Bally's Intralot shareholders approved the acquisition of evoke, parent of William Hill and 888, at its Athens AGM on 17 September. The 52p-per-share deal values evoke at £243.1m and passed with 99.585% in favour, matching evoke's own 99.63% in August.
£243.1m for William Hill and 888 together is the number to keep. These are two of the best-known brands in British betting, with a combined history approaching two centuries, and the equity is being valued at roughly what a mid-sized technology supplier fetches. That is the clearest available mark on what UK-exposed online assets are worth once the duty rise is priced in.
The near-unanimity on both sides — 99.585% and 99.63% — says the shareholders had stopped seeing an alternative. Votes at that level are not endorsements of price; they are the absence of a competing bid and the presence of leverage. evoke has carried significant debt since the William Hill international acquisition, and a buyer willing to take that on removes a refinancing problem as well as an equity one.
For competitors the relevant change is what emerges afterwards. A Greek-American owner running William Hill alongside Intralot's lottery business has no legacy UK retail commitments to defend and no London listing to answer to quarterly. That is a freer hand on cost than any of its UK-listed rivals have, entering a year in which cost is the whole game. The delisting also thins London's gambling sector further, which matters for how the industry is covered and valued.
The Gambling Commission's industry activity report for April 2025 to March 2026 shows GGY up 4.4% to £17.5bn across all channels, driven by online while physical premises declined. Excluding lotteries, GGY rose 4.7% to £13.2bn.
This is the last clean read of the UK market before the duty rise works through, which makes it the baseline every submission this autumn will argue from. Both sides will use it. The industry will point to operators falling 1.1% while yield rises, as evidence of a market already consolidating under pressure. The reform side will point to £17.5bn and 4.4% growth as evidence that the sector is not struggling at all.
Slots are the exposed line. At £4.8bn they are 84% of online casino and about 27% of the entire market including lotteries. Any intervention on stake limits, spin speed or bonus mechanics lands almost entirely here, and there is no other vertical large enough to absorb the loss. Remote betting at £2.4bn is less than half the size, and within it horse racing at £769.3m is now comfortably smaller than football at £1.2bn — which is the number the racing industry's funding argument has to contend with.
The structural story is channel shift continuing rather than accelerating. Remote at £8.3bn is roughly 63% of non-lottery yield and grew 6.9% while premises declined. That is a steady migration, not a collapse, and it means the land-based estate is still large enough that a machine games duty rise has real employment consequences — which is precisely the argument Entain and Betfred are making with it.
Prime Minister Hun Manet said all online gambling operated from within physical casinos will stop from early October, citing fraud and money laundering conducted through the technology. No exact date was given.
Read the framing rather than the policy. Cambodia is not describing a gambling-harm intervention; it is describing an anti-fraud operation, and the distinction explains the target. Online operations licensed through physical casinos have functioned as legal cover for businesses whose customers are elsewhere in Asia, and some of that infrastructure overlaps with the scam compounds that have drawn sustained regional and international pressure. Closing the licensing route is a cheaper answer than policing what runs on it.
The likely effect is displacement rather than reduction. Volume of this kind moves to whichever neighbouring jurisdiction is most permissive next quarter, which has been the pattern every time a Southeast Asian market has tightened. For the regional picture that means the Philippines, which is simultaneously closing 8,000 disguised payment merchants, becomes a less available alternative than it was a year ago.
The immediate commercial exposure sits with suppliers rather than operators. A platform provider, aggregator or payments business with receivables from a Cambodian licensee is now three weeks or less from its counterparty losing the basis on which it operates, with no published date and no regulator notice to plan against. That is a collections question this month, not a strategy question next year.
Eeze's Aarti Carl says a growing number of gaming companies are establishing themselves in Bucharest, describing a migration of operational functions into the Romanian capital.
Romania's attraction is the combination that Malta used to own: a licensed domestic market, EU membership, English-capable technical staff, and a cost base well below Western Europe. The difference is scale — Romania's own regulated market is substantial, so a company based there is not purely an offshore back office.
The timing is awkward and worth holding in view. The same week the industry describes Bucharest as a hub, the regulator is on its third president in two years and working through an audit that raised questions over close to €1bn in unpaid fees and taxes.
