A sports picks service sells predictions about sporting events to paying subscribers, usually monthly, usually delivered through Discord or Telegram, usually billed through a creator-commerce platform such as Whop.
The pitch is that you have an edge and are renting it out. The honest description is that you are selling a high-variance information product into a market where the underlying activity has a negative expected return for almost everyone who acts on it, where your marketing is regulated in three of your four biggest markets, and where your payment processor may classify you as a prohibited business.
None of that makes it illegal. Most of it makes it harder than the marketing suggests. This guide covers the arithmetic that determines whether a track record means anything, the advertising rules that actually apply in the US, UK and Australia, the platform and payment restrictions, the subscription churn that quietly kills these businesses, and the conflict of interest buried in the affiliate revenue that most services rely on.
Standard American odds on a point spread or total are minus 110. You risk 110 to win 100. The break-even win rate is 110 divided by 210, which is 52.381 per cent.
A bettor who picks at random wins about half the time. At minus 110 that produces a return on turnover of minus 4.55 per cent. This is the vig, and it is the reason the phrase "I hit 55 per cent last season" is not, on its own, evidence of anything.
Here is why. If a thousand people each post 100 coin-flip picks at minus 110, the binomial distribution says roughly 18.4 per cent of them will finish at 55 per cent or better. Around 184 people out of a thousand will have produced, by chance alone, a record good enough to sell. About 31 per cent of them will finish above the break-even rate and can truthfully claim to have been profitable.
Now run it the other way. Suppose you genuinely win 55 per cent of your minus 110 bets, which would be an exceptional long-run rate. How many picks does it take to demonstrate that you are better than break-even, rather than lucky? Using a one-sided test at the 5 per cent level with 80 per cent power against a null of 52.381 per cent, the answer is roughly 2,240 bets. If your true rate is 54 per cent, roughly 5,880 bets. If it is 57 per cent, roughly 720.
Two things follow. First, almost no picks service has a statistically meaningful record, because almost none of them have posted several thousand graded picks. Second, the services that have posted several thousand picks are the ones worth examining, and they are rare.
This audit found no published, independently maintained dataset of picks-service performance across the industry. Anyone quoting an industry-wide hit rate is estimating.
Sportsbook hold, the share of money wagered that the book keeps, has risen sharply in the US since legalisation.
The American Gaming Association reported total commercial sports betting revenue of 16.89 billion dollars in 2025, an increase of 22.6 per cent on the previous year. Legal Sports Report, working from AGA-published handle and revenue figures, calculated implied national hold of roughly 10.2 per cent in 2025, against 9.2 per cent in 2024, 8.0 per cent in 2022 and 6.9 per cent in 2019.
Hold rising from around 7 per cent to over 10 per cent in six years is not a small change. It means the average dollar wagered now loses about half as much again as it did in 2019. The main driver is the shift towards parlays and same-game parlays, which carry a far higher theoretical hold than a single point spread.
The practical consequence for a seller: the bar for genuinely adding value has moved up while the marketing language has not. A service built on same-game parlays is selling a product with a structurally worse expected return than one selling standard sides, whatever the advertised win rate looks like, and subscriber results will reflect that.
A screenshot is not a record. A spreadsheet you maintain is not a record. A Telegram channel where messages can be edited or deleted is not a record.
The last of these is a large part of why regulators and payment processors treat this category as high risk, and why the more serious operators in it have moved to third-party verification.
The current standard is account-synced tracking. Pikkit's BookSync, for example, connects directly to sportsbook accounts and imports every bet placed, across a stated 30-plus sportsbooks, rather than relying on manual entry. The same product reports closing line value: whether the price taken was better than the price the market settled at before kick-off.
Closing line value matters more than win rate over short samples. Consistently taking prices better than the closing line means beating the market's final estimate, and that signal emerges in far fewer bets than a win-rate edge does. A record showing a positive win rate but negative closing line value is a record of good luck.
Synced tracking is a real improvement. It is not proof:
If you are going to sell picks, the defensible version is this: every pick timestamped and published before the event, at a stated price, with a stated stake, graded automatically, on a platform you do not control, with the whole record public including the losing months. Anything short of that is a marketing asset, not a record.
