Ten minutes to kickoff. You are watching a line move on a page you built yourself, and somewhere a bettor who found your tool is watching the same thing. He decides to pay you. That small moment is the whole business in miniature.
If you code, you have probably felt the squeeze lately. Contract work dries up. Hiring managers want ten years of experience in a framework that is five years old. AI tools write the boilerplate you used to bill for. You want a product of your own, something that earns while you sleep and answers to nobody's sprint board.
Sports data is a hard place to do that, and the guide is honest about why: the leagues own the data, the feeds are expensive, and bettors who win get limited and cancel. The opening right now is in narrow, cheap tools built on free or low tiers, while prediction markets reshuffle what data people need. The builder who learns the cost structure first can move when a gap appears. The one who waits has to learn it under pressure.
SportsGameOdds publishes a free tier, and third-party comparisons in mid-2026 put its paid tiers at 99 to 499 dollars a month, billed on objects returned. Start-up costs here run from $100 to $2,000, profit typically arrives two to four months in, and monthly income runs from $1,000 to $30,000. Plan your life around the bottom of that range.
Tonight, sign up for the free tier, pull the odds for a single match, and count the objects that come back. That count is your first real cost number.
Before you get excited, look at the two biggest companies in this world, because both publish their accounts. In 2025 Sportradar reported total revenue of 1,290 million euros and profit for the period of 100 million euros, or 7.8% of revenue. Its sport rights costs in the second quarter alone were 106.2 million euros against 318 million euros of quarterly revenue, and 210.2 million euros across the first six months. That works out at roughly a third of its revenue going straight to rights holders. Genius Sports reported 2025 group revenue of 669.5 million dollars against cost of revenue of 515.6 million dollars, and a net loss of 111.6 million dollars, wider than its 63.0 million dollar loss in 2024.
I want to be fair with you about those figures, so here are two caveats. Sportradar's sport rights line is a specific disclosed cost, and it runs at around a third of revenue. Genius Sports does not break out data rights separately. Its 515.6 million dollar cost of revenue includes rights payments alongside other direct costs, so 77% is the ceiling of what rights could be. The true rights figure sits somewhere below it.
Even if you read them as kindly as possible, the picture stays the same. These are the incumbents. They hold the exclusive contracts, the scale, the sportsbook customers and the public listings. One of them still hands roughly a third of its revenue straight to rights holders, and the other is still losing money on nearly 670 million dollars of turnover.
Keep that in your head before you write a single line of code. It is the ground you will be building on.
If you only read one story about how sports data rights can sink a company, make it IMG Arena. Endeavor Group built a betting data business on a stack of exclusive rights deals with sports federations. In March 2025 it agreed to give the entire business to Sportradar. Endeavor handed it over and paid for the privilege. Under the announced structure Sportradar paid no consideration at all. Endeavor paid a total of 225 million dollars instead, made up of 125 million dollars in cash to Sportradar and up to 100 million dollars in prepayments to certain rights holders. The deal closed on 3 November 2025.
The reporting at the time was blunt about why. Endeavor could not find a buyer willing to take on the rights commitments, so it chose to pay its way out of them rather than keep funding them. Before completion, IMG Arena was also asked to exit a loss-making European Leagues Association agreement.
Sit with that for a moment. A company inside one of the largest sports agencies in the world signed multi-year data rights contracts, could not make them pay, and spent a quarter of a billion dollars to be set free. If a business that size had to pay 225 million dollars to walk away, what would a multi-year data commitment do to your own household budget? That is the honest base rate you are working against if your plan involves buying sports data rights. Start small, keep your commitments monthly, and let paying users earn you the right to sign anything longer.
You have probably heard that scores and statistics are free, because facts cannot be copyrighted. The first half of that is broadly right in the United States. The second half does not follow, because the facts were never what was being sold. What buyers pay for is speed, completeness, contractual certainty and the right to say the word "official".
It helps to pull apart three separate things that get licensed, because people mix them up all the time:
The commercial structure that has grown up around all this is exclusive, long, and expensive. Here is who holds what.
Here is what that means for you as a small operator. It is simple, and a little sobering. For the major North American and European leagues, there is no window in which you can buy official low-latency data directly from the league. The league has already sold it, exclusively, to a counterparty with a nine or ten-figure balance sheet. Your only routes are a sublicence from that counterparty, a downstream feed with delay, aggregated bookmaker odds in place of event data, or unofficial collection.
Which of those three layers are you actually planning to sell? Write the answer down now, because it decides almost every cost and legal question in the rest of this page.
The leagues have more than commercial muscle on their side. Several US states have written official league data into law for certain wager types, which turns a commercial preference into a legal requirement for licensed operators.
The Kansas Legislative Research Department listed six states requiring official league data for proposition bets, in-play wagers, or both: Arizona, Illinois, Michigan, North Carolina, Tennessee and Virginia. Sportshandle's coverage lists a wider set including Kansas, Massachusetts and New York. Tennessee's statute requires licensed operators to use official league data exclusively for live betting, subject to an exception. Sources differ on the precise current list and on the scope of each mandate, and the mandates have been amended since they were first enacted. So check the current statute and regulator guidance in each state you care about rather than relying on any list, including this one.
