? Gambl…. okay fine, not going there in this piece, otherwise it would be 5000 words, and who wants that.]

To add fuel to the fire, the headlines driving additional regulatory scrutiny have recently gotten uglier.

The Wall Street Journal reported in June that Polymarket ran a covert influencer campaign built on fake winnings.

The Journal reviewed 1,105 videos from ten creators tied to a Polymarket marketing contractor. Around 70% showed a bet, totalling roughly $1.9 million.

The catch? None of them were real.

118 of the videos showed creators "winning" almost $900,000. Had the bets been genuine, those same creators would have lost more than $166,000.

The fakes were staged on dummy sites built to look like Polymarket, designed to push viewers onto the real, offshore platform. The creators were mostly college-age, paid two to three thousand dollars a month, and told not to disclose the arrangement. One student posted a video of "winning" $100,000 on a bet that Trump would say the word "McDonald's" that month.

On 26 June 2026, the National Association of Consumer Advocates sued Polymarket over the campaign, alleging they deliberately targeted college students, a group with problem-gambling rates around twice that of adults. Polymarket say they are auditing their promotional content and have launched an internal investigation.

Meta have 2 billion users. And addictive game loops like Farmville were key to Facebook’s success in the late 2000s. But my hunch is that Meta would not stop at gamified loops. If the numbers looked good ‘on paper’, why would they not turn the paid funnel on?

Crucially, they have not ruled out a real-money prediction market.

A paid product, however, might not be a sure path to revenue.

The logic is that raw distribution is the moat in this market, and Meta starts with the billions of users that Kalshi and Polymarket have spent years and fortunes chasing. On that logic, Meta would win by default, and the only real obstacle is then regulatory.

I think that logic is wrong.

The thing that decides this market is the type of customer that is being acquired, and Meta have the wrong one.

Distribution does not = volume.

And volume is exactly what prediction markets require to drive revenues.

Most founders I have spoken to building in this market could not tell you who their actual customer is, and that includes incumbents with the biggest reach.

Breaking this down further: a prediction market is a high-volume, low-margin business. It makes a thin slice on an enormous churn of money.

Compare that to how a platform like Coinbase makes money.

Coinbase a platform that aims to generate wealth, with high margin and lower volume, for retail customers buying crypto. Of course, there are also market makers and professional traders that use their pro terminals and APIs, but most of their customers are retail.

They can sell you Bitcoin, and take a big margin on it. And consumer transactions are 40.1% of Coinbase’s revenue, per their latest SEC filings.

Prediction markets cannot replicate the big margin products that their adopters are making their money from.

A prediction market earns a thin take on each trade, and far less than that on the most liquid contracts. To build a real business on a thin take, you need volume that recurs, which means you need the same people coming back to trade again and again.

The customers who recur are the ones who are good at it, the market makers and the sharps, and they are extracting from everyone else. The customers who actually fund the thing, ordinary retail, are by definition the ones losing, and people who lose money do not come back forever. A wealth-building product aims to keep customers for a long period of time. A wealth-extraction product like a prediction market churns customers regularly.

That difference is why the distribution keeps failing to convert.

Coinbase have enormous reach and launched prediction markets, and it has not worked for them, because you cannot take a customer who came to you to build wealth and move them into a product that is designed to extract it. Gemini, another crypto exchange, have seen similar challenges.

As I see it, here are the many different types of customers that prediction markets cater for, and these platforms do not understand the segmentation properly.

There might be more, but this is what I’m currently seeing: 

Building horizontally for all of these audiences is very difficult. 

And each audience needs something different.

Fundamentally, this is a product, technical and business model problem depending on who the customer is.

Let’s discuss some of the problems we see in existing platforms, for example.

Kalshi began offering parlays to their customers about 9 months ago. 

Or, as Kalshi call them, ‘Combos’. 