There is a structural problem with proving an edge by betting: sportsbooks restrict accounts that win.
Massachusetts became the first US state to regulate the practice directly. Following an investigation that began in 2023 and a public roundtable in 2024, the Massachusetts Gaming Commission adopted a rule, effective 1 June 2026, requiring licensed operators to tell customers when and why their betting has been limited. In evidence to the Commission, operators put the share of accounts carrying limitations at 0.64 per cent.
That figure cuts both ways. It suggests limiting is rare across the whole customer base. It also suggests that a consistent winner sits in a very small group the books can restrict without commercial consequence.
For a picks seller this is not abstract. The ability to maintain a synced, verified record depends on being able to place the bets you publish, at the prices you publish, in size. Once the main accounts are limited to trivial stakes, the public record stops reflecting anything a subscriber could replicate. Sellers who reach that point typically stop publishing synced records and revert to screenshots, which is why a service that once published verified results and no longer does is worth a second look.
In most jurisdictions, selling opinions about sporting events is not a licensable gambling activity. You are not accepting wagers, not holding customer funds against outcomes, and not operating a pool.
Cross into handling money and the analysis changes completely, and fast.
The US Securities and Exchange Commission has litigated exactly this. In SEC v. Nevada Sports Investment Group, LP (Litigation Release No. 24264, September 2018), the Commission charged an entity formed under a 2015 Nevada law that permitted Nevada-based businesses to pool money from investors anywhere in the world for sports wagering. The entity had raised over one million dollars from more than 30 investors, solicited generally over the internet, and failed to determine whether its investors were accredited. It was charged with an unregistered offering of securities under Sections 5(a) and 5(c) of the Securities Act of 1933, and settled to a permanent injunction.
18 U.S. Code section 1084 is frequently cited in this context, usually inaccurately. The operative text of subsection (a) reads:
"Whoever being engaged in the business of betting or wagering knowingly uses a wire communication facility for the transmission in interstate or foreign commerce of bets or wagers or information assisting in the placing of bets or wagers on any sporting event or contest ... shall be fined under this title or imprisoned not more than two years, or both."
Two limits matter. The prohibition attaches to a person "being engaged in the business of betting or wagering", which on its face targets bookmakers rather than commentators. And subsection (b) provides that nothing in the section prevents transmission "of information for use in news reporting of sporting events or contests, or for the transmission of information assisting in the placing of bets or wagers on a sporting event or contest from a State or foreign country where betting on that sporting event or contest is legal into a State or foreign country in which such betting is legal."
Subsection (c) then states that the section creates no immunity from prosecution under state law.
The honest summary: a subscription tipping service, selling analysis and not taking wagers, is not the target of section 1084, and the news-reporting and legal-to-legal carve-outs in subsection (b) point away from it. That is not a guarantee, and this is a criminal statute. Anyone building a business on the point should take advice in their own jurisdiction rather than rely on a guide.
US Advertising Rules: The FTC Is the Real Constraint
For a US-facing picks service, the binding rules are consumer-protection rules, not gambling rules.
The Endorsement Guides (16 CFR Part 255) require disclosure of material connections between an endorser and an advertiser, and require that endorsements reflect performance consumers can generally expect. A testimonial from a subscriber who received a free membership, a payment or anything else of value is a paid endorsement and must be disclosed clearly and conspicuously. A testimonial describing an atypical run of results does not become compliant because small print says results are not typical; the FTC's own examples make clear that a disclaimer which fails to communicate what a customer can generally expect is inadequate.
The Rule on the Use of Consumer Reviews and Testimonials, announced by the FTC on 14 August 2024 and effective 60 days after Federal Register publication, prohibits creating, buying, selling or disseminating fake or false consumer reviews and testimonials, including AI-generated ones; providing compensation or incentives conditioned on a review expressing a particular sentiment; undisclosed insider reviews from officers, managers, employees or agents; company-controlled sites presented as independent review portals; suppressing negative reviews through unfounded legal threats or intimidation; and buying or selling fake indicators of social media influence such as bot-generated followers or views.