Those two facts are not contradictory. Operational relocation follows cost and talent; licensing risk follows the regulator. A company can rationally move its engineering to Bucharest while treating the Romanian licence itself as a separate and less settled question.
Sportsbook supplier BETBY has entered the Polish regulated market via a deal with local operator Betters, supplying its full sportsbook along with an esports feed.
Poland is a hard market to enter and an easy one to overlook. Online casino is a state monopoly through Totalizator Sportowy, so the entire private opportunity sits in sports betting, and the licensing regime has kept the supplier field narrow for years.
That narrowness is exactly why a supplier deal matters here more than it would in Italy or Spain. Betters is buying a competitive sportsbook rather than building one, which raises the product floor for every other licensed Polish operator at once.
For anyone tracking CEE, this is the shape entry takes in a closed regime: not a licence application, but a supply agreement with someone who already holds one.
Lula's administration is preparing a provisional measure that would prohibit online casino games while leaving sports betting and its sponsorships legal, with the final language expected before the 4 October first round. The National Association of Gaming and Lotteries puts roughly three quarters of industry revenue in the products that would be banned.
The shape of the measure tells you who drafted it. Casino and sports betting produce broadly similar problem-gambling profiles, so a ban that removes one and protects the other is not a harm-reduction design. It is a design that takes the product with the fewest visible defenders and leaves the one that puts logos on football shirts. Brazilian reporting describes the measure as coming from the political wing of Planalto with the technical staff sidelined, and the text does exactly what that description predicts.
The instrument matters as much as the content. A medida provisória takes legal effect on publication and only later goes to Congress for conversion into law. That is what makes it attractive five weeks before a first round: the announcement lands immediately and the legislative fight happens afterwards, possibly under a different government. It is also what makes it dangerous for an operator, because there is no consultation window in which to negotiate.
What a licensee should be doing this week is separating two exposures that are usually modelled together. The first is revenue: if roughly three quarters of your Brazilian revenue is casino, a sports-only Brazil is a different business with a different margin, and the question is whether it clears your cost of capital at all. The second is the claim: the licence was paid for, it covers both activities, and the state's obligation to refund the unused portion is contractual rather than discretionary. Those are two separate conversations with two separate sets of advisers, and the second one has a deadline set by the publication date.
Television directors have opened a campaign in Brasília to stop the measure, briefing affiliate owners who also hold seats in Congress to argue the order is unconstitutional. The networks are not just losing ad revenue: Globo, SBT and Band hold stakes in betting platforms including BetMGM, Bet do Milhão and BandBet.
Equity changes the character of a lobby. A broadcaster losing advertising revenue complains; a broadcaster holding a stake in the advertiser litigates. Globo, SBT and Band are in the second position, and that converts what would have been a commercial grievance into a constitutional challenge with balance-sheet motivation behind it.
The congressional route is the detail worth noting. Brazilian broadcast affiliates are frequently owned by regional political families, and several of those owners hold seats. Briefing them to argue unconstitutionality is not lobbying from outside the legislature — it is activating members who already sit inside it and who have a personal financial interest in the outcome. That is a materially stronger mechanism than an industry association writing letters.
The football side adds the public argument the broadcasters cannot make credibly themselves. A manifesto titled "the end of Brazilian football", signed by clubs and federations, frames the ban as an attack on the sport rather than a defence of consumers. Whether or not that is fair, it is the frame most likely to reach voters in the five weeks that matter, and it is being carried by the same networks that would otherwise have to report the government's case neutrally.
Each operator paid R$30m for a five-year licence. If Planalto closes online casino before those terms expire, the state must refund proportionally to the unused period — and online casino is roughly 80% of licensee revenue.
The refund obligation is what turns a political gesture into a budget problem. A provisional measure takes effect on signature, but the Fiscal Responsibility Law requires the government to identify where the replacement revenue comes from — and doing that five weeks before an election means naming either a tax rise or a spending cut.
That is why the reporting describes a legal manoeuvre being designed: a grace period that defers the binding effect, pushing the definitive decision to Congress. It protects the government from an immediate legal-certainty crisis while still allowing the announcement.