In a traditional sportsbook, parlays are where a user picks multiple lines and wraps them into one wager at higher odds. Like a lottery ticket for sports gamblers.

It’s the most popular bet type for retail bettors and the highest-margin, most profitable product that a sportsbook offers.

But this product was built for a centralised sportsbook. 

I can pick 10 legs and know IF I win, it is paid out. 

But how does that work with event swaps?

One would think that Kalshi would take the liquidity from their book and also ask market makers to make markets for these combos. 

Actually, they are shifting this liability to market makers - meaning that they ‘make the markets’ for these parlays. 

That is fundamentally a worse product offering than a centralised sportsbook. 

Kalshi, who are an ‘everything prediction market’, still haven’t been able to crack this, and it’s mostly because of the business model. 

As per Legal Sports Report: 

“Parlays on Kalshi have grown from a negligible share of trading activity in the fall of 2025 to more than 20% of weekly sports volume during the opening week of March Madness, marking a new high since the product launched, according to data compiled by Legal Sports Report.”

On the surface, this seems pretty good.

Parlays have gone from 0% to 20% of total sports volume on Kalshi in ~6 months.

But, again, as per Legal Sports Report:

“Kalshi generates most of its revenue through transaction fees rather than hold, so parlay activity does not inherently increase margins the way it does for sportsbooks. In practice, Kalshi earns roughly the same fee — about 1% — on a trade, whether it’s a parlay or a straight bet, regardless of outcome. Sportsbooks, by contrast, might hold around 5% on a straight bet and 20% or more on a parlay.”

Essentially, the feature that Kalshi have built is servicing sports bettors, but not necessarily their own business model properly. Especially when you have to cut in market makers and heavily incentivise them to make the markets for parlays. 

Nevertheless, volume is volume, and it’s something Kalshi desperately need to keep growing to grow into their huge valuation. 

But that’s just one feature set of a traditional sportsbook that a prediction market is trying to replicate, and struggling to find the right business model for. The customer acquisition cost here is HIGH, and if sportsbooks are making 45-65% of their revenues from parlays. The amount of money required to drive customers away from them is huge. 

This is just one example of the minutiae that are required for a $40bn valuation prediction market business to be competitive with existing products. And still, it’s an unideal way for them to compete, because the core primitives of event swaps are not conducive to building a very profitable parlay product. 

Let alone talking about building for day traders, market makers with deep pockets, and having the infrastructure to support the likes of JPMorgan and Goldman Sachs, who have both indicated prediction markets are of ‘interest’ to them. 

But even in the usual swap markets, some big market makers are getting savaged.

0xDevin on Twitter explains this below:

TL/DR → Market makers on Kalshi employ something called ‘shading’. This is where market makers see retail flow going into one side of a contract, and try to make the price ‘bad’, removing any edge. But the market makers are getting gazumped by sharp pro sports traders and bettors.

Devin’s research showed the following:

“The description circulating among prediction market traders is blunt: market makers shade the retail-heavy moneyline side to horrible odds, moving the entire market, then get picked off for millions of contracts on the other side, then shade the other side to balance inventory. They also leave millions of contracts exposed in the order book when lineups drop.”

Betfair exchange solved this by giving market makers a 5-9 second delay, and taxing their top 1% of traders fairly heavily. Even Polymarket have a 3-second delay for market makers. Kalshi have no guardrails at all.

This brings us on to Robinhood.

Robinhood, like Coinbase, are a digitally native broker. One started in stocks, the other in crypto. But now, the lines are blurred. The overlap in their products is great.

But Robinhood are seeing great success in their prediction market compared to Coinbase.

Why?

Because Robinhood have captured the hearts and minds of retail traders and investors. And a subsection of them has a gambling-adjacent mindset. The GameStop saga is a good example of that. The venue with the right customer beats the venue with the most customers, every single time.

Their exchange, Rothera, a joint venture with SIG (the biggest market maker on Kalshi), is now firing on all cylinders, with every metric growing exponentially.