That last item lands directly on this category. Buying followers to make a new picks account look established is specifically enumerated in a trade regulation rule that carries civil penalties. The maximum civil penalty under Sections 5(l) and 5(m)(1)(A) of the FTC Act stands at 53,088 dollars per violation following the January 2025 inflation adjustment, with no further increase applied for 2026.
Separately, several US states require registration or licensing for anyone marketing sportsbooks on a commission basis. The categories, fee levels and background-check requirements are set by each state's gaming regulator and differ substantially, so the requirement has to be checked state by state rather than assumed. Being paid to send customers to a sportsbook is a regulated activity in a meaningful number of US jurisdictions. Selling picks alone generally is not.
In practice a platform will stop you long before a regulator writes to you.
Google Ads treats gambling-related content as a restricted category requiring certification. Advertisers must apply for certification, provide licensing documentation from the relevant authority, keep that licence valid, notify Google if it lapses, and apply separately for each country targeted. Aggregator and affiliate sites that provide information about or comparisons of gambling services are permitted under the policy, but the destination must not itself offer gambling services or link to gambling services the advertiser owns. Ads must not target minors and must carry responsible-gambling information.
Meta defines online gambling and games as any product or service where anything of monetary value is included as part of a method of entry and prize, and allows ads only from accounts that have obtained authorisation through the Authorisations and Verifications tab in Meta Business Suite, supported by evidence that the activity is appropriately licensed or otherwise lawful in the territory targeted. Critically for picks sellers, the authorisation requirement extends to ads whose destination landing pages contain promotions for online gambling, even where the page itself accepts no wagers. A picks landing page carrying sportsbook affiliate offers is inside that.
The consequence is that the standard paid-acquisition playbook is largely unavailable. Most picks services grow through organic social content, free channels and cross-promotion with other sellers, which is slower, less predictable, and far more dependent on the personality at the front of it. Budget for that, because it is the difference between a business you can scale with money and one you can only scale with time.
UK Rules: Inside the Advertising Code Without Needing a Licence
Great Britain draws the licensing line where you would expect. The Gambling Commission licenses those who provide facilities for gambling. A tipping service does not provide facilities for gambling and, on that basis, is not an operating-licence holder.
The advertising position is different, and it catches people out.
The CAP Code's Section 16 scope note states that the section's rules apply to marketing communications by gambling operators licensed in Great Britain and by firms authorised to provide spread betting products, and "includes marketing by third parties (for example, affiliate marketers) acting on an advertiser's behalf". It then adds: "Although they do not apply to marketing communications for non-gambling operators, the ASA may draw on the principles established in the rules to assess whether ads for products likely to encourage gambling (for example, betting tipsters) meet the standards required by the general social responsibility provisions of the Code."
Betting tipsters are named in the Code. The ASA will assess a tipster's advertising against Section 1 social responsibility, drawing on the Section 16 principles. Those principles state that marketing communications must not:
- portray, condone or encourage gambling behaviour that is socially irresponsible or could lead to financial, social or emotional harm (16.3.1)
- exploit the susceptibilities, aspirations, credulity, inexperience or lack of knowledge of children, young persons or other vulnerable persons (16.3.2)
- suggest that gambling can provide an escape from personal, professional or educational problems such as loneliness or depression (16.3.3)
- suggest that gambling can be a solution to financial concerns, an alternative to employment or a way to achieve financial security (16.3.4)
- portray gambling as indispensable or as taking priority in life, for example over family, friends or professional or educational commitments (16.3.5)
- suggest that gambling can enhance personal qualities, or is a way to gain control, superiority, recognition or admiration (16.3.6)
- be likely to be of strong appeal to children or young persons, especially by reflecting or being associated with youth culture (16.3.12)
- include a child or young person, and no-one who is or seems to be under 25 years old may be featured gambling or playing a significant role (16.3.14)
Rule 16.3.4 is fatal to the standard picks marketing script. "Quit your job", "replace your income" and "financial freedom through betting" all describe gambling as an alternative to employment. That is the wording of the prohibition, almost exactly.