For an operator the practical question is which of those two things actually arrives. An MP with a long carência is an announcement; an MP with immediate effect is a balance-sheet event. The text will tell you which within days of publication.
The Attorneys General of Alberta and Ontario — the only two provinces with multi-licence markets — will argue that Ontario should be allowed to pool iGaming and DFS players with other jurisdictions. Flutter is among those joining them.
Liquidity is the whole argument. Poker and daily fantasy are network products: their economics depend on the size of the player pool, and a province-sized pool is too small to sustain competitive prize structures or deep cash games. That is why Ontario poker has underperformed relative to the market's overall success.
The opposition is revealing. The provincial lottery corporations objecting are state-owned monopolies in provinces that have not opened to private operators, and a ruling for Ontario makes their closed model look more clearly like a choice rather than a necessity.
Flutter joining the attorneys general is the commercial tell: PokerStars is the asset that gains most from pooling, and Flutter rarely litigates in public without a revenue line attached.
Sixty days into Alberta's regulated iGaming market, more than 50 operators have registered but only around 24 are live, according to mkodo's commercial director reflecting on early geolocation data.
The 50-to-24 gap is the most useful thing in this data. Registering is cheap and signals intent to investors; launching requires payments, geolocation, compliance and marketing spend in a province of roughly five million people. Half the field has decided the second step can wait.
That matters for anyone modelling Alberta's steady state. The competitive intensity operators budgeted for assumed a full field at launch, and it has not materialised — which means acquisition costs in the first quarter are probably lower than planned, and the crowding arrives later.
Alberta is also the control case for Ontario. Same country, later launch, smaller population, and a market design that learned from four years of Ontario data. What diverges between them is the useful signal.
Internal documents and six former employees told the New York Times that DraftKings built a machine-learning model in 2023 ranking customers by how much they were expected to lose after receiving a promotion. Around $400m of AI-automated promotional spend in 2025 was directed at the highest-scoring users.
The technique is ordinary and that matters to how you read this. Expected value per promotion is the basic arithmetic of retention marketing: a free bet costs something, returns something, and the return varies by customer. The inputs described are the standard ones. Nobody in the industry should claim to be surprised that such a model exists.
What makes this different from a routine CRM story is the combination of three things. There is a revenue concentration figure — 42% from 3.8% of users — which quantifies exactly how dependent the business is on a small, heavy-losing cohort. There is a spend figure, around $400m in 2025, showing the model was not a pilot but the allocation mechanism for the promotional budget. And there is the shelved harm-detection work: a data scientist built a model to identify players in trouble, and the presentation was cancelled on the day it was due, with two further efforts dropped. Building the capability and then declining to deploy it is materially worse, evidentially, than never having built it.
The quotes are what will follow the industry around. One former analyst told the Times "it is as predatory as it sounds", adding "if you lose more, we give you more, so you keep playing more". Jayden Butts, who tested the model, is quoted saying "the best investment would be a problem gambler". DraftKings rejects the characterisation, says promotions target customers showing "sustained, engaged use" rather than losses, and called Butts's analysis preliminary and inconclusive.
The regulatory vehicle already exists. The SAFE Bet Act, introduced by Senator Blumenthal and Representative Tonko in September 2024, would prohibit AI tracking of individual habits and personalised promotional offers outright. It has gone nowhere for two years. A story with these numbers and these quotes is what moves a stalled bill onto a markup calendar, and Baltimore's April 2025 consumer-protection suit shows the municipal route is live too.
NOTE ON SOURCING: the New York Times original is paywalled and was not read directly. This account is reconstructed from secondary coverage of that reporting, and every figure here is attributed to the Times via those reports rather than independently verified.
Genius Sports has launched Prediction.com, a comparison site aggregating pricing across eleven prediction-market apps, built on the Legend media network it bought in February 2026 for $1.2bn. The listed venues include Kalshi, Polymarket, Novig, Underdog, FanDuel Predicts and Fanatics.
Comparison layers change where margin sits, and the UK sportsbook market is the worked example. Before Oddschecker, a book competed on brand, product and marketing. After it, a meaningful share of customers began their journey on a page that displayed every price side by side, and the book's job narrowed to being the best number in a row. Acquisition rent moved to whoever owned that page.