This matters to you even if you never plan to sell to a sportsbook, because mandates guarantee the leagues a captive institutional buyer for the official feed, and that sets the floor price for the whole market. Every downstream product, yours included, is priced above a floor that a regulator helped build. You will feel it in every quote you get.
This is where most independent projects live, and where most founders misjudge their exposure. Here is the honest summary. Scraping sports data is rarely a criminal computer-misuse problem in the US, frequently a contract problem everywhere, and a serious intellectual property problem in the UK and EU.
Have you actually read the terms of service on the sites you plan to pull from? Most people skip them, and the cases below show why you should open them tonight.
The Computer Fraud and Abuse Act is the statute people worry about. In Van Buren v United States (2021) the Supreme Court read the "exceeds authorized access" provision narrowly, treating it as a gates-up-or-down question. In hiQ Labs v LinkedIn the Ninth Circuit held, twice (most recently in April 2022 after remand in light of Van Buren), that automated collection from publicly accessible pages that need no account does not amount to access "without authorization" under the CFAA.
So far, so encouraging for you. Then look at what happened next.
In November 2022 the district court held that provisions of a website user agreement prohibiting data scraping and the creation of fake profiles are enforceable as a matter of contract. The parties then entered a consent judgment: 500,000 dollars against hiQ, findings of liability under California common law trespass to chattels and misappropriation, and injunctive relief that effectively ended hiQ's ability to scrape LinkedIn at all.
hiQ won the big constitutional argument and lost the business anyway. That is the right picture to carry for how this usually goes. Your main risk sits in the terms of service you clicked through, and in the tort claims that survive alongside them. The CFAA comes a long way behind. Note too that the consent judgment is a stipulation with no judicial finding behind it, so it carries no precedential weight. It tells you about commercial outcomes. It tells you nothing about what a future court must hold.
Two decisions matter here, and both help you as a builder, within limits.
In NBA v Motorola (2d Cir 1997) the court held that professional basketball games are not original works of authorship protected by the Copyright Act, that the NBA's state-law misappropriation claim was preempted except for a narrow "hot news" category, and that sending real-time scores and statistics to pagers was not misappropriation. In C.B.C. Distribution and Marketing v Major League Baseball Advanced Media (8th Cir 2007, certiorari denied 2008) the court held that although using players' names and statistics for profit touched Missouri right-of-publicity law, the First Amendment overrode those rights in the context of fantasy baseball products.
US-focused writing about sports data almost always leaves this part out, and it is the part most likely to cause you real trouble if you are a British or European founder.
Directive 96/9/EC created a sui generis database right, separate from copyright, that protects substantial investment in obtaining, verifying or presenting the contents of a database. The UK put it into law through the Copyright and Rights in Databases Regulations 1997 and kept a domestic version after leaving the EU.
Two cases mark out the boundary for you.
In British Horseracing Board v William Hill (Case C-203/02, 9 November 2004) the Court of Justice held that "investment in the obtaining of the contents" means resources used to seek out existing independent materials and collect them, and leaves out resources spent creating the materials in the first place. BHB's spending on selecting horses, checking eligibility and drawing up race lists was investment in creating data, which falls outside obtaining it, and so it did not attract the sui generis right. Fixture lists and entry lists generated by the organiser of an event therefore sit outside the right.
Then came the Football DataCo litigation against Stan James and Sportradar. The Court of Appeal of England and Wales held that database right did subsist in the Football Live database of live match events, and that Sportradar and Stan James were jointly liable for the acts of UK users who reached the data through a link on Stan James' website. Live match-event data collected by observers counts as obtained, because the observers record events that already happened. And so it is protected.
This distinction will shape your product. Fixtures and schedules: probably unprotected. Live in-play event data compiled by human or automated observers: likely protected in the UK and EU, with no US equivalent.
Jurisdiction was settled in a separate case. In Football Dataco v Sportradar (Case C-173/11, 18 October 2012) the Court of Justice held that sending data from a server in one member state to a user in another counts as re-utilisation, and that the act takes place in the country where the alleged infringer intended to target the public. So hosting your servers outside the UK or EU will not put you outside the right if you are marketing to British or European users.
The third route is sending someone to the ground with a phone. You might think that is the clever workaround. It has been fought all the way to a conclusion, and the conclusion is unhelpful for small operators like you.
Football DataCo and Genius Sports sued Sportradar and, unusually, individual data scouts as representative defendants. The claims were breach of confidence and trade secrets, breach of contract (the ground regulations and ticket conditions), trespass to land, inducement of breach of contract, and conspiracy to cause loss by unlawful means. Sportradar counterclaimed on competition law grounds, alleging abuse of dominance in the award of the exclusive contract.
