The number on your screen was assembled, not chosen
Price two evenly matched teams at 1.91 each and the implied probabilities add up to 104.7%, not 100%. That extra 4.7 percentage points is the whole business model, and it was put there by software you have never heard of, running somewhere you will never see. Understanding how sportsbook odds work starts with accepting an awkward fact: the brand on your app is usually not the company calculating the price.
Odds come out of a backend odds engine — data feeds, probability models, risk tools and a trading desk — and most operators license that engine from a specialist rather than building it. A current example landed in October 2026, when BETBY announced a partnership with Lebull in the regulated Polish market, powering the brand’s sportsbook with its technology, betting content, features and managed services. Lebull keeps its own frontend. The pricing machinery sits behind it.
From data feed to price: how odds compilers do the work
Odds compilers turn raw probability estimates into sellable prices in three steps: a model produces a fair probability for each outcome, margin is added so the book’s prices sum above 100%, and the result is published as decimal, fractional or American odds. The model does the estimating. The margin does the earning.
Models and the data that feeds them
The input is an API data feed: team and player statistics, historical results, line-ups, injuries, weather, and in many cases the prices other books are showing. A model ingests that and outputs a probability — say 55% for the home side. Convert 55% to a price and you get 1/0.55 = 1.82. That is the fair price, the one that would return nothing to the book over the long run. No sportsbook posts it.
The feed matters more than most bettors assume. A price is only as good as the data behind it, which is why liquidity feeds and official data rights are fought over hard in this industry. A book pricing a lower-tier league off thin data is pricing with wider uncertainty, and it protects itself by widening the margin instead.
Where human traders still override the machine
Automation handles volume; traders handle judgement. A model has no opinion on a rumour, a manager’s press conference, or a sudden flood of one-sided money from a few accounts. Trading rooms watch exposure across markets and move lines when the book’s liability on one outcome grows uncomfortable — not because the true probability changed, but because the book’s position did.
This is the part that gets flattened in most explainers. Odds are a blend of estimated probability and commercial risk management, and on big events you genuinely cannot tell from the outside which of the two moved a line. BETBY’s arrangement with Lebull includes support from its in-house trading team, including bespoke markets built for effectively any event, which is the same function described from the supplier side.
The brand is the skin, the odds engine is the body
Sportsbook software splits into two layers: the frontend a bettor touches (brand, app, design, account journey, promotions) and the backend that prices markets, manages risk and settles bets. These are routinely supplied by different companies, and the split is invisible from the betting slip.
Which means two sportsbooks that look nothing alike can be quoting the same engine. If you have ever opened three apps and found identical prices on an obscure volleyball match, you probably were not seeing three independent opinions. You were seeing one engine with three paint jobs. Line shopping across brands only helps when the brands actually run different pricing, and a surprising share of them do not.
The Lebull deal is a tidy illustration of the modular approach. The operator preserved the interface and user journey its players already knew, and integrated BETBY’s sportsbook capabilities behind the scenes. Same shop window, new engine room.
Why almost nobody builds their own odds engine
Operators license betting technology because building a competitive odds engine means funding a permanent engineering team, a data rights budget, a 24-hour trading room and a settlement system, all before taking a single bet. Licensing turns that into a running cost with a launch date measured in months.
There are three pressures doing the work here:
- Speed to market, because a regulated licence that is sitting idle earns nothing.
- Market specificity, because the sports, leagues, payment habits and compliance reporting that matter in Poland are not the ones that matter in India or Brazil.
- Breadth of coverage, because players expect thousands of events and in-play markets that a small in-house book cannot price alone.
A turnkey sportsbook platform solves all three at once. The trade-off is real: license the engine and you inherit someone else’s pricing, someone else’s risk appetite, and someone else’s roadmap. That is why the more interesting deals are modular rather than full turnkey — the operator keeps what it is good at and rents the rest.
What the BETBY and Lebull deal actually says
BETBY’s October 2026 agreement with Lebull extends the provider’s position in the regulated Polish market, with Lebull described as a Tier 1 B2C betting brand. BETBY supplies the sportsbook technology, betting content, features and managed services; Lebull retains its existing frontend. The integration gives Lebull access to BETBY’s event and market coverage, its in-house trading team and bespoke market creation, plus Betby Games, the provider’s proprietary esports product.
Read that as a template rather than a headline. A B2B sportsbook provider wins a market by signing recognisable local brands, and the local brand gets depth of markets it was never going to compile alone. Nothing in the announcement tells a bettor the prices will be sharper. It tells them the market list will be longer and the plumbing more reliable.
Real-time repricing and the latency problem
In-play odds update in well under a second because the pipeline is automated end to end: an event data feed pushes a state change, the model reprices every affected market, and the new numbers publish to the app without a human touching them. Humans set the rules; the machine applies them thousands of times a match.
The risk is latency. If the feed reaches the engine slower than it reaches a bettor watching a stream, the old price is briefly wrong and exploitable. Books defend that gap with bet delays of a few seconds on in-play slips, automatic market suspension around goals, penalties, wickets and timeouts, and stake limits that shrink as volatility rises. Those pauses that annoy you mid-match are the risk system doing exactly what it was bought to do.
The margin hiding in plain sight
The overround is the amount by which a market’s implied probabilities exceed 100%. Take that 1.91 / 1.91 two-way market: 1 ÷ 1.91 = 52.36% each, 104.71% combined. The 4.71 points above 100 is the book’s margin, which works out to roughly a 4.5% edge on money staked in that market.
| Market | Decimal odds | Implied probability |
|---|---|---|
| Team A | 1.91 | 52.36% |
| Team B | 1.91 | 52.36% |
| Total | — | 104.71% |
Margins vary by sport, league and market type. Headline markets on major competitions tend to be priced tightest; niche props and obscure fixtures carry more, because the book is less certain and the competition is thinner. No market is ever posted at a true 100%, and no amount of better technology changes that — the margin is the product, not a flaw in it.
What better tech does and does not change for you
Faster engines and bigger B2B platforms give you more markets, quicker settlement and fewer frozen screens. They do not shift the maths in your favour. If anything, sharper pricing and real-time risk management close the gaps a bettor could once find, while the overround sits where it always sat.
So the useful habit is not chasing the slickest app. It is checking whether the prices you are comparing come from genuinely different engines, and reading the margin in a market before you back anything in it. Over time the book’s edge wins; that is arithmetic, not pessimism. Bet only money you can afford to lose, set deposit and loss limits before you need them, and use cool-off or self-exclusion tools if betting stops being entertainment. If you want to see the margin for yourself, add up the implied probabilities on the next market you open. The number above 100 is the part nobody advertises.
