Cost to Build a Sports Betting App Like DraftKings or Bet365 in 2026

Someone quotes you $80,000 to build a betting app. Someone else quotes $900,000. Both are describing a sports betting app. Neither is lying.

The spread exists because a betting app is not one product. An app like DraftKings is really three products: daily fantasy, a real-money sportsbook, and a casino, each in a different regulatory tier. An app like Bet365 is defined by live, in-play betting priced in real time across dozens of jurisdictions. An app like Dream11 is a game of skill that survives ten million people locking lineups in the minutes before a match. An app like Polymarket is not a book at all. It is an exchange where users trade against each other, living in a derivatives lane, not a gaming one.

The title names the two brands everyone searches for, but this guide prices the whole family. What you are really pricing is a regulated financial platform, and the sports part is the thin visible layer on top. Get clear on which product you are building, and the cost stops being a mystery.

The cost, up front

Cost tracks the product model, not the brand you are copying. All figures below are software only.

The range inside each tier moves with the number of sports and markets, the depth of the compliance integrations, how much trading you rent vs own, and how many states you launch in.

A license is not engineering. These are separate budgets, and they are usually larger than the build.

The single most useful mental model: the software is 20-30% of what it takes to operate a real-money betting product. Price the build as the down payment, not the purchase.

According to the American Gaming Association, Americans wagered $166.94 billion on sports in 2025 and operators kept a record $16.96 billion, up 22.6% year over year. More than 30 states plus Washington, D.C. now permit some form of legal betting. Globally, the sports betting market is projected to reach roughly $187 billion by 2030, per Grand View Research, with online the fastest-growing segment.

Who actually builds a betting product

Most people building betting products are not trying to beat DraftKings or Bet365 head-on. They already own something expensive and want to monetize it. Five types of operators build custom.

Tribal and regional casinos launching their own brand. They already hold a license and want their brand, their floor loyalty program, and their own player data, not a white-label skin. The build question is about ownership, not permission.

Media and audience brands. A sports media company with millions of registered users owns the one thing incumbents pay $250-$750 to get: a warm audience. A custom product ties betting or fantasy accounts to existing profiles, keeps the data, and feeds behavior back into their CRM. A turnkey skin cannot do that.

International operators entering a new market. An established book from the UK, Europe, or Australia has trading expertise but needs a locally compliant build with the right geolocation, KYC, tax reporting, and market-specific rules. This is re-platforming plus compliance, not a from-scratch build.

Social and sweepstakes operators. A virtual-currency model operates in most states without a gaming license. It is the lowest-friction way to prove an audience and monetize before committing to real-money licensing. Many operators start here deliberately.

Fantasy-first operators in skill-game markets. Where fantasy is treated as a game of skill rather than gambling, operators launch contests without a gaming license, survive on contest fees instead of odds risk, and scale on match-time concurrency rather than trading. The regulatory tier is lighter, but the tax and data-licensing realities are unique to each market.

"These record revenues and tax contributions demonstrate the broad appeal of regulated gaming markets."

• Bill Miller, CEO, American Gaming Association, 2025 industry report.

The catch for a new entrant is the same regulation driving that legitimacy is what makes the market expensive to enter.

How a sportsbook thinks

Before you price a sportsbook, you need to understand how it thinks. Every expensive line item in the build exists to serve one of these four machines.

The odds engine: the book never needs to predict the game. Odds are prices, and the book is the shop that sets them. It starts with a model that estimates the true probability of each outcome, then adds its own cut. Here is the plain version: on a coin-flip-style market, both sides might be priced at -110, meaning you risk $110 to win $100. The implied chance on each side is about 52.4%, and the two sides add to about 104.8%. That extra 4.8% is the overround, the book's built-in margin, and on standard markets it works out to the book keeping roughly 5% of everything staked. This is the vig: the fee for arranging the bet, baked into the price so the bettor never sees a bill. The critical part is that the book does not need to know who will win. It sets the opening price, watches where the money goes, and moves the price to balance the action. If too much money lands on one side, that side gets more expensive and the other gets cheaper, until the money roughly evens out. A balanced book pays winners out of losers' stakes and keeps the vig either way. The odds engine is a money-balancing machine, not a prediction machine.

Risk management: the desk that watches the machine. Algorithms set the prices, but a risk team sets the boundaries. Every market carries exposure limits: how much the book can lose on a single outcome. Every player carries limits too. And the system watches for sharp action, bettors who win consistently enough that their bets are treated as information. When sharp money hits one side, the book moves the price faster than the algorithm alone would, because those bettors are usually right. Winning accounts get their limits cut; this is standard practice, not a defect. The risk team can also override the algorithm entirely: a star player limps off in warm-ups, weather turns, a suspicious pattern of bets lands from new accounts in one city. The override is a human decision, and the build must give the risk team the console to make it: pause a market, move a price manually, void suspicious bets. Almost no new operator runs this desk in-house on day one, which is why you rent trading and risk from a licensed provider (OpenBet, Kambi, Amelco, GR8 Tech). You are renting the risk team's judgment, not just their software.