Taking affiliate commission from a British-licensed operator adds a second problem. Social responsibility code provision 1.1.2 of the Gambling Commission's Licence Conditions and Codes of Practice requires licensees to take responsibility for third parties with whom they contract for any aspect of their business related to the licensed activities, and to take all reasonable steps to ensure marketing restrictions are not breached. The Commission's guidance on affiliates states that where this fails it "may take regulatory action against the business, the relevant Personal Management Licence holders, or both".
The operator carries the regulatory risk for your advertising. Operators respond by writing broad termination rights into affiliate contracts and using them. Your revenue can end because of one post.
Australia: The Rules Are Tightening Right Now
Australia is the jurisdiction where the direction of travel is clearest.
The existing framework is the Interactive Gambling Act 2001, enforced by the Australian Communications and Media Authority, which prohibits the provision and the advertising of prohibited interactive gambling services to Australians. Promoting an offshore, unlicensed sportsbook to an Australian audience is not a grey area.
The Murphy review, "You win some, you lose more", reported in June 2023 with 31 unanimous recommendations including a phased total ban on gambling advertising and a national regulator. The government's response was tabled in May 2026 and was widely criticised, including by crossbench members and public health groups, for falling well short of that.
On 2 April 2026 the government announced a package of measures, and the Interactive Gambling Amendment (Gambling Reform) Bill 2026 was introduced to Parliament on 2 July 2026 and referred to the Environment and Communications Legislation Committee, with a report due 17 August 2026. As summarised by the law firm Hall and Wilcox, the Bill would:
- ban wagering advertising during live sport coverage between 6am and 8.30pm, from five minutes before to five minutes after
- cap television wagering ads at three per 60-minute period during daytime hours
- ban radio wagering ads during school drop-off and pick-up windows
- prohibit wagering branding on sporting uniforms and at venues
- prohibit "notable persons", including current and former professional athletes, celebrities, social media influencers and prominent individuals, from promoting wagering services
- ban distinct promotional references to specific odds
The reported maximum civil penalties are substantial: AUD 364,000 for individuals, AUD 1,820,000 for bodies corporate, and AUD 2,730,000 for anti-avoidance breaches. Prohibitions would commence on 1 January 2027 with a phased implementation over three years.
Two caveats. This is a Bill, not law, and the final text may change or the Bill may not pass. And the restrictions target the promotion of wagering services rather than the sale of tips as such. But the "notable persons" prohibition is drafted to capture social media influencers promoting wagering, which is precisely what a picks seller carrying affiliate links is doing.
If Australia is a meaningful part of your audience, build on the assumption that influencer promotion of wagering will be restricted and that affiliate revenue from Australian traffic is on a timer.
Canada: Ontario Has Already Banned the Endorsement Model
Ontario, which operates the largest regulated online gaming market in Canada, amended Standard 2.03 of the Registrar's Standards for Internet Gaming to prohibit the use of athletes in internet gaming advertising and marketing, and to restrict the use of celebrities who would likely be expected to appeal to minors. The Alcohol and Gaming Commission of Ontario announced the change in August 2023 and it took effect on 28 February 2024. Legal analyses of the amended standards note that the restriction on celebrity promoters extends to social media influencers.
Canadian provinces regulate separately, so Ontario is not Canada. But it is the largest market in the country and it has already closed off the endorsement-led growth model.
Payments: You May Be a Prohibited Business
This is the part that ends more picks services than any regulator does.
Stripe's restricted businesses list places the following under Gambling, in its prohibited category, verbatim: "Games of chance including gambling, internet gambling, casino games, sweepstakes and contests, and fantasy sports leagues with a monetary or material prize"; "Games of skill including video game and mobile game tournaments or competitions, darts, card games and board games with a monetary or material prize"; "Payments of an entry or player fee that promise the entrant or player will win a prize of value"; and "Sports forecasting or odds-making with a monetary or material prize".
The qualifier "with a monetary or material prize" matters and should not be glossed over. A plain subscription to analysis, with no prize element, is not obviously inside that entry. But the same list also prohibits, under unfair, deceptive or abusive acts or practices: "'Get rich quick' schemes, including investment opportunities or other services that promise high rewards to mislead consumers; schemes that claim to offer high rewards for very little effort or up-front work; and sites that promise fast and easy money"; "Businesses that make outrageous claims, use deceptive testimonials, use high-pressure upselling, or use fake testimonials"; and "Negative option marketing, negative option membership clubs and reduced price trials with unclear or hidden pricing".