Applying the model to prediction markets compresses something already thin. Exchanges run on take rates measured in fractions of a percent rather than the 8-10% hold of a sportsbook. A venue that must also be the cheapest visible price on a comparison page has very little room left, which accelerates exactly the consolidation that Robinhood's own VP said he expects among the twenty-plus venues now live.
The position Genius has built is the interesting part. It sells official data to the venues and now owns the shelf where their prices are compared, which means it earns on supply and on discovery simultaneously. It also did this without new capital — Legend was bought in February for other reasons and repointed. For a sportsbook the uncomfortable extension is obvious: the same engine, run by the same company, is one product decision away from showing contract prices and sportsbook prices on a single screen, and that is a comparison most books would rather their customers never made.
Macquarie now expects taker volume — recreational and retail flow that removes liquidity — to reach $190bn this year, up from its previous $169bn estimate. Makers are the professional liquidity providers on the other side.
The taker/maker split is the detail that makes these forecasts readable. Headline "volume" on an exchange double-counts: every contract has a buyer and a seller. Taker volume isolates the side that behaves like a bettor — someone crossing the spread to get a position now — which is why it is the right number to set against sportsbook handle.
At $190bn, taker volume would exceed the handle of most individual US sportsbooks by a wide margin. But it is a volume figure, not a revenue figure: at the take rates seen over the NFL opening weekend, the revenue implied is a fraction of what the same handle produces for a book holding 8-10%.
The upgrade being driven by new venue launches rather than by growth at existing ones is the part worth watching. It suggests the number is being pushed by supply, and supply is what Robinhood's own VP expects to consolidate.
EKG forecasts about $40.1bn wagered on the NFL in the 2026 season, split roughly $31.7bn through regulated sportsbooks and $8.4bn through prediction markets.
This is the most useful number in the prediction-markets debate because it is falsifiable. Every other framing — valuations, download charts, volume records — measures attention. EKG has put a share figure on the season that can be checked in February.
The 21% is also higher than most sportsbook executives were assuming twelve months ago, which is the real content. A category that did not meaningfully exist for NFL betting two seasons ago taking a fifth of handle is a structural shift even if sportsbooks keep the majority.
Read it alongside the download data: prediction markets are winning the acquisition contest while sportsbooks keep the volume. That gap closes or it does not, and this season is where it gets measured.
Citizens and Sensor Tower data for NFL Week 1 shows Polymarket at 592k downloads and Kalshi at 415k, ahead of DraftKings on 265k and FanDuel on 139k.
Downloads are an acquisition metric, not a revenue one, and the two diverge sharply here — sportsbooks still expect roughly four fifths of the season's handle. But acquisition is the leading indicator, and the leader on it is not the leader on handle.
The likely explanation is partly structural: prediction markets operate in all fifty states, so their addressable install base is roughly double that of a sportsbook restricted to regulated states. A like-for-like comparison would need to normalise for that, and this data does not.
What it does establish is that the category no longer needs to be explained to consumers. An app that has to teach the product does not top the chart in Week 1.
Fred Done confirmed in The Sunday Times that Betfred will not renew its rugby league sponsorship, held for almost a decade and reported to be worth millions a year. He warned the bookmaker will also leave horseracing's five British Classics if machine games duty rises in next month's budget.
Sponsorship is the first budget line to move because it is the easiest to stop. It renews on a cycle, it has no redundancy cost, and cancelling it produces a headline — which for an operator arguing against a tax rise is a feature rather than a side effect. That does not make the decision insincere, but it does mean the sequencing tells you about negotiating tactics as well as about economics.
The Classics threat is the one with real leverage. Rugby league is a regional sport with limited political weight; British horseracing is a national industry with a funding model that depends directly on betting, a levy mechanism written into statute, and constituency MPs in racing towns. Betfred walking away from the five Classics would force a conversation the Treasury has so far avoided, which is why the threat is stated conditionally and the rugby league decision is stated as done.
Watch what this does to sponsorship pricing rather than to Betfred. If the largest independent bookmaker exits two sports properties in one budget cycle, rights holders across British sport have to reprice gambling inventory downward — and that reduces the value of exactly the sponsorship assets the industry cites when arguing that it funds sport. The argument and the retreat undercut each other.
