A sportsbook welcome offer is a one-time acquisition cost paid to a new customer. Liquidity — how large a bet a book will accept and how fast it repriced that bet — and line quality — how close a posted price sits to the market consensus — are structural, apply to every wager for the life of the account, and are far harder for an operator to fake. That is why our scoring model weights liquidity and odds quality above promotional size: national sportsbook hold has climbed from 6.9% in 2019 to roughly 10.2% in 2025, which means the house’s structural take on every dollar wagered has grown faster than almost any individual promotion can offset.
Key takeaways
- U.S. national sportsbook hold rose from 6.9% in 2019 to approximately 10.2% in 2025, and briefly topped 11% again in single states during April 2026, driven largely by same-game parlays and in-play markets.
- DraftKings’ own 2025 annual report attributes margin expansion partly to “improved promotional reinvestment,” meaning promo costs are being managed down even as hold rises — a double squeeze on bettor value.
- Massachusetts regulators found a direct correlation between winning and having bet limits reduced, even though only about 0.64% of accounts were formally restricted.
- Our algorithm scores line quality (overround vs. consensus) and inferred liquidity ahead of bonus size because these affect every wager, not just the first one.
Table of contents
The hold percentage reality check
Hold — the share of total handle a sportsbook keeps after paying winners — is the single clearest structural metric in sports betting, and it has moved in one direction for years. National hold rose to roughly 10.2% in 2025, up from 9.2% in 2024, 8.0% in 2022 and 6.9% in 2019. That is not a rounding change; it is an industry-wide repricing of every dollar wagered.
The trend is also volatile at the state and operator level. In April 2026, Massachusetts reported FanDuel holding 13.40% and DraftKings 13.12% in a single month, a reminder that same-game parlays can push effective margins well above the headline national figure for specific books and specific weeks. The same tracker shows how quickly hold can also fall when competitive or structural pressure increases: national hold dropped to 9.2% in August 2026 from 10.8% a year earlier, a decline the American Gaming Association linked partly to competition from prediction markets rather than any voluntary pricing improvement.
Operators themselves describe the mechanics behind rising margins in their own disclosures. DraftKings’ FY2025 annual report notes that its Sportsbook Net Revenue Margin rose from 6.0% in 2024 to 7.1% in 2025 “as a result of increased Sportsbook hold percentage and improved promotional reinvestment”. Read plainly, that single sentence says two things at once: the book kept more of every dollar wagered, and it spent proportionally less of that extra margin buying back customer goodwill through bonuses. A bettor evaluating “value” purely by the size of a sign-up offer is looking at the smaller, shrinking half of that equation.
What liquidity means in a sportsbook
Liquidity in this context has two practical meanings for a bettor. The first is depth: how much money a book will accept on a given side before moving the price or capping the bet. The second is responsiveness: how quickly a book updates its line in reaction to new information or heavy action elsewhere in the market. A book with thin liquidity might post an attractive number, but if it only holds that price for a $50 stake before slashing the limit, the headline odds are close to irrelevant for anyone betting meaningful size.
Part of why liquidity has become harder to assess from the outside is the shift in bet-type mix. Industry estimates compiled from state filings suggest live, in-play betting has grown from the high-30s percent of handle in 2022 to roughly half of handle by 2026, though the exact current share is contested since it is not uniformly reported by regulators. In-play markets are priced and repriced in seconds, so the liquidity a book can genuinely support — as opposed to the liquidity implied by a static pre-match limit — becomes the more meaningful figure, and it is one retail operators rarely publish.
The same structural shift toward parlays and live markets is a documented driver of the hold increase described above. One industry analysis of AGA data concluded that the transition toward these product types pushed average hold from the 6–7% range in 2021 to 9–11% by 2025. Liquidity and hold are therefore not separate concepts from a scoring perspective — the same product mix that compresses available liquidity on simple markets is what inflates the operator’s take on complex ones.
How the algorithm weights liquidity and lines vs. promos
Our data-scraping engine (described in more depth on the data-scraping hub) pulls posted odds across major markets at regular intervals, which lets the algorithm compute two things a promotional headline cannot capture: the operator’s average overround versus a consensus benchmark, and the frequency and magnitude of limit-related complaints surfaced through sentiment analysis of player forums and regulator complaint logs. Both carry more weight in the composite Trust Score than bonus size, because both apply to every wager an account places, not just the first one.
| Signal | What it captures | Why it outweighs promo size |
|---|---|---|
| Line quality / overround | Deviation of posted odds from market consensus across tracked markets | Affects the expected value of every bet placed, not a one-off credit |
| Inferred liquidity | Frequency of limit-reduction complaints, market suspension patterns, odds update latency | Determines whether a quoted price is actually available at meaningful stake size |
| Payout and withdrawal speed | Time from cash-out request to funds cleared, scraped from support logs and complaint data | A structural reliability measure independent of any single promotion |
| Bonus terms (NLP-parsed) | Wagering requirements, game weighting, withdrawal caps extracted from T&Cs | Scored for fairness, but capped in overall weight versus structural signals |
This approach is consistent with how we treat other operator-quality signals across the site. The same logic underpins why payout speed makes up 30% of the total Trust Score and how predatory bonus terms get penalized rather than rewarded — recurring, structural behavior outweighs a single attractive number on a landing page.