Read those three together and the standard picks-service funnel, with its income claims, its screenshot testimonials and its one-dollar trial that rolls into a monthly charge, is described almost line by line. Terminations in this category are usually justified under those clauses rather than the gambling clause.
PayPal places "Gambling, Gaming, Prize Draws and Contests" in its pre-approval category, requiring written permission and permitting the activity only where operator and customers are located exclusively in jurisdictions where it is lawful.
The upshot is that most picks services end up on a creator-commerce platform that has already solved payment acceptance for them, and accept that platform's terms and its ability to switch them off.
What Whop Actually Costs
Whop is the dominant storefront for this category. Its published fee schedule, checked in August 2026, is specific:
- Card payments: 2.7 per cent plus 0.30 dollars per successful domestic transaction, plus 1.5 per cent for international cards and a further 1 per cent where currency conversion is required
- ACH debit: 1.5 per cent, capped at 5 dollars
- Financing partners such as Klarna and Afterpay: 15 per cent per transaction
- 3D Secure: 0.03 dollars per transaction. Machine-learning fraud detection: 0.07 dollars per transaction
- Dispute or chargeback: 15 dollars per dispute. Early dispute alert: 29 dollars per alert
- Payment orchestration: 0.8 per cent. Billing automation: 0.5 per cent. Tax calculation and remittance: 2 per cent where tax is collected
- Affiliate processing: 1.25 per cent per transaction
- Payouts: next-day ACH 2.50 dollars; instant bank deposit 4 per cent plus 1.00 dollar; crypto or Venmo 5 per cent plus 1.00 dollar; bank wire 23.00 dollars
One disagreement is worth flagging. Several third-party reviews describe Whop as charging a flat 3 per cent platform commission on top of card processing. Whop's own pricing page and fee documentation, checked for this guide, list processing, payout, fraud, optimisation and affiliate fees but show no separate flat platform commission. Where a platform's own fee page and third-party summaries disagree, model both scenarios and get the answer in writing before launch.
On the published card rate, a single monthly charge nets out as follows:
- 19 dollars: 0.81 fee, 18.19 net, effective 4.28 per cent
- 49 dollars: 1.62 fee, 47.38 net, effective 3.31 per cent
- 99 dollars: 2.97 fee, 96.03 net, effective 3.00 per cent
- 299 dollars: 8.37 fee, 290.63 net, effective 2.80 per cent
The flat 30 cents makes cheap tiers disproportionately expensive. A 19-dollar tier surrenders 53 per cent more of its revenue to processing than a 299-dollar tier does.
The dispute fee is the one to watch. A single chargeback on a 99-dollar subscription costs the 99 dollars plus 15 dollars, a swing of 114 dollars. That is the entire net margin on 1.19 other subscriptions.
Chargebacks Are the Existential Risk, Not a Nuisance
Sports picks generate disputes at rates most subscription categories never see, for an obvious reason: the customer's assessment of value is set by whether they won money last week.
Card scheme monitoring sets the outer boundary. Under the Visa Acquirer Monitoring Program, with thresholds effective 1 June 2025, the Excessive Merchant threshold in the US, Canada, EU and Asia-Pacific is a VAMP ratio of 220 basis points (2.2 per cent) combined with a monthly count of fraud and disputes of at least 1,500. Visa's own fact sheet states the Excessive Merchant threshold reduces to 150 basis points from 1 April 2026 in those regions. The ratio counts fraud reports plus disputes divided by settled card-not-present transactions.
A small picks service will never reach 1,500 disputes a month, so it will not be identified by Visa directly. That is not reassurance. Your acquirer or platform sits above you in the chain, applies its own far tighter limits, and does so because your disputes count towards their portfolio ratio. Platforms in this category commonly act at ratios well under 1 per cent. The threshold that matters is set by whoever processes for you, and it is much lower than the scheme's.