It settled in 2022. Sportradar agreed to stop unofficial in-stadia scouting of Premier League and Football League matches, and took a sublicence from Genius for a delayed secondary feed running to 2024. Genius kept the exclusive right to provide low-latency official Football DataCo betting data through 2024. Football DataCo kept the right to license its data as it saw fit in future. The financial terms were confidential.
Think about who lost there. Sportradar is a listed company with over a billion euros of revenue and a serious competition-law argument. It lost the ability to collect English football data unofficially and ended up as a customer of its competitor, on a delayed feed. If that is how it went for Sportradar, a two-person startup sending a mate with a phone stands no chance.
The practical rule
Take away the case names and your working rule is short:
- Historical and post-match statistics from public sources carry low legal risk in the US and moderate risk in the UK and EU, and they have low commercial value.
- Live in-play event data is the valuable product and is the most protected, by database right in the UK and EU, by exclusive contract everywhere, and by ground regulations at the venue.
- Bookmaker odds are a different asset again. They are the bookmaker's own commercial output. Sportsbook terms of use routinely ban automated access, scraping and commercial reuse. You will not be prosecuted, but you can be blocked, sued in contract, and cut off from the affiliate relationship that may be paying your bills.
- Whatever the law says, the other side usually answers scraping with technical and commercial moves before it reaches for lawyers: rate limits, IP bans, feed changes and account closure. If a competitor's engineer can switch off your supply chain on a Tuesday afternoon, you do not yet have a durable business.
What Data Actually Costs, at Every Tier
So now you know who controls the feed and on what terms. Next comes the number that lands on your card every month.
Here is the price ladder as published in mid-2026. Check each of these against the vendor's own page before you build a model on them, because API pricing changes without notice.
The Odds API. Free tier at 500 credits per month covering most bookmakers. Paid tiers at 30 dollars per month for 20,000 credits, 59 dollars for 100,000, 119 dollars for 5 million and 249 dollars for 15 million. Watch the unit: a credit and a request are different things. One request returns live and upcoming games for a given sport, market and bookmaker region, and the credit cost scales with the number of markets and regions you query.
API-Football and the wider API-Sports family. A free plan at 100 requests per day, with paid plans starting at 19 dollars per month. Third-party comparison sites report a top self-service tier around 39 dollars per month for 150,000 requests. Treat those higher-tier numbers as third-party reporting until you confirm them yourself.
SportsGameOdds. A free tier plus paid tiers reported by third-party comparisons in the range of 99 to 499 dollars per month, billed on objects returned rather than requests.
OpticOdds. No published pricing. The pricing page is a multi-step quote request form that asks for your company size and the services you need before anyone will talk numbers with you. So no entry price is quoted here. Figures float around betting communities, but nothing could be confirmed against a primary source, and a number you are told not to budget with is worse for you than no number at all. Assume low thousands per month is possible, and get a real quote before you plan around it. Additional sports are charged separately, but none of those figures could be confirmed against a primary source and none should be used for budgeting.
Sportradar and Genius Sports. No public price list at all. Sportradar's developer portal offers trial access through its marketplace and says plainly that its APIs are a business-to-business service, built to sit behind your servers and never to be called directly from a client application. Genius Sports sells through commercial agreements. Enterprise sports data contracts in this tier are negotiated, typically annual, typically with minimum commitments, and typically include limits on redistribution and on serving end consumers. No list price is published, and no verifiable figure for what a small buyer pays could be found. If someone quotes you a confident number for what Sportradar charges a startup, they are guessing.
The gap between the two halves of that ladder is the single most important economic fact you will face in this business. On one side there is a self-service tier at 20 to 250 dollars a month that gives you aggregated bookmaker odds and basic fixture and result data. On the other there is an enterprise tier for official low-latency event data that starts at a negotiated annual contract. There is very little in between, and the products bettors actually pay for tend to sit on the wrong side of the gap.
How API pricing bites in practice
The headline price rarely hurts you. The multiplication does.
Picture a positive-expected-value tool covering four sports, three betting markets per sport, and two bookmaker regions, refreshing every sixty seconds during the active window. The billing unit on most odds APIs multiplies by markets and regions. That gives you 4 x 3 x 2 = 24 billable units per refresh cycle. At one refresh a minute, that is 1,440 units an hour, 34,560 a day, and just over a million a month if you run continuously. On a 5 million credit plan at 119 dollars a month you are fine. Now add player props (which can mean dozens of markets per game in place of three) and refresh every fifteen seconds because line movement is the product, and you multiply that by something between twenty and eighty. Suddenly you are shopping for the 249 dollar plan, and then you are quickly past it.
Here is the general shape to remember: cost scales with markets multiplied by books multiplied by refresh frequency, and all three of those are exactly the dimensions along which your product competes. Every feature request from a serious user makes your gross margin worse. That is a strange and unpleasant property for a software business, and it is the specific reason data bills kill these projects long before infrastructure bills do.
So how many markets and regions does your first version really need? Be strict with yourself here, because every extra one multiplies straight into your monthly bill.
