US sportsbooks kept an average of 10.16% of every dollar wagered in 2025, up from 9.2% in 2024, based on the American Gaming Association’s own revenue and handle figures. That national number, however, hides enormous month-to-month and operator-to-operator swings in the actual price a bettor faces. Our algorithm treats a posted line as a perishable data point rather than a fixed rating: it re-scrapes odds continuously, converts them to implied probabilities, and recalculates the embedded margin so that GamblScout’s scores reflect current pricing conditions, not a one-time snapshot.
Key takeaways
- The implied national sportsbook hold rate rose from 9.2% in 2024 to roughly 10.16% in 2025, using the AGA’s own reported revenue ($16.96 billion) and handle ($166.94 billion) figures.
- Margin is not stable even at the aggregate level: AGA tracker data shows national hold falling to 8.1% in one recent June after sitting at 12.5% in the same month a year earlier.
- Academic research shows the standard overround formula is best treated as a floor on expected bettor losses, not a precise one, because bookmakers price longshots with a larger margin than favorites.
- Commercial data feeds now support in-play markets with over 99% uptime across a full football season, which is the infrastructure layer our own scraping has to keep pace with.
- GamblScout’s algorithm logs implied margin continuously across operators and markets, feeding that time series into our broader scoring weights rather than relying on a single sampled price.
Table of contents
What sportsbook margin actually measures
A sportsbook builds a profit margin into every market by quoting odds whose implied probabilities sum to more than 100%. Sum the inverses of the decimal odds on all outcomes and subtract 100%, and the remainder is the overround — the theoretical edge a book holds if it takes balanced action on both sides. Academic work on this calculation is explicit that gamblers commonly use the overround to estimate a bookmaker’s profit margin, and that this figure is what most published bookmaker margin guides are built on.
The catch is that the formula assumes every bet on an event carries the same expected loss rate for the bettor. Research on this exact question finds that assumption breaks down in practice:
if bookmakers set higher profit margins for bets with lower probabilities of winning (as implied by the evidence on favourite-longshot bias), then an equal-weighted average of loss rates across each of the available bets will be higher than predicted by this widely recommended calculation.
In plain terms, the headline overround on a match tends to understate how much a bettor loses on average once they start betting longshots, props, or underdog lines rather than the favorite. The same research concludes
the overround formula provides an obvious formula to use but our findings suggest it should be treated as a minimum likely average loss rate.
That distinction matters for how we score operators. A single “average margin” figure for a sportsbook is a floor, not a ceiling, and it can vary sharply depending on which market segment — spreads, moneylines, same-game parlays, player props — is being sampled.
How US sportsbook holds climbed past 10 percent
At the national level, the trend is unambiguous. Legal US sportsbooks reported
commercial sports betting revenue up 24.8 percent to $13.78 billion as Americans legally bet a total of $149.90 billion on sports throughout the year
in 2024, which works out to an implied hold of roughly 9.2%. A year later, the American Gaming Association’s own release confirmed
Sports Betting revenue rose to $16.96 billion, a 22.8 percent increase, on a total handle of $166.94 billion (+11.0 percent)
— an implied national hold of about 10.16%. Revenue is growing roughly twice as fast as handle, which means operators are extracting a larger cut of every dollar wagered even as overall betting volume growth slows.
This is consistent with what a hub page on the macro economics of iGaming would predict: as a regulated market matures, promotional intensity fades and product mix shifts toward higher-margin bet types, and the aggregate take rate rises even without any single price becoming more “expensive” in isolation.
Why the same market swings hard from month to month
The annual average also masks large seasonal variance. The AGA’s own quarterly tracker shows that in one recent reporting period,
June handle increased 26% to $12.59 billion, June revenue, however, declined 18.3% due to a substantially lower hold rate of 8.1%, down from 12.5% last year.
A national hold rate that moves from 12.5% to 8.1% in the same calendar month a year apart, driven largely by which side of close games and parlays happened to land, shows why a margin figure captured once and left in a static review is close to meaningless within weeks. Outside the US, the pattern of large but volatile sports-betting yields is similar: the UK Gambling Commission’s own quarterly data recorded that
real event betting GGY grew by 16 percent YoY to £625 million
in a single quarter, reflecting how sensitive operator revenue is to the calendar of major sporting events rather than a fixed pricing policy.
The technical problem: odds that change by the second
None of this variance would be trackable without infrastructure built for constant change. Modern sportsbooks price markets off commercial data feeds rather than manual trading alone, and the vendors behind those feeds compete explicitly on update speed and uptime. Genius Sports, one of the two dominant suppliers of official league data and live odds to sportsbooks, reports that in the 2024 English Premier League season, Genius Sports delivered 99%+ in-play market uptime, which the company frames as
over 33 hours per season, and 23 whole extra matches
of tradable live markets compared with a lower-uptime feed. That single-digit percentage of downtime is treated as a meaningful competitive differentiator in an industry where a stale line for even a few seconds during live play can be picked off by faster bettors or expose the book to arbitrage.