Three operational conclusions follow. Refund quickly and without argument, because a refund costs the sale while a chargeback costs the sale plus 15 dollars plus ratio. Make renewal terms unambiguous and cancellation easy, because "I did not know it renewed" is among the most common dispute reasons in subscription businesses and is exactly the practice Stripe describes as negative option marketing. And do not sell annual plans on the back of a hot streak: a 1,200-dollar annual charge disputed in month three is a catastrophe where twelve separate monthly charges would not have been.
The Churn Problem Nobody Models
Recurly's cross-industry benchmark data, published from its network in July 2026, puts overall subscriber churn at 3.60 per cent per month, split into 2.34 per cent voluntary and 1.25 per cent involuntary. Industry medians run from 3.22 per cent for SaaS to 4.99 per cent for education, with digital media and entertainment at 4.14 per cent.
This audit could not find any published churn benchmark specific to sports picks services. Anyone quoting one is estimating, and so is the reasoning below. But the direction is not in doubt, for four structural reasons:
- Cancellation is triggered by outcomes, not by usage. A losing week produces cancellations regardless of how good the analysis was, and variance guarantees losing weeks.
- The product is seasonal. A service built on one league faces a cliff at the end of its season. Subscribers who leave in the off-season are re-acquisitions, not retentions.
- The subscriber's bankroll depletes. Because the underlying activity has a negative expected return for most participants, a meaningful share of customers simply run out of money to bet with. Very few subscription categories systematically erode their own customers' ability to pay.
- Involuntary churn runs higher. Card declines correlate with financial stress, and this customer base skews towards it.
Work the arithmetic at plausible rates. Average subscriber lifetime is approximately one divided by the monthly churn rate:
- 5 per cent monthly churn: 20 months average lifetime, cohort half-life 13.5 months
- 10 per cent: 10 months, half-life 6.6 months
- 15 per cent: 6.7 months, half-life 4.3 months
- 20 per cent: 5 months, half-life 3.1 months
- 30 per cent: 3.3 months, half-life 1.9 months
At 20 per cent monthly churn on a 99-dollar plan, gross lifetime revenue per subscriber is about 495 dollars before fees, refunds and disputes. To hold a flat 200 subscribers you must acquire 40 new paying members every month, 480 a year, purely to stand still. If a free channel converts to paid at 2 per cent, that means roughly 2,000 new free members a month. Those conversion figures are illustrative rather than sourced; measure your own.
This is the mechanism that kills picks services. Not a losing month. The permanent, unpaid acquisition treadmill needed to replace subscribers who leave whatever the results are.
A Worked Example at 200 Subscribers
Take a service with 200 members paying 99 dollars monthly on Whop. Gross monthly revenue is 19,800 dollars.
- Whop card fees at 2.7 per cent plus 0.30 per transaction: 594 dollars. Net 19,206.
- Assume 8 per cent of the base are international cards, adding 1.5 per cent on those 16 transactions: about 24 dollars.
- Refunds at 5 per cent of transactions: 10 refunds, 990 dollars.
- Disputes at 1.5 per cent of transactions: 3 disputes, costing 297 dollars in reversed revenue plus 45 dollars in fees.
- One next-day ACH payout: 2.50 dollars.
Net collected: roughly 17,847 dollars, about 90.1 per cent of gross. That is before any of your own costs.
Now the costs that are always understated:
- Data and tools. Odds feeds, historical databases and modelling tools are the real cost base of a service doing actual work. This audit did not find a single defensible figure, because pricing ranges from free public data to enterprise feeds costing thousands per month. Budget an estimate and verify it against actual quotes before you rely on it.
- Your own bankroll. Maintaining a verified record means betting real money at real stakes. Accumulating the roughly 2,240 graded bets needed to demonstrate a 55 per cent edge, at 50 dollars risked per bet, turns over 112,000 dollars. At a genuine 55 per cent rate the expected profit across that sample is about 5,600 dollars with a standard deviation of about 2,248 dollars, so even a genuinely skilled bettor has roughly a 1-in-150 chance of finishing the whole sample down. At a true 53.5 per cent rate, expected profit falls to about 2,393 dollars against essentially the same variance, and the chance of finishing in the red rises to about 14 per cent. You need capital to survive that, and it is capital at risk, not a deductible business expense.
