Sportsbook Trading Operations: How the Desk Makes Money

• A trading desk does three jobs: compile odds, manage lines as money and information arrive, and control liability through limits and player profiling.
• The overround is the built-in margin. A -110/-110 market carries a ~4.5% theoretical hold; US blended holds now average around 10%, driven almost entirely by parlays.
• Parlays and same-game parlays are the profit engine: New Jersey reported a 24.2% parlay hold in September 2024 versus 4.4% on everything else.
• In-play now accounts for roughly half of online betting turnover globally, and reportedly more than half in mature markets.
• Under roughly EUR50M annual GGR, don't build an in-house desk. Managed trading from Kambi, Sportradar, Altenar or Digitain covers 95% of the workload; in-house makes sense only when trading is your actual edge.

Sportsbook Trading Operations: How the Desk Actually Makes Money

The 2026 FIFA World Cup reportedly took more than $50 billion in wagers worldwide, and US sportsbooks called it the biggest betting event in history -- US handle alone equaled roughly ten Super Bowls. Every one of those bets passed through a trading function: a price was compiled, a liability tracked, a limit applied, and a margin baked in before the bettor ever saw the odds.

That trading function is the profit engine of a sportsbook, and the least understood part of the business. Operators obsess over acquisition costs and platform features, then treat trading as a black box. That's how you end up with a 4% hold in a market where the leaders print 10%.

Here's how a trading desk works in practice -- odds compilation, overround mechanics, in-play risk, cash-out pricing, parlay margins, player factoring -- and the decision every operator eventually faces: build your own desk or buy managed trading from suppliers like Kambi, Sportradar, Altenar or Digitain.

What a trading desk actually does

Strip away the jargon and the function has three jobs.

Odds compilation. Someone -- increasingly, something -- has to turn a probability estimate into a price. For a Premier League match, that starts with a model output (say, home win 48%, draw 26%, away 26%), gets adjusted for team news, and has margin applied. Models generate the baseline for tens of thousands of events; human traders adjust the head of the curve: big soccer leagues, the NFL, marquee races.

Line management. Once a price is live, it has to move. Sharp money is information: if respected accounts keep hitting the over, your total is probably wrong. Traders and automated engines shade lines off bet flow, competitor prices, and news. The goal isn't a perfectly balanced book -- that's a myth -- it's a price accurate enough that the margin does its work over volume.

Risk and liability management. Every open bet is a liability. The desk tracks exposure per market, event, and customer, and decides where the book holds a position versus where it moves the line, cuts limits, or (rarely) hedges. This is where player profiling comes in.

Our guide on how to launch a sportsbook covers where trading sits relative to platform, payments and licensing.

Overround and margin: the worked example

The overround is the sum of implied probabilities in a market beyond 100%. It's the theoretical margin, applied before a single bet lands.

Take the standard US point spread at -110 both sides. Each -110 price implies 110/210 = 52.38% probability. Two sides: 104.76% total book. That 4.76 points of overround means a theoretical hold of about 4.5% of handle (4.76/104.76).

A three-way soccer market works the same way. Price it 2.10 / 3.40 / 3.60 in decimal odds:

• Home 1/2.10 = 47.62%
• Draw 1/3.40 = 29.41%
• Away 1/3.60 = 27.78%

Total book: 104.81%, so roughly a 4.6% margin. A desk sets target overround per sport and market tier -- tight on main lines where price competition is brutal, wider on derivatives and long-tail sports where bettors don't shop.

Here's the part operators get wrong: theoretical margin and actual hold are different animals. Actual hold depends on results variance and, far more importantly, product mix. Per Legal Sports Report, US sportsbooks pushed blended hold to roughly 10% in 2025 -- double the theoretical margin on a straight bet -- and the gap is almost entirely parlays. New Jersey's annual hold hit 9.65% in 2025, up from 8.54% a year earlier, per iGamingBusiness.

If your product mix can't reach those parlay numbers, you're competing on straight-bet margin against operators who don't have to -- the squeeze we covered in sportsbook margin compression.

Pre-match vs in-play: where the volume went

Pre-match trading is the classic discipline: compile early, take positions from sharps, refine the closing line. In-play is a different sport entirely: prices update every few seconds off live feeds, markets suspend around dangerous moments, and no human can price 200 concurrent events by hand.

The volume shift is done arguing with us. Roughly 47% of global online wagers were placed in-play in 2024, and in mature markets the share has reportedly already crossed 50% of turnover. Live betting is the main event; pre-match is the warm-up.

Two operational consequences:

• Latency is a risk parameter, not a UX metric. If your feed is two seconds behind the courtsider in the stadium, you're running a charity. Bet delays, automated suspension triggers and feed redundancy are trading decisions.
• In-play margin is structurally higher. Live markets carry wider overround than pre-match main lines because bettors accept it -- the bet is impulsive and time-boxed.

This is also why live video rights matter commercially -- watch-and-bet drives in-play turnover, which we broke down in live streaming for sportsbooks.

Cash-out: the margin you pay twice

Cash-out looks like a customer feature. It's a trading product. When a bettor cashes out, the operator computes the current fair value of the open bet from live prices, then applies a margin to that value. Since the original bet already carried an overround, a cashed-out bet has been margined twice.

Let's be blunt: cash-out is one of the quietest profit lines in the book. The bettor sees "lock in a win"; the desk sees a second bite at the margin plus reduced variance. The trading decisions are how aggressively to margin the cash-out price (too greedy and usage drops), which markets get it, and when to pull it -- engines suspend cash-out the moment underlying markets suspend.