The provisional measure would strip roughly three quarters of licensed operators' revenue from a framework completed in 2024 — and because each licence cost R$30m for five years, the state has to refund what it takes away while the Fiscal Responsibility Law makes it name the replacement revenue.
The design tells you who is writing it. Sports betting survives, along with its sponsorships and its visibility; online casino, which the industry association puts at around three quarters of revenue, does not. That is not a harm-reduction architecture — casino and sports betting produce similar problem-gambling profiles — it is a structure that removes the product with the fewest visible constituencies while protecting football. Reporting in Brazil describes the measure as coming from the political wing of Planalto with technical staff sidelined, which is consistent with what the text does.
The fiscal arithmetic is where it gets difficult. Operators paid R$30m each for five-year licences covering both activities. Closing one of those activities early creates a proportional refund obligation, and the Fiscal Responsibility Law requires the economic team to identify a compensating source for the revenue it gives up — five weeks before a first round. Licence fees and taxes have already delivered close to R$10bn. The reported workaround is a grace period that defers binding effect and pushes the final decision to Congress, which would let the government announce a ban without immediately paying for one.
The opposition is the part operators should study rather than celebrate. It is not the gambling lobby. It is Globo, SBT and Band, which hold equity in BetMGM, Bet do Milhão and BandBet and have briefed affiliate owners sitting in Congress to argue unconstitutionality, alongside nine major clubs warning of insolvency. When the broadcasters who shape election coverage and the clubs that carry the shirts are both funding the constitutional challenge, the probability of a clean, immediately binding ban falls sharply — and the probability of a messy, deferred, litigated one rises.
DraftKings built a machine-learning score ranking customers by how much each promotion was expected to lose them, and the practice is now documented in the New York Times with internal papers and a named former analyst — which converts routine promotional economics into a discoverable record every operator will be asked about.
Strip the reporting back and the model is ordinary. Expected value per bonus is the basic unit of CRM: you work out what a free bet costs, what it returns, and who to send it to. The inputs described — play frequency, daily balance, ratio of amount staked to amount lost, plus a churn probability — are the inputs any competent retention team uses. Nothing here is technically novel.
What is new is the evidentiary position. There is an internal name, "elasticity". There is a start date, September 2023, and a test cohort of roughly 5,000 players. There is a named former employee who says he began questioning whether the high-scoring customers were also the vulnerable ones. That combination is what a regulator subpoenas and a plaintiff's lawyer builds on, and it exists now whether or not the model was ever misused.
The commercial consequence is not about DraftKings. It is that every operator's targeting stack is now presumptively in scope, and the defensible position is documentary rather than technical. An operator that can show its model excludes players whose loss rate is accelerating, that the exclusion is tested, and that someone senior signs off on the rule, is in a different position from one that simply never wrote the rule down. The gap between those two operators is currently invisible and is about to stop being.
Genius Sports has launched a price-comparison site across eleven prediction-market venues — the Oddschecker model applied to a category whose legal status is still being litigated — and it built it on an asset it already owned rather than on new capital.
Prediction.com sits on the Legend media network Genius acquired in February for $1.2bn. That matters more than the launch itself: Genius did not decide to enter this category and then buy the means, it found it already had them. A media network bought for betting affiliation repoints at prediction markets with no incremental investment, which is the cheapest possible option on a category that might not survive its court cases.
Comparison layers do a specific thing to a market. Once price is visible side by side, the venue competes on cost and the comparison site captures the acquisition rent. That is precisely what happened to UK sportsbooks after Oddschecker, and it is why margin in that market migrated toward whoever owned the customer's first click. Applying it to exchanges that already run on thin take rates compresses a thin thing further.
The strategic read connects to what Robinhood's own VP described three weeks ago: the exchange is a pipe, and the value sits with whoever holds the customer. Genius is now positioned on both sides of that — it sells the official data the venues need and owns the shelf where their prices are compared. For a sportsbook the uncomfortable implication is that the same comparison engine, run by the same company, is one product decision away from displaying sportsbook prices and contract prices on the same screen.