Two habits are worth building in from day one. First, adaptive polling: refresh often only for events starting soon or with volatile lines, and back off hard everywhere else. Second, fan-out separation: poll the upstream API once and serve all your users from your own cache. Never let your user count drive upstream calls. If your architecture makes an upstream request per user page load, your marginal cost per user becomes an API bill instead of a hosting bill, and no consumer price point will save you.
Infrastructure Cost Per User: The Part That Is Genuinely Cheap
You have seen that data is expensive. Now for some good news, stated precisely, about what costs you very little.
Published rates as of mid-2026:
- Supabase. Free tier: 500 MB database, 5 GB egress. Pro: 25 dollars per month, 8 GB database, 250 GB egress included, then 0.09 dollars per GB egress, 0.03 dollars per GB cached egress, 0.125 dollars per GB additional database storage. Team: 599 dollars per month.
- Vercel. Fast Data Transfer: first 1 TB included, then 0.15 to 0.35 dollars per GB depending on region. Edge Requests: first 10 million included, then 2.00 to 3.20 dollars per million. Additional team seats 20 dollars per month.
Now model a real-time odds product with me. Say a logged-in user with the board open receives 40 KB of JSON per minute over a websocket or polling connection, and spends 90 minutes a day in your product. That is 3.6 MB a day, roughly 108 MB a month of egress per active user. At Supabase Pro's 0.09 dollars per GB overage, that comes to under one cent per user per month in data transfer. A thousand daily-active users is roughly 108 GB, comfortably inside the 250 GB included allowance.
Compute is just as forgiving, as long as you have separated fan-out from upstream polling. One ingestion worker keeping a normalised odds table up to date, and a read path served from cache, will run on a small instance for tens of dollars a month at this scale.
So a plausible fully loaded infrastructure cost for a thousand active users lands in the region of 50 to 150 dollars a month. That range is my estimate worked out from the published rates above, and you should treat it as unsourced. Set it beside an odds API bill of 119 to 249 dollars a month, plus any premium feed, plus payment processing at typical card rates, and the shape becomes obvious: at small scale your data bill exceeds your infrastructure bill by a wide margin, and infrastructure only starts to matter if you get the polling architecture wrong.
Is your current design making one upstream call every time a user loads a page? If it is, fix that before anything else, because that single choice decides whether your cost per user is a fraction of a cent or an API credit.
The costs that really do scale badly are the ones people forget: websocket connection concurrency if you use a managed realtime service billed per connection, egress if you push the full board state instead of deltas, and background job minutes if you recompute models on every tick instead of on change.
Cheap hosting means something real at your kitchen table. With Supabase's free tier and its 25 dollar Pro plan carrying your early users, your outlay can stay small enough that a slow first season leaves the rent alone, as long as you cap data and the other costs too. Keep it that lean until paying users ask you for more.
B2C Pricing: What Bettors Actually Pay
That maps out your cost side. What follows decides whether anyone hands you enough money to cover it.
You can see real prices without guessing, because subscription tiers are disclosed on app store listings.
The OddsJam iOS listing shows in-app subscription tiers including Trends at 19.99 dollars per month, Fantasy Optimizer at 59.99 dollars, Sharp Money and Gold at 199.99 dollars, Positive EV Global at 399.99 dollars, and Positive EV Monthly and Platinum at 499.99 dollars, with a seven-day free trial. The listing carries an 18+ age rating, a "Frequent Simulated Gambling" content descriptor, and a problem-gambling helpline notice.
That is the ceiling of the consumer market, and it tells you something specific: the high price points are attached to the features that need the most expensive data. Arbitrage, positive-EV and middling tools need wide bookmaker coverage refreshed fast. The 19.99 dollar tier is trends and historical data, which is cheap to source. The 499.99 dollar tier is the live cross-book board, which costs a great deal to feed.
Three things follow for you if you come in underneath.
You cannot win on price alone. Undercutting a 199 dollar product with a 49 dollar product only works if your data costs less, and if your data costs less, your product is worse at the very thing users are paying for. A leaner cost structure gives you little edge in this market, because the biggest cost is a licence fee that does not care how efficient you are.
Your market is small and self-selecting. The people who will pay you 200 dollars a month for a betting tool are people betting enough volume that a small percentage edge beats 200 dollars a month. That is a narrow group. They talk to each other, compare tools obsessively, and cancel without sentiment. Do you personally know ten people who bet at that level? If you cannot name them, your first job is finding where they gather before you write any more code.
Free tiers cost you real money here. In most SaaS, a free tier costs you hosting. In this one, a free tier that shows live odds costs you API credits per refresh, per market, per region. Build your free tier around delayed or aggregated data, or around historical analysis, and keep the live board behind the paywall.
There is no reliable public data on what independent sports analytics tools earn. Private companies in this niche do not file revenue publicly, and the numbers that float around (annual recurring revenue claims in forum posts, founder interviews, launch-day screenshots) are self-reported, unaudited and picked because they look impressive. There is no denominator anywhere: nobody publishes how many people built a betting tool and how many of those are still trading two years later.