- Time. Producing daily analysis for a paying base is a full working day, seven days a week, in season.
- Moderation and support. A paid channel of 200 needs moderating. A free channel of 2,000 feeding it needs more.
At 200 subscribers this is a real income. It is also a job with no holidays during the season, a customer base that re-evaluates its purchase every week, and a growth channel most ad platforms will not sell you.
The Affiliate Question: Where the Money Actually Comes From
Most picks services do not make most of their money from subscriptions. They make it from sportsbook affiliate commission, which is a share of the losses of the customers they send.
The scale of that industry is public. Better Collective, one of the largest listed sports betting affiliates, reported 337 million euros of revenue in 2025 against 1.2 million new depositing customers referred that year, and stated that 73 per cent of fourth-quarter new depositing customers were generated under revenue-share agreements. Dividing revenue by referred customers gives roughly 281 euros per new depositing customer for that year. That is arithmetic on published figures, not a per-customer rate the company discloses: revenue-share income accrues from earlier cohorts as well as new ones, so the true lifetime value of a single referral is a different and unpublished number.
Better Collective also introduced Value of Deposits as a reported metric and put it at 820 million euros in the fourth quarter of 2025, the total deposited into revenue-share accounts by its referred users in three months.
Two commercial models exist:
- CPA, a one-off payment per qualifying first-time depositor, usually tied to a minimum deposit and a wagering threshold inside a defined window. Public reporting of specific US CPA rates is thin. Figures in the range of 100 to 500 dollars per depositor are widely quoted in industry commentary, but this audit could not verify them against a primary source, so treat them as an estimate rather than a fact.
- Revenue share, an ongoing percentage of the operator's net gaming revenue from your referred players. Some agreements carry negative carryover, meaning a month in which your referred players win is carried forward as a deficit against future commission. Others reset monthly. That clause is worth more than the headline percentage and is frequently not mentioned until the contract arrives.
Now the part that is rarely stated plainly. Under revenue share, you are paid in proportion to how much money your subscribers lose. A subscriber who follows your picks and wins reduces your affiliate income. A subscriber who chases losses, bets parlays and reloads increases it. The subscription and the affiliate stream pull in opposite directions, and the affiliate stream is usually the larger one.
That conflict is not hypothetical. It shapes product decisions: whether to push high-hold parlays, whether to encourage line shopping across books that do not pay you, whether to promote deposit bonuses carrying wagering requirements. A service genuinely trying to improve its subscribers' results will make less affiliate money than one that is not. Anyone claiming both streams are aligned has not read their own contract.
If you take affiliate money, three things follow. Disclose it clearly and conspicuously on every placement, as the FTC Endorsement Guides require. Check whether your state requires registration to be paid on a commission basis for sportsbook referrals. And accept that the contract can be terminated at the operator's discretion, because in Great Britain the operator carries the regulatory liability for your marketing under SR code 1.1.2, and elsewhere the commercial incentive runs the same way.
Tax: Different Rules, Same Mistake
The consistent error is treating the picks business like gambling. It is not. In most jurisdictions betting winnings and trading income are taxed differently, and selling subscriptions is unambiguously trading income.
- United States. Payment apps and online marketplaces must file Form 1099-K where payments for goods and services exceed 20,000 dollars in more than 200 transactions, although they may report lower amounts. If customers pay you directly by credit, debit or gift card, the IRS states your card processor files a 1099-K "no matter how many payments you got or how much they were for". Self-employment income also carries self-employment tax on top of income tax.
- United Kingdom. There is a 1,000 pound trading allowance per tax year; gross trading income at or below that generally need not be reported, subject to exceptions where Self Assessment registration is still required. Above it, Self Assessment applies. VAT registration is compulsory once taxable turnover in the previous 12 months exceeds 90,000 pounds, or if you expect to exceed it in the next 30 days. Note the asymmetry: a punter's winnings are not taxed in the UK, but your subscription revenue is. Selling the tips and following them are taxed completely differently.
- Australia. GST registration is required once GST turnover reaches 75,000 Australian dollars or more.
- Canada. You stop being a small supplier and must register for GST/HST once taxable revenues exceed 30,000 Canadian dollars across four consecutive calendar quarters.