Bet builders and same-game parlays: the profit engine

If one product explains modern sportsbook economics, it's the bet builder -- the same-game parlay in US terminology. The margin math is brutal: every leg carries its own overround, and multiplying legs compounds the margin. Correlation pricing adds another layer -- the operator prices the legs' dependence (a striker to score and his team to win are correlated), and that pricing is itself margined.

The state data quoted above tells the story: 18-24% hold on parlays versus 4-5% on straights, a four-to-five-times difference. In some US states over recent 12-month stretches, parlays reportedly made up about a third of handle but roughly two-thirds of revenue. For desk economics, that means:

• SGP pricing capability is not optional. If your platform or trading supplier can't price correlated legs across thousands of events, you're locked out of the industry's highest-margin product.
• Recreational mix drives hold. The more of your handle comes from casual multi-leg bettors, the higher and more stable your blended hold.
• Suppliers know this, which is why Kambi sells Bet Builder as a standalone module and every managed-trading pitch now leads with SGP depth.

Liability limits, profiling and factoring

No book takes unlimited bets. The desk sets maximum win limits per market tier -- high on Premier League match odds, low on Finnish third-division corners -- because limits scale with price confidence and market liquidity.

Player profiling ("factoring" in the UK trade) assigns customers a factor that scales their limits. A flagged sharp might be factored to 10% of standard stakes on soft markets; a recreational parlay customer gets full limits everywhere. Profiling runs on signals: bet timing versus line moves, closing-line value, market selection, and win rate on early prices.

Two caveats. First, factoring is commercially rational but reputationally loaded -- regulators including the UKGC have scrutinized how operators restrict winners while marketing hard to losers, and affordability-driven limits are a separate obligation tied to responsible gambling rules. Build restriction logic you can explain to a regulator, not just to your CFO. Second, profiling isn't a substitute for pricing. If sharps beat you constantly, your prices are wrong; banning the messengers doesn't fix the model.

Automation vs human traders

Modern trading is a pyramid. At the base, algorithms price the long tail -- table tennis, esports qualifiers, minor leagues -- fully automatically, thousands of events with zero human touch. In the middle, traders handle what the automation flags: suspicious flow, feed disagreements, big liabilities. At the top, senior traders make judgment calls on marquee events, novelty markets and model overrides.

The ratio keeps shifting toward the machines. What automation still doesn't do well: pricing news (a manager sacked an hour before kickoff), pricing genuinely new products, and the gray zone where a "sharp" pattern might actually be a bonus-abuse ring -- a problem that overlaps with fraud tooling, covered in our AI fraud detection ROI piece.

Build vs outsource: the real decision

Here's the choice, stripped of vendor marketing. In-house means you employ compilers, in-play traders and risk analysts across time zones, license raw data feeds, and own margin strategy end to end. Managed trading means a supplier's desk prices and risk-manages your book for a revenue share or fee, and you steer through configuration.

The supplier tiers, briefly. Kambi historically sold the full sportsbook bundle and has now modularized -- Managed Trading, Odds Feed+ and Bet Builder are standalone products, so you can take the trading layer without the whole stack. Sportradar runs Managed Trading Services (MTS) as a risk-and-pricing layer that bolts onto an engine you already operate, and can trade third-party content through feed convergence. Altenar bundles a 24/7 trading and risk team into its platform with multi-feed redundancy -- a mid-tier turnkey route. Digitain plays the turnkey value tier: fast launch, managed trading included, popular in emerging markets. See Kambi vs Sportradar for the top two and Digitain vs Altenar for the mid-tier.

The honest heuristic: a competent in-house desk is a multi-million-euro annual commitment before it prices a single event better than Kambi's or Sportradar's shared desk. If trading isn't the axis you compete on, outsource it and spend the money where your edge lives. Bring trading in-house when fixed costs beat the revenue share at your scale and you have a genuine thesis for why your prices will differ. Several tier-1 US operators made that migration off Kambi; almost nobody below tier 1 has.

How to build a trading function: step-by-step

1. Pick your trading model -- Decide between fully managed (Digitain, Altenar turnkey), hybrid (Sportradar MTS or Kambi Managed Trading with your own margin configuration), or full in-house. Anchor the choice to projected GGR: below roughly EUR50M a year, in-house fixed costs rarely beat a revenue share.
2. Contract your data and pricing feeds -- Line up official data rights for core sports, at least two feed sources for redundancy, and a pricing feed or model stack for the long tail. Latency SLAs matter more than price lists; get them in writing.
3. Set margin and liability frameworks per sport -- Define target overround by sport and market tier, maximum win limits per tier, and automated suspension rules for in-play. Document it -- this framework is what traders or supplier configuration execute daily.
4. Stand up risk and player-profiling workflows -- Implement liability dashboards, factoring rules with clear criteria, and an escalation path for suspicious flow, including integrity-body reporting. Every restriction decision must be explainable to a regulator.
5. Automate the long tail, staff the head -- Let algorithms run minor markets untouched; hire traders only for supervision, marquee events, and model exceptions. A hybrid desk of 3-6 people over a managed service covers most mid-size operators.
6. Measure, reprice, repeat -- Track theoretical versus actual hold weekly by product, monitor closing-line value against sharp books, and review parlay mix monthly. A desk that doesn't measure itself is just a cost center with opinions.