What you can learn from audited filings is narrow but useful. The two largest firms in the sector publish their accounts, and one of them lost 111.6 million dollars in 2025 on 669.5 million dollars of revenue. That tells you nothing about what a solo product earns. It does tell you that scale, exclusive rights and sportsbook customers are not enough on their own to make money here. Treat any specific income figure attached to this niche, including any you find elsewhere on the internet, as unverifiable unless it comes from a filing. Protect yourself from the screenshot that makes you quit your job too early.
Keep the honest version of success in view. A small tool with a few loyal subscribers could cover your hosting, your coffee habit and the streaming bill, and give you the quiet pride of a product with your name on it. That is enough to start, and it is far safer than quitting on a screenshot.
The Churn Problem Nobody Mentions: Your Best Customers Get Limited
This is the problem that sets betting tools apart from ordinary analytics SaaS, and it gets its own section because you will not see it coming until it shows up in your cohort data.
If your tool works, your users win. If your users win, sportsbooks restrict their accounts. Once they are cut down to trivial stake sizes, they cannot act on your signals any more, and they cancel. Your success becomes the reason they leave.
This is documented. The Massachusetts Gaming Commission opened an inquiry into the practice in July 2023 and held a public roundtable in May 2024; reporting at the time noted that the state's active licensed operators did not attend. A further session followed in September 2024. Operator data later supplied to the Commission showed that account restrictions affected around 0.64% of accounts, and that above-average winnings were linked to a higher likelihood of restriction. In December 2025 the Commission approved a regulation, effective 1 June 2026, requiring sportsbooks to notify customers within 48 hours of restricting an account and to give a specific, individualised reason in place of a boilerplate one, applied retroactively to accounts limited before that date.
Read that 0.64% figure carefully, because it is easy to take as reassuring. It is 0.64% of all accounts, and the overwhelming majority of those are casual losing accounts that no book has any reason to touch. Among the people your product exists to serve (bettors placing volume and beating the closing line) the rate is necessarily far higher. The Commission's own finding was that winning is linked to restriction.
So your business model has a ceiling built in: the better your product works for a given customer, the shorter that customer's useful life. Plan for it on purpose. What does your product still do for a customer the week after their account gets limited? The products that survive this tend to broaden into bankroll tracking, closing-line-value analysis, and multi-account management, which keep their value after limitation. Pure live arbitrage and positive-EV alerts lose theirs.
Watch the jurisdiction, too. The Massachusetts rule is a Massachusetts rule. Great Britain, Australia and most other markets have their own, different approaches to how much freedom operators have over stake limits. Do not assume any of this carries over to where your customers live.
B2B: Selling to Operators Instead of Bettors
The other path is to stop selling tools to bettors and sell technology to the industry instead: odds screens for media companies, pricing or risk tools for smaller operators, model outputs for traders, or data visualisation for broadcasters.
The economics get better for you in some ways and much worse in others.
Better: contract values are larger, churn is set by contract terms and has nothing to do with a customer's mood after a losing weekend, and you are out of the limiting problem entirely. Worse: your customers are licensed entities, and doing business with them pulls you inside their regulatory perimeter.
Here is the concrete version, using Tennessee as the example because its fee schedule is published in statute and regulation. Under Tennessee's sports gaming rules, vendors must register, and the fee for the first three-year registration period is 150,000 dollars, payable in instalments. Renewal fees are then banded by vendor category: 150,000 dollars for geolocation services, online sports wagering platform software or player account management; 75,000 dollars for risk management services, oddsmaking services, or sports data feeds or software; 45,000 dollars for independent testing laboratories; and 30,000 dollars for KYC or payment processing, vendors with direct interface to patron accounts or platforms, and vendors paid under revenue-share arrangements.
Read that middle band again, slowly. "Sports data feeds or software" is named outright as a registrable vendor category with a five-figure renewal fee, in one state, on top of a six-figure initial registration. Colorado runs a similar structure, with a Vendor Major licence required for entities that manage, administer or control wagers on a sports betting system. Every state has its own scheme, its own fees and its own definitions.
B2B stays possible for you. The catch is this: the moment your software touches a licensed operator's wagering process, you inherit a licensing cost structure that assumes you are a company with a compliance department. Selling read-only analytics to a media company is a very different proposition from selling anything that touches pricing or wager acceptance. Does your software only show numbers to readers, or does it feed into how a book sets prices and takes bets? Know which side of that line you sit on before you sign your first contract, and get local advice, because every jurisdiction draws the line differently.
A single B2B contract for a media company's odds screen behaves very differently from a few hundred app subscribers. Suppose one brings in $1,000 a month, renews, and leaves a surplus. Part of that one invoice could pay for your children's after-school clubs every month. Know which side of the regulatory line your product sits on first, and keep your family budget independent of the next renewal, so a lost contract never reaches the kids.