Geolocation compliance: the bet legally happens somewhere. Here is the part that surprises every founder from outside the industry. A bet is not placed where the person lives, or where their account was created. It is placed where their phone was at the moment they tapped confirm. The law requires that place to be inside a licensed jurisdiction, and the app must prove it, fraud-resistantly, on every single bet. GPS alone is not enough, because GPS can be faked. The standard check layers GPS with Wi-Fi signals, IP data, and device integrity signals, and refuses the bet if the answer is uncertain. Regulators audit this the way banks audit transactions. What it forces into the architecture: the location check runs on every bet placement, not just at login; the app needs a clear, fast flow for the legitimate user standing near a state border who gets blocked; every state gets its own rules configuration; and testing happens on real devices near real borders, not in an emulator. Geolocation is a compliance layer wearing a map's clothing, and getting it wrong is a licensing event, not a bug report.

Real-time settlement: grading the bet the moment the game ends. When the final whistle blows, the book must decide who won and credit the wallet, in seconds for in-play, in minutes for pre-match. This is called grading, and the referee is the licensed data feed (Sportradar, Genius Sports, SportsDataIO). The feed declares the score, the engine applies the market rules, and the wallet moves. The hard part is the edge cases: matches abandoned at halftime, stats corrected after the fact, a tennis player retiring hurt. Each needs a written rule and a review queue, because the first disputed settlement handled badly becomes a support crisis and a regulatory complaint. Instant, trusted settlement is what makes a book feel like a book. Delays make it feel like a scam.

This is the teardown we run in discovery before we ever quote you. If a vendor cannot explain the odds engine, the risk desk, geolocation, and settlement for your market, they have not thought about your product yet.

Where the budget actually goes

Four builds plus the compliance layer that ties them together. Typical split for a single-state sportsbook:

The visible app is the minority of the work. Teams that budget for screens and discover the compliance layer mid-build are the ones that run over.

What drives the price up or down

Licensing and the regulatory timeline

This deserves its own section because it gates everything. Each US state licenses operators individually, and you usually need market access through a licensed casino or tribal partner before you can even apply. The engineering takes 24-36 weeks; the license and vendor certifications take 12-18 months. Run the license track from day one, in parallel with development. A team that finishes the build and then starts the license application has built a product that cannot legally take a bet. Fantasy and sweepstakes models exist precisely to dodge this timeline: they launch in months because they sit outside gaming licensing.

Geolocation compliance

Geolocation is a certified vendor integration (GeoComply is the category standard), not a map feature. The cost sits in integration, per-state rule configuration, and testing: the check must pass a regulator's audit, handle border-town edge cases gracefully, and never accept a bet it cannot verify. Multi-state launch multiplies the configuration work, because each state is its own ruleset. Under-budget this and you find out during certification, the most expensive possible time.

Live in-play betting

In-play is the Bet365 differentiator and the single biggest cost multiplier in the category. Pricing and re-pricing tens of thousands of markets in real time as a match unfolds, accepting bets against fast-moving odds, offering cash-out priced dynamically off live odds, and settling instantly demands low-latency infrastructure and real-time risk controls. The layer adds $150,000-$400,000 and 12-20 weeks, and it is the phase where latency and risk mistakes are most expensive. Retrofitting in-play onto an architecture designed for pre-match only is a rebuild, not an add-on. If live betting is the goal, the architecture anticipates it from V1 even if the feature ships in V2.

KYC, AML, and responsible gaming

You must verify identity, age, and source of funds, and screen against sanctions and exclusion lists. This is a certified onboarding flow tied to identity vendors, not a signup form. Responsible gaming rides on the same infrastructure: deposit limits the player sets, cool-off periods, self-exclusion with defined durations, and integration with state exclusion registers so an excluded player cannot re-register under a new email. These are legal requirements, not features, and they must be built into the first real-money release. A broken exclusion flow is a licensing risk, not a bug.

Sports-data feeds

Live scoring and settlement run on licensed data (Sportradar, Genius Sports, SportsDataIO), an ongoing monthly cost from hundreds to thousands of dollars depending on sport and coverage, starting before you have a single paying user. The feed is the referee for grading, so confirm the provider before the scoring and settlement engines are built: different providers use different data structures and update frequencies, and switching later means rebuilding.

One state vs many

Every state you add is a separate license, a separate compliance configuration, and separate tax reporting. Multi-state is a program of licensing and compliance work, and the software has to be built for that variation from the start: per-state rule configs, per-state reporting, per-state tax handling. Treat a second state as a new launch, not a configuration flag.

V1, V2, V3: what to build and when

The operators who ship phase the work. They prove an audience before taking on licensing, and they rent trading before building it.

How the timeline breaks down

Total engineering: 16-22 weeks for a fantasy MVP, 24-36 weeks for a single-state sportsbook. The end-to-end launch is gated by the license track, not the code.