Cross-border digital subscriptions add another layer, because consumption taxes on digital services are generally due where the customer is rather than where you are, and selling to consumers in several countries can create registration obligations in each. Check locally. This is a paragraph in a guide, not tax advice.
Discord hosts most of these communities, and the reasons servers get actioned are usually not the ones people expect.
Discord's Community Guidelines rule 26 says "Do not coordinate or participate in illegal gambling", and its Gambling Policy Explainer defines a violation as requiring all four of: a payment or wager of real-world value to participate; prizes of real-world value awarded; an outcome predominantly determined by chance; and the activity being prohibited by applicable gaming law. A subscription picks channel does not meet that test, because there is no wager to participate and no prize.
Rule 25, "Do not organize, promote, or engage in the selling or facilitation of sales of regulated or potentially dangerous goods", is a closer fit where the server functions as a sales channel for a regulated service.
In practice, removals in this category follow from mass direct-message promotion, bought engagement, scam reports from members, and the payment platform pulling out first. The gambling policy is rarely the mechanism. The consequence is the same either way: a Discord community is a tenancy, not an asset. Keep your subscriber list somewhere you own it, as email addresses collected with consent, so that a server removal is a disruption rather than a closure.
What a Defensible Service Looks Like
If you are going to do this, the version that survives scrutiny has a specific shape:
- One sport, one bet type, one staking rule, published in advance. Narrow scope makes the record legible and the edge testable.
- Every pick timestamped before the event, with the price and the stake, on a third-party platform. Publish closing line value alongside win rate. Never delete a pick.
- The full record, including the losing months, permanently public. A record that begins at your best month is a marketing asset, not a record.
- No income claims, ever. No "quit your job", no "replace your salary", no betting slips presented as typical outcomes. In Great Britain that is rule 16.3.4. In the US it is the Endorsement Guides and the reviews rule. In Australia the direction of travel is a prohibition on influencer promotion of wagering altogether.
- Clear renewal terms, easy cancellation, fast refunds. This is chargeback management, and chargeback management is survival.
- Affiliate relationships disclosed on every placement, or not entered into at all.
- Responsible gambling information on the page and a hard age gate, at 18 or 21 depending on jurisdiction.
- An email list you own, so that no single platform decision can end the business.
That service will grow more slowly than a competitor making income claims. It will also still exist in two years.
Who should skip this
Be direct about it.
Do not do this if you cannot show several thousand graded, timestamped, third-party-verified picks. Below that sample size your record is statistically indistinguishable from luck, and selling it as skill is precisely the conduct the FTC's testimonials rule and the CAP Code's social responsibility provisions exist to address.
Do not do this if the marketing you have in mind involves income claims. The single most effective message in this category, that betting can replace employment, is explicitly prohibited by CAP rule 16.3.4 in Great Britain, sits squarely inside the FTC's deceptive-claims remit in the US, and is described almost verbatim in Stripe's prohibited-business list. If that message is the plan, you do not have a business, you have an enforcement timeline.
Do not do this if you need the income to be stable. Churn here is outcome-driven and seasonal, and you will spend most of your working time replacing subscribers rather than serving them.
Do not do this if you intend to pool subscriber money. The SEC has litigated it repeatedly. Selling information and managing a bankroll are different businesses under different legal regimes, and blending them converts a lawful subscription into an unregistered securities offering.
Do not do this if you are uncomfortable being paid more when your customers lose more. Under a revenue-share affiliate deal, that is the arrangement. If you would not say it out loud to a subscriber, do not sign it.
Do not do this if your audience skews young. The under-18 protections in the CAP Code, Ontario's Standard 2.03, Meta's and Google's advertising policies and the Australian Bill all converge on the same point, and a following built on youth-appeal content cannot be insured against them.
Do not do this if you were told it was passive. It is a seven-day-a-week content job in season, plus moderation, plus support, plus the personal bankroll risk of maintaining a record worth showing.
The operators who make this work over years share three traits: a genuinely narrow specialism, a verified record long enough to be boring, and marketing so conservative it reads as an underclaim. Everything else in this category is variance being sold as skill, and variance eventually presents its bill to whoever is holding the subscription.