The Affiliate Route and What It Costs
Many consumer betting tools are really affiliate businesses wearing a software front end. They earn cost-per-acquisition or revenue share for sending signed-up depositors to sportsbooks, and the subscription is almost a side line.
That shifts your regulatory position again, and the position varies enormously depending on where you operate.
In the United States, several states require affiliates to register or be licensed. Industry sources report cost-per-acquisition affiliate registration fees of nothing in New Jersey, 200 dollars in Michigan, 500 dollars in Tennessee, Virginia, Colorado and Indiana, 1,000 dollars in West Virginia and 2,500 dollars in Pennsylvania. These figures come from law firms and industry commentators rather than from the regulators' own fee schedules, and at least one of them sits awkwardly beside Tennessee's statutory vendor registration fee of 150,000 dollars for the first three-year period. The likeliest explanation is that affiliates fall into a separate and cheaper registration category from technology vendors, but that could not be confirmed from a primary source. Check with the specific regulator before you budget either number.
In Great Britain, affiliates are not licensed by the Gambling Commission themselves. The obligation is pushed onto the operator through the Licence Conditions and Codes of Practice. Social responsibility code provision 1.1.2 makes licensees responsible for the actions of third parties they contract with for any aspect of the licensed business, requires contract terms obliging the third party to behave as if bound by the same licence conditions and codes of practice, requires the third party to supply information the licensee needs for regulatory reporting, and requires the licensee to be able to terminate promptly where an affiliate has breached a relevant advertising code. Condition 1.1.3 adds requirements for third parties providing user interfaces through which customers access remote gambling facilities.
In practice, a British operator will push its entire compliance burden onto you by contract, audit you, and drop you without much argument if a regulator or the Advertising Standards Authority takes an interest in something you published. You carry the compliance cost and get none of the licence's protections.
Two more points apply everywhere. Affiliate income is concentrated and volatile. Deals get renegotiated, revenue-share tails are cut when operators change platform, and a single regulatory change in a single market can wipe out a meaningful share of your revenue overnight. And affiliate revenue conflicts directly with your product promise. A tool that tells users which book has the best price is structurally at odds with a deal that pays you more for sending them to one particular book. Could you look a user in the eye and explain why your tool pointed them at that book? People in this niche are unusually good at sensing the answer, and once your reputation goes, it does not come back.
Distribution Risk: The App Stores Can Say No
Everything so far assumes people can reach your product. Getting it in front of them is a separate approval, and someone else grants it.
If your plan includes a mobile app, read the store policies before you build.
Apple. App Review Guideline 5.3.4 requires that apps offering real money gaming or lotteries have the necessary licensing and permissions in the locations where the app is used, be geo-restricted to those locations, and be free on the App Store. It also says plainly that "illegal gambling aids, including card counters, are not permitted on the App Store." Guideline 5.3.3 prohibits using in-app purchase to buy credit or currency for real money gaming. Guideline 5.3 opens by warning that this category is one of the most regulated on the store and that you should expect extra review time.
An odds and analytics tool does not itself offer real money gaming, so 5.3.4 does not automatically apply to you. But "gambling aids" is the review category you sit closest to, and the outcome rests on a reviewer's judgement about whether your app assists gambling that is legal where it is used. You cannot settle that question with certainty in advance.
Google. The Real-Money Gambling, Games and Contests policy is more explicit and more restrictive. It requires approved gambling apps to be published by an approved governmental operator or a licensed operator holding a valid operating licence in each country where the app is distributed. Critically for you, it prohibits apps that provide gambling support or companion functionality. The policy names "functionality that assists with wagering, payouts, sports score/odds/performance tracking". And it gives as an explicit violation example "a dedicated sports odds tracker app containing integrated gambling ads linking to a sports betting site."
That example describes, almost word for word, the default business model for a consumer sports analytics app. Assume mobile distribution through Google Play is difficult to impossible for an ad-and-affiliate-funded odds tracker, and plan for web-first delivery, progressive web app installation, or a materially different product scope.
That hits your customer acquisition hard. You lose store search, store-driven install flows, and the trust signal of being in the store. You have to win every user through channels you own or rent. If Google Play turned you down tomorrow, where would your next hundred users come from? Name that channel now, and start building it alongside the product.
Payment Processing: The Other Switch Someone Else Controls
Stripe's list of restricted businesses prohibits, among other things, "Games of chance including gambling, internet gambling, casino games, sweepstakes and contests, and fantasy sports leagues with a monetary or material prize"; "Sports forecasting or odds-making with a monetary or material prize"; and, in its Japan-specific restrictions, "Consultation and advisory or prediction services relating to online gaming or gambling". It separately prohibits "get rich quick" schemes, including "sites that promise fast and easy money".
A subscription analytics tool that does not itself take wagers or award prizes falls outside the first two prohibitions in a straightforward reading. But you are one marketing page away from the third and fourth. Landing copy promising guaranteed profit, risk-free arbitrage or a specific monthly return moves you from an adjacent category into a prohibited one, and enforcement skips the warning letter. You get an account review, a payout hold and a termination.