Winning accounts and the limiting problem
Liquidity is not just about posted limits; it is about whether an operator will continue to honor those limits once a bettor starts winning. This is where regulatory data is most revealing. The Massachusetts Gaming Commission’s review of operator data found that approximately 0.64% of the state’s 2.1 million online wagering accounts, or roughly 13,400 accounts, had been restricted. The headline number is small, but the commission’s own analysis went further: it specifically examined whether bettors who tend to win have their limits raised while those who tend to lose have theirs decreased, and concluded that both happen.
Separate reporting on the same data set found the restrictions were not always a hard ban. Some restricted accounts were capped at 90% or 75% of the normal limit, while roughly 12% of affected accounts were cut more severely, and bettors were typically not informed why. Partly in response, Massachusetts adopted a first-in-the-nation rule, requiring licensed sportsbooks to notify a player within 48 hours whenever their betting activity is limited and explain the restriction.
From a scoring standpoint, this is exactly the kind of structural reliability issue that a promotion cannot compensate for. An operator can offer a generous bonus to new depositors while simultaneously cutting limits on its most consistently accurate customers a few months later. Because this behavior is largely invisible in marketing material, our algorithm leans on patterns extracted via natural language processing of player complaints and sentiment data to approximate how aggressively an operator limits winners, since regulator-level disclosure like the Massachusetts program remains the exception rather than the rule.
Why promotional spend is a weak quality signal
The dollar figures operators spend attracting customers are large, public, and easy for a review site to lead with — which is precisely why they are a weak proxy for value. DraftKings’ 2025 annual report shows sales and marketing expense rising 9.1% to $1.4 billion in 2025, up from $1.3 billion in 2024. Flutter, FanDuel’s parent, reported a similar pattern: $3.68 billion in global sales and marketing expense in 2025, up 15% from $3.21 billion in 2024, with approximately $1.32 billion of that attributable to U.S. operations. These are acquisition budgets, not value signals — they tell a bettor how aggressively a brand is advertising, not how its odds, limits, or withdrawal process will treat a long-term customer.
Industry-wide advertising volume tells a more nuanced story than individual operator budgets. A Nielsen study commissioned by the AGA found that total sports betting advertising spend fell 5% year-over-year, with ad volume across all channels down 27% from its 2021 peak. Media-tracked ad volume and reported sales-and-marketing expense are not identical line items — the latter includes promotional credits, free-to-play contests, and personnel costs that Nielsen’s tracker does not capture — but the divergence is instructive: traditional advertising is retreating industry-wide even as individual operators’ total marketing expense lines continue climbing, which is consistent with budgets shifting toward targeted promotional reinvestment rather than broad brand advertising. That shift reinforces the point above: promotional spend is being allocated more surgically, not necessarily more generously, even as hold climbs in the background.
For more on how operator economics shape consumer outcomes industry-wide, see our macro economics of iGaming hub, and for related algorithmic weighting decisions, the longevity bonus article covers a similar case where a structural, non-marketing signal outranks surface-level claims.
Frequently asked questions
What is sportsbook hold and why do you weight it over bonuses?
Hold is the share of total wagers a sportsbook retains after paying winning bets. It is a structural, recurring margin that applies to every wager placed, unlike a bonus, which is a single credit tied to a new account. With national hold near 10.2% in 2025, the long-run cost of a high-hold book typically exceeds what any single promotion returns.
Does a low vig always mean better value for a bettor?
Generally yes, assuming similar market coverage and reliable payouts, because a lower overround means a smaller built-in edge against the bettor on every wager. It is not an absolute guarantee, since low-margin pricing can coincide with lower limits on recreational-facing markets or narrower market coverage, both of which our algorithm also scores separately.
Why do sportsbooks limit winning bettors instead of simply raising the vig?
Raising the vig for everyone would hurt casual, recreational volume, which is the core of most retail operators’ revenue. Regulatory data from Massachusetts found a correlation between consistent winning and having bet limits reduced, suggesting operators prefer targeted restriction of a small number of accounts over broad price changes that would affect everyone.
How does your algorithm score liquidity if it cannot see an operator’s internal limits?
We cannot access an operator’s internal risk engine, so liquidity is inferred indirectly: through odds-update latency across tracked markets, the frequency of market suspensions, and sentiment analysis of limit-related complaints in player forums and regulator filings, as detailed on our NLP methodology hub.
Does this mean low-margin “sharp” books always rank at the top?
Not automatically. A book can have tight margins on headline markets but score poorly elsewhere — on payout speed, game coverage, or bonus-term fairness — all of which feed the same composite score described in our payout speed weighting article. Line quality is one heavily weighted input, not the entire score.
Our data signals
For sportsbook scoring specifically, our algorithm combines scraped, time-stamped odds across major markets (used to compute average overround versus a consensus benchmark), update-latency measurements during live events, and NLP-derived sentiment from player complaints referencing account limits or line suspensions. These are weighted alongside the payout-speed and bonus-term signals used across the rest of our review methodology, consistent with the approach described throughout the Scoring System & Algorithmic Weights hub.
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.