For an odds-tracking system, this means the target is not static: a sportsbook can post, suspend, and repost dozens of prices on a single market within a single match, and margins on props and same-game parlays can shift independently of the core spread or total. Any monitoring approach that samples odds once a day, or even once an hour, on live markets is effectively working with expired information.
How our algorithm tracks margins 24/7
Our scraping layer, described in more technical detail in The Architecture of a Casino Scraper, applies the same core principle to sportsbook pricing: high-frequency, structured collection rather than periodic manual sampling. For odds specifically, the pipeline does four things on a continuous loop:
- Ingestion — pulling posted prices across pre-match and in-play markets for each tracked operator at short, fixed intervals, with faster polling during live events.
- Normalization — converting American, decimal, and fractional odds formats into a single implied-probability representation so margins are comparable across operators and jurisdictions.
- Margin calculation — summing implied probabilities per market to compute the overround, then storing that value as a time-stamped series rather than a single figure.
- Deviation flagging — comparing each operator’s margin against a rolling market average for the same fixture and bet type, surfacing outliers where a book’s price is unusually generous or unusually tight relative to peers.
That time series is what feeds our broader scoring model, covered on the Scoring System & Algorithmic Weights hub — an operator’s margin behavior is one weighted input among several, alongside signals from our NLP and sentiment analysis work on posted terms and our fair play and fraud detection checks, rather than a standalone score.
What margin tracking reveals — and what it doesn’t
Continuous margin tracking is useful for spotting relative pricing patterns across operators and over time, but it has real limits. A market-average overround tells you nothing about limits, promotional adjustments to specific prices, or how a book prices exotic same-game parlays versus straight bets. The table below illustrates the kind of aggregate contrast our data shows at the national level using confirmed AGA figures, without implying any single operator’s price on any single bet.
| Year | Reported handle | Reported revenue | Implied national hold |
|---|---|---|---|
| 2024 | $149.90 billion | $13.78 billion | ≈9.2% |
| 2025 | $166.94 billion | $16.96 billion | ≈10.2% |
Sources: AGA State of the States 2025; AGA 2025 annual results release. Implied hold calculated as revenue divided by handle.
These are national aggregates, not per-operator or per-bet margins, and they combine straight bets, parlays, and props into a single blended figure. Our own operator-level data disaggregates this further, but the direction is consistent with the academic point above: as a market’s product mix shifts toward higher-variance bet types, the blended hold rate rises even if the price on a plain moneyline bet barely moves. A market-structure study of online betting pricing similarly finds that
for games where there is a strong favourite will be slightly lower than for games that are effectively toss ups
, meaning the margin embedded in a coin-flip match is often marginally higher than in a lopsided one — the opposite of what casual intuition might suggest.
Frequently asked questions
What is the difference between overround, vig, and hold?
Overround and vig (or “juice”) refer to the same thing: the built-in margin calculated from posted odds before a bet settles. Hold is a retrospective, accounting-based figure — actual revenue divided by actual handle over a period — which reflects real betting outcomes, not just posted prices, and can diverge from the theoretical overround depending on which side of bets won.
Why has the US national sportsbook hold rate kept rising?
Reported AGA figures show implied national hold moving from about 9.2% in 2024 to roughly 10.2% in 2025 as revenue grew faster than handle. This pattern typically reflects a shift in bet mix toward parlays and in-play wagering, plus reduced promotional spending as markets mature, rather than a uniform price increase on every bet type.
Does a lower average margin always mean better value for a bettor?
Not necessarily. A low blended margin can be driven by heavy volume on low-margin markets like standard spreads while props or parlays at the same operator carry a much higher built-in edge. Margin data is most useful compared like-for-like, market type against market type, rather than as a single operator-wide number.
Can an individual bettor track margins the way an algorithm does?
Manually, yes, on a small scale: calculating implied probabilities from posted odds and comparing them across a handful of books before placing a bet. What’s hard to replicate manually is doing this continuously across dozens of operators and thousands of markets, which is the scale problem automated tracking is built to solve.
Methodology note
For this topic, GamblScout’s algorithm ingests posted odds across pre-match and in-play markets at short polling intervals, normalizes them into implied probabilities, and stores the resulting overround as a time series per operator and market type rather than a single sampled value. That series is cross-referenced against rolling peer averages to flag outliers and is weighted alongside our other signals described on the Data Scraping & The Technical Engine hub before contributing to an operator’s overall score.
Gambling involves risk. Only play with money you can afford to lose and use the deposit limits and self-exclusion tools available in your jurisdiction.