What protects you: keep your marketing copy factual and free of return promises; say clearly that the product provides information and does not accept wagers; set up a second processor relationship before you need it; keep a cash buffer big enough to cover a rolling reserve if one is imposed; and read the restricted business list for every country you sell into, because it varies by country, as the Japan clause shows. If your processor froze your payouts next month, how long could your savings keep paying the data bill? That number is the size of the buffer you should be building.
Prediction Markets: The 2026 Wildcard
Something is shifting underneath this whole sector, and it is worth your time to understand it even if you never touch it.
Sports event contracts on CFTC-regulated designated contract markets, offered by exchanges such as Kalshi, have created a parallel venue for something that looks a great deal like sports betting, regulated federally instead of state by state. The legal position in mid-2026 is genuinely unsettled, and the courts are split.
On 6 April 2026 a divided panel of the Third Circuit held that the CFTC has exclusive jurisdiction over sports-related event contracts, finding them to be "swaps" under the Commodity Exchange Act and holding New Jersey's gambling laws field- and conflict-preempted as applied to trading on a registered designated contract market. In January 2026 a Massachusetts Superior Court reached the opposite conclusion, rejecting the swaps label as "overly broad" and granting an injunction effective 8 March 2026. On 8 July 2026 a federal judge in Manhattan declined to block New York's regulators, holding the state's gambling laws not preempted as applied to sports contracts; New York's Attorney General sued Kalshi on 31 July 2026, alleging an illegal gambling operation. The CFTC Chair announced in January 2026 that the agency would go ahead with a prediction markets rulemaking.
Here is why this matters to you as a data product builder. If sports event contracts settle into a federally regulated national venue, the data needs change: settlement sources, resolution criteria, and cross-venue price comparison between exchange contracts and sportsbook odds all become products in their own right. It also opens a distribution channel that sits outside state-by-state gambling affiliate registration. Equally, if the state-law challenges succeed, that channel narrows sharply.
Do not build a business that depends on the outcome. Do keep watching it, because it is the most likely source of a genuinely new product category in this sector over the next two years.
New categories reward whoever is already standing nearby with working code. If exchange contracts settle into a stable venue, the builders with settlement and cross venue comparison tools half built will ship in weeks. Everyone else will spend those weeks reading documentation. Keep a prototype warm.
Tax and Structure
I will keep this brief, because it is dull, and because it is exactly where small SaaS businesses like yours get caught.
- UK. The VAT registration threshold is 90,000 pounds of taxable turnover in a rolling twelve months, or where you expect to exceed it within the next 30 days. Digital subscription services supplied to UK consumers are standard-rated.
- European Union. VAT on electronically supplied services to consumers is due where the customer is located, and for non-EU suppliers there is no registration threshold. The non-Union One Stop Shop scheme lets you register in a single member state and account for VAT across the bloc, but you still have to work out and evidence each customer's location. If you sell to EU consumers, this obligation starts with your very first sale, whatever your revenue.
- United States. Whether SaaS is subject to sales tax varies by state, and economic nexus rules mean you can pick up a filing obligation in a state where you have no physical presence once you cross that state's sales or transaction thresholds. There is no national rule. Get advice keyed to the states where your customers actually are.
- Everywhere. If any part of your revenue is affiliate commission from gambling operators, expect banks and processors to look harder at where your money comes from, and expect operators to ask for tax documentation before they pay you.
None of this is optional, and none of it is expensive to get right early. Getting it wrong for three years and then fixing it is what costs you.
A Realistic Cost Model
These are illustrative ranges built from the published prices above plus my reasoned estimates. None of the totals is a sourced figure. Treat every number as an estimate, and rebuild it with live quotes before you commit to anything.
Scenario one: historical analysis tool, no live data. Self-service statistics API at 19 to 39 dollars a month, hosting at 25 to 50 dollars, domain and email at 5 dollars, payment processing at roughly 3% of revenue. Total fixed cost under 100 dollars a month. You can genuinely bootstrap this. It is also the least defensible product, because the data is cheap for everyone, your competitors included, and its value to a bettor is modest.
Scenario two: live odds comparison, one region, moderate market coverage. Odds API at 119 to 249 dollars a month, hosting and database at 50 to 150 dollars, error monitoring and support tooling at 50 dollars, plus your time. Call it 250 to 500 dollars a month fixed before marketing. At a 49 dollar price point you need roughly ten paying subscribers to cover costs and perhaps forty to make it worth doing. Could you find forty bettors willing to pay 49 dollars every month, and keep them after their accounts are limited? That is achievable for a focused builder, and it is where most independent projects like yours should aim.
Scenario three: multi-region, player props, low latency, arbitrage detection. This is the product that competes with the 199 to 499 dollar tier, and it needs a data package that never appears on a self-service pricing page. Enterprise odds feeds in this class are negotiated annually with minimum commitments, and no confirmable figure for what such a contract costs could be found. Budget on the assumption that the annual data commitment is the largest line in your business and that you owe it whether or not you win customers, which is precisely the structure that destroyed IMG Arena.