And no doubt, Rothera, which is part-owned by SIG, will protect the market maker with the guardrails it needs.

To summarise:

Presumably, because they think 2bn+ customers would convert to substantial volumes. Because honestly, I don’t buy that Zuck would keep this ‘free’ if it showed even a semblance of traction.

But again… Meta does not have the right customer. Facebook and Instagram are where people scroll, message and post. They are not there to take financial positions or bet, even they are ported to a sister brand.

Let’s run a thought experiment.

Picture the funnel. You are on Instagram, you see a market, you are nudged across to a real-money venue. You deposit $500. You lose $200. You might never come back.

That is the base case. Retail loses, and retail that loses does not stay. The volume charts you see from the big venues do not capture churn. And retail churn is the hardest thing to solve for in this industry.

Meta's version of that funnel, should they turn real money taps on, is potentially problematic from a business perspective, too. People who came to scroll or speak with friends and get nudged into a bet or a trade. Casual money is exactly the money that is deposited once, lost, and disappears.

We have a quarter of a century of evidence for this, and it is sitting in the UK. Betfair Exchange has been running a peer-to-peer prediction market since 2000. It is a genuinely good product with tighter pricing than the bookmakers around it. Twenty-five years on, it is still a fraction of the UK betting market it was supposed to replace, because the same thing happened.

Market makers and quants set up shop, retail kept losing to them, and retail churned out, going back to traditional sportsbooks. Saving 1% on the spread did not keep ordinary punters in the seat.

This is the real unsolved problem. How do you stop retail churning out?

Until someone answers that, prediction markets do not become mass-market consumer businesses, especially not as integrated or standalone products for incumbents.

A last note, which may not have dawned on people.

Meta makes money from advertising, predominantly. Direct advertising via their platforms, using campaigns, or indirect (paid promotions from content creators) drive big money to Meta.

If Meta launch a real money prediction market, are they going to give up the advertising revenue that is being pushed in this direction?

Is that potential loss in advertiser revenue worth the ‘hope’ that prediction markets drive the necessary volumes that move the needle for Meta?

Zuck is one of the all time great CEOs when it comes to moving his stock price up. But even he has had huge failures, most famously with the 10s of billions lost with the metaverse.

But even a company as big as Meta need to show the market they are looking for growth.

For Zuck, it seems as though AI and prediction markets are the verticals Meta will pursue.

This newsletter is for informational purposes only and is not financial, business or legal advice. These are the author's thoughts & opinions and do not represent the opinions of any other person, business, entity or sponsor. Any companies, platforms, markets or projects mentioned are for illustrative purposes unless specified.

The contents of this newsletter should not be used in any public or private domain without the express permission of the author.

The contents of this newsletter should not be used for any commercial activity, for example - research report, consultancy activity, or paywalled article without the express permission of the author.

Please note, the services and products advertised by our sponsors (by use of terminology such as but not limited to; supported by, sponsored by or brought to you by) in this newsletter carry inherent risks and should not be regarded as completely safe or risk-free. Third-party entities provide these services and products, and we do not control, endorse, or guarantee the accuracy, efficacy, or safety of their offerings.

It's crucial to provide our readers with clear information regarding the inherent nature of services and products that might be covered in this newsletter, including those advertised by our sponsors from time to time. When you trade on prediction markets (including event contracts, opinion markets and other speculative instruments) your capital is at risk. Risks associated with prediction markets include price volatility, loss of capital (the value of your position could drop to zero), illiquidity, complexity, evolving regulation and lack of protection. Many prediction market operators do not currently operate in a fully regulated industry, and availability varies by jurisdiction. Therefore, please be aware that when you place funds on prediction markets, you may not be protected under financial compensation schemes and protections typically afforded to investors when dealing with regulated and authorised entities to operate as financial services firm. Nothing in this newsletter constitutes a recommendation to place, hold, or close any position on any market.