The tech stack, without the sales pitch

Backend: Node.js or similar with PostgreSQL for the wallet ledger and bet records, Redis for live odds caching and session state. The ledger must be exact; this is a financial product.

Real-time layer: WebSockets for live odds, bet status, and settlement pushes. Polling collapses under real load and cannot serve in-play.

Odds and trading: OpenBet, Kambi, Amelco, or GR8 Tech for licensed odds and risk management. Almost nobody builds this in-house at launch.

Data feeds: Sportradar, Genius Sports, or SportsDataIO for scores, stats, and settlement truth. Confirm the provider before building the scoring engine.

Geolocation: GeoComply is the category standard. Fraud-resistant, audited, verified on every bet placement.

KYC and AML: Certified identity vendors for age, identity, source of funds, and sanctions and exclusion screening, wired into onboarding and payouts.

Payments: Regulated deposit and withdrawal rails with the reconciliation and reporting a gaming regulator expects. Betting payments carry higher decline rates and stricter rules than ordinary e-commerce.

Notifications: Push for bet settlement, cash-out offers, and lineup-lock reminders. Cheap per message, high value for a habit product.

The business math

A betting business earns in ways that look nothing like a subscription app. Here is the math in plain terms.

The hold. The sportsbook keeps a percentage of everything wagered. Across the US market in 2025, operators held roughly 10% of handle: for every $100 bet, the industry kept about $10. That 10% is before promotions. Free bets, boosted odds, and acquisition bonuses cut deeply into it, which is why a book's marketing budget and its margin are the same conversation.

The vig. The hold over time comes from the vig baked into every price. On a standard two-sided market priced at -110 each way, the book keeps about 5% of everything staked on that market. Think of it as the table fee: the price of arranging the bet. Promotions that boost odds are the book voluntarily giving some of the vig back to buy market share.

The fantasy rake. Daily fantasy takes a cut of contest entry fees, typically 10-15%. Dream11-style platforms take a platform fee of 15-25% of the prize pool. There is no odds risk: the platform runs a pari-mutuel pool and never loses on a result. Revenue scales with contest volume, not with how the match plays out.

The retention equation. DraftKings reported average revenue per monthly payer of around $97 in Q4 2024, while online sportsbook acquisition costs run $250-$750 per funded player. That gap is the whole business: you lose money acquiring a bettor and earn it back over months of play. The loyalty, cash-out, and CRM layer is not a nice-to-have. It is the payback mechanism.

The rent-vs-own crossover. A trading provider taking 15% of gaming revenue on a book generating $20 million a year is $3 million annually flowing to your vendor. Against a custom trading and platform build in the low seven figures, the crossover comes fast. Below a few million in revenue, renting almost always wins.

White-label vs custom: when each wins

Rent a turnkey platform when you are entering one jurisdiction, want to launch in months, do not want to run a trading and risk desk, and your handle is unproven. Paying a revenue share beats fixed engineering until you have volume.

Build custom when your handle is large enough that the platform's cut exceeds the cost of owning the stack, when you need to control odds and margin to differentiate, when you want to own player data and CRM because retention economics are the whole game, or when you run multiple brands on one platform.

One guide, four products: which model are you pricing?

The brands in the title are reference points, not interchangeable products. Here is the family this guide covers.

If fantasy is your entry point, our fantasy sports platform development guide goes deeper on that model's data providers, DFS compliance, and build-vs-buy math. The prediction-market fork (on-chain vs licensed exchange) is the one product decision that changes your regulatory universe; make it with specialist counsel before the build begins.

Where betting builds go wrong

The license treated as a parallel track. Teams build for six months assuming the license will sort itself out. It will not. The license, the market-access deal, and the vendor certifications gate the launch, and the regulatory path takes 12-18 months. Sequence the license first or you finish a product that cannot legally take a bet.

Retention underpriced. Acquiring a bettor at $250-$750 only works if you keep them. The loyalty, bonusing, and CRM layer gets deferred to later, which means the operator spends heavily to acquire players who churn before they are profitable. Design the retention layer into V1, even when the money-making product is two phases away.

In-play treated as a later feature. Teams budget for a pre-match book and add live betting later, then discover it needs a different architecture: low-latency pricing, real-time risk, settlement in seconds. Retrofitting in-play is a rebuild. If live betting is the goal, the architecture anticipates it from V1.

How RaftLabs approaches sports betting builds

We build the software layers that make regulated betting products work: compliant DFS and fantasy platforms, player wallets, KYC and onboarding, and the player-management and loyalty systems that decide whether an operator keeps the bettors it pays to acquire.

We are direct about the boundary. We do not sell a gaming license, and we do not run your trading desk. For most operators, renting odds and trading from a licensed provider is the right call, and we integrate with it. What we build is the product around it, the compliance layer beneath it, and the retention engine that turns an app into a business.

The right first step is a scoping call: which product model you need, what your licensing and market-access situation looks like, and the realistic timeline to launch. Get those three clear and the build is straightforward.

Request a 30-minute scoping call and we will tell you what to build first, what it will cost, and what the phased roadmap should look like.