Every week you spend planning the perfect multi region product is a week you could have spent learning what one paying user wants. Start with scenario one, keep the data bill tiny, and let real customers tell you what to build next. Open the free tier tonight and ship something small before the weekend.
Who should skip this
You have now seen the data bill, the price you can realistically ask, and the gates standing between you and a paying user. Time to hold all three up against your own life.
I will say it plainly, because you deserve a straight answer.
- Anyone who cannot absorb the data cost for twelve months without revenue. Once you leave the self-service tier, sports data contracts are commitments you owe in full, whatever your usage. If your plan needs month-three subscription revenue to fund the data bill, the plan does not work.
- Anyone building mainly on scraped live data. Prison is very unlikely for you. The real problems are that in the UK and EU it engages a database right that has already been enforced successfully against a listed competitor, and that everywhere it hands your supply chain to someone else to switch off.
- Anyone whose growth plan runs through mobile app stores. Google Play's policy names sports odds tracker apps with gambling ad links as a violation example. Apple's category is judgement-based and slow. If you have no viable web acquisition channel, you have no business.
- Anyone uncomfortable with the customer relationship. Your paying customers are gambling. Some share of any gambling population is being harmed by it. You will get support emails from people in distress, and if you are honest with yourself you will find some of your revenue comes from people who should not be spending it. If that troubles you, it should, and you would be wise to choose a different niche.
- Anyone who needs a stable regulatory environment. State mandates change, affiliate registration rules change, prediction market jurisdiction is being fought over in three courts at once, and league data contracts turn over on multi-year cycles that reshape the supply market when they do.
- Anyone without either a genuine data edge or a genuine distribution edge. Building the software has become the easy part. If your only asset is that you can build the app, you are competing on the cheapest dimension there is, against incumbents whose advantage is contracts you cannot sign.
Be honest with yourself: could your household carry the data bill for twelve months with no revenue coming in? If the answer is no, stay in scenario one until it is yes.
If You Are Still Going
If you have read this far and still want in, good. Here is a sequence that works with the economics.
Start where the data is cheap and the analysis is hard. Historical performance, closing line value tracking, bankroll analytics, model backtesting against archived data. These need statistics APIs at tens of dollars a month, far below live feeds at thousands. They also keep their value for your users after their sportsbook accounts are restricted, which tackles the churn problem head on.
Prove people will pay before you buy expensive data. A waiting list with deposits, or a manual service you deliver by spreadsheet, tells you more about demand than a finished product does, and costs you nothing in licence fees.
Separate ingestion from serving on day one. One process polls upstream and writes to your store; every user reads from your store. That is the difference between a marginal cost per user of fractions of a cent and a marginal cost per user of an API credit.
Measure your credit consumption before you measure anything else. Know, for each feature, how many billable units it burns per hour. When you later have to cut costs, you will need to know which feature to cut.
Read the actual terms of every feed you use. Redistribution rights, caching rights, whether you may serve the data to consumers at all, attribution requirements, and what happens on termination. Several self-service APIs restrict redistribution in ways that clash with the product people instinctively build.
Get advice for each jurisdiction before you take money from operators. Vendor registration, affiliate registration and advertising rules differ by state and by country, and the fee schedules are large enough to matter to a small business like yours.
Write your marketing copy as if a payment processor's risk team will read it, because one will. No guaranteed returns, no risk-free framing, no income claims.
Decide early whether you are a subscription business or an affiliate business. Trying to be both muddies your product promise and confuses your regulatory position. Pick one and be honest about it on your pricing page.
Picture the first stranger paying for a closing line value tracker you built on cheap historical data. Something shifts in how you describe yourself that day: you own a product, and its pricing page carries your name. Start there, where the analysis carries the value, and let money you have already collected decide when to pay for expensive live feeds.
What Would Change This Picture
Three things are worth your attention, and you control none of them.
Rights renewals. The current exclusive contracts have end dates. When a major league's data rights come up, the market briefly reopens. Nothing suggests the leagues will sell to small buyers like you, but the terms downstream distributors offer do move when they are competing for a renewal.
Prediction market resolution. If federal jurisdiction over sports event contracts is settled in favour of the exchanges, a large neighbouring market opens with different data needs and a different regulatory perimeter. If it goes the other way, the state-by-state structure hardens.
Regulatory attention to limiting. Massachusetts is the first state to require individualised justification for account restrictions, effective 1 June 2026. If other regulators follow, the useful life of your winning customers gets longer, and the unit economics of consumer betting tools improve materially. If they do not, the churn ceiling stays where it is.
Everything else comes down to how well you execute against a cost structure you do not control, in a market whose most valuable input has already been sold, exclusively, to someone else. Go in with your eyes open, keep your commitments small, and let your paying users tell you when to grow.