Across 2,500 Stake Crash rounds CasinoDrop recorded on 17–18 August 2026, the apparent return on a game designed to pay back exactly 99.00% ranged from 97.60% to 108.00%, depending on which cash-out target you measured it at. Every one of those readings was normal. That is the problem with how to check slot RTP by playing: the number you can look up takes two minutes to find, and the number you could measure yourself would take weeks of continuous play to pin down.
What RTP is, and what it is not
Return to player is a design figure. Over a very large number of spins, a 96% game is built to return 96 units for every 100 staked, and the remaining 4 units are the house edge. It is an average across every player and every session, not a description of what any one session will do.
The Gambling Commission splits the idea in two, and the distinction is the one this whole article turns on. Theoretical RTP is the figure the game was designed and advertised with. Actual RTP is what live play produced — total wins divided by total turnover. In the Commission’s own worked example, a game designed at 91.68% that took £1,200,000 in turnover and paid £1,085,000 in wins produced an actual RTP of 90.42% over that window. Nothing was wrong with the game.
“This is the designed return to player percentage of the game, it will also be the advertised RTP of the game as displayed in the player facing rules, as per RTS 3C.” — Gambling Commission, How to calculate return to player (RTP)
Random-number testing is a separate job. Laboratories such as eCOGRA treat the RNG as the thing that keeps each spin independent of the last. RTP is the share of stakes the maths is built to return. A perfectly fair RNG and a 92% RTP sit in the same game without contradiction — random does not mean generous.
The check itself: read it off the game
The figure that matters is the one attached to the copy of the game in front of you. UK remote rules require information that lets a player judge the chance of winning — house edge, RTP, or event probabilities — to be easily available before they commit, and the Commission’s implementation notes put that content in game artwork, help or how-to-play pages, or supporting material reachable from the game menu. Malta requires full game rules no more than one click from the play page. New Jersey requires the rules of play to be readable without placing a wager.
On a land-based machine the same idea applies in a clunkier form. The Commission’s player-facing guide says category, %RTP, and whether the machine is compensated or random must be available on every machine, as a label, a notice or a help screen. New Jersey’s Division of Gaming Enforcement tests slots against a minimum theoretical return. Staff can point you at the help pages; nobody at the venue can change the certified maths in the cabinet.
Why the same slot has more than one RTP
The artwork is the product; the paytable maths is a setting. A studio builds one game and then certifies several RTP configurations of it. The operator picks the one that suits its licence floor, tax position and margin target. Reels, features and the max-win headline stay identical.
This is not a leak from someone’s database — the suppliers publish it. Hacksaw Gaming’s own game-data sheet for Wanted Dead or a Wild lists four official RTPs, and Densho from the same studio follows the same four-band pattern.
| Certified build | Wanted Dead or a Wild | Densho | House keeps (derived) |
|---|---|---|---|
| Top band | 96.38% | 96.40% | 3.62% |
| Second band | 94.55% | 94.38% | 5.45% |
| Third band | 92.33% | 92.33% | 7.67% |
| Bottom band | 88.42% | 88.34% | 11.58% |
Treated as a price, the gap is not small. The 96.38% build keeps 3.62% of stakes; the 88.42% build keeps 11.58%. That is more than three times the hold on the same title, with the same max win and the same bonus names. In April 2026, iGaming Business reported the same lever in industry language: operators deploy 96%, 94% or 92% versions of one game to take more margin without changing the product. Helen Walton of G Games, quoted there, described RTP as a pricing lever and said UK standard RTPs had moved from around 96% toward 94%, with 92% now often the requested maximum. Eyal Loz of RubyPlay put the arithmetic plainly in the same piece: cutting 96% to 92% doubles the operator’s take from four cents to eight cents in the euro. Those are single-source industry observations rather than a rule change.
Studio marketing pages usually show one figure. Pragmatic Play’s Sweet Bonanza page publishes 96.48% and stops there. That is the default the studio wants associated with the brand, not a guarantee that every casino runs it.
Why you cannot measure RTP by playing
This is where most RTP advice stops, and it is the part that decides whether the rest is any use. Players routinely treat last night’s result as evidence about a game’s return. The Commission’s own monitoring guidance is the antidote, and it is worth working through rather than quoting.
The Commission’s worked example takes a game designed at 91.68% with a volatility (standard deviation) of 5.6, and asks how far actual RTP can sit from design without anything being wrong. Its published answers are about ±4.91 percentage points after 50,000 plays, ±1.74 after 400,000, and ±1.10 at one million. Those three points come from a single formula — 1.96 × 5.6 ÷ √n — which reproduces all three exactly, so the same formula can be run at the scales a real player actually reaches.
500 plays: 1.96 × 5.6 ÷ 22.4 = ±49.09 pp
50,000 plays: 1.96 × 5.6 ÷ 223.6 = ±4.91 pp → matches the published 4.91
1,000,000 plays: 1.96 × 5.6 ÷ 1,000 = ±1.10 pp → matches the published 1.10
This is also why “hot” and “cold” machines are a misreading of variance. A high-volatility 96% game can empty a bankroll faster than a low-volatility 94% game, because the 96% is concentrated in rare features. RTP is the average; volatility is how lumpy the path to it is. For a different kind of long-run drain built into a paytable, see how progressive jackpots divert a slice of every stake into a pool most sessions never touch.
What it looks like when you can check
Slots do not publish their round outcomes, so nobody outside the studio and its test house can audit a slot’s RTP from play — including us. Crash games are the opposite case: the payout rule is a single published formula, every settled round is public, and CasinoDrop records them. That makes them the closest thing to a control experiment for the question this article asks.
Stake Crash is built to a 1% house edge, so at every cash-out target the designed return is exactly 99.00%: cash out at 2x and the model gives a 49.50% chance of getting there, and 2 × 0.495 = 0.99. Across the 2,500 rounds in our feed, every one of the sixteen published targets landed inside its 95% interval.
The hit rates behave. The trouble starts the moment you convert them into an apparent RTP, which is exactly what a player does when they judge a game by results.
| Cash-out target | Rounds reaching it | Apparent RTP | 95% range on that reading |
|---|---|---|---|
| 1.5x | 1,660 | 99.60% | 96.82 – 102.38% |
| 2x | 1,230 | 98.40% | 94.48 – 102.32% |
| 5x | 515 | 103.00% | 95.07 – 110.93% |
| 10x | 244 | 97.60% | 85.97 – 109.23% |
| 50x | 54 | 108.00% | 79.51 – 136.49% |
| 100x | 26 | 104.00% | 64.23 – 143.77% |
Every game in that table is the same game, running the same 99.00% model, over the same 2,500 rounds. Read at the 10x target it looks like a 97.60% game. Read at the 50x target it looks like a 108.00% game — apparently paying out more than it takes in. Both readings are wrong, and both are entirely normal, because the rarer the outcome you measure, the fewer of them a sample contains and the wider the honest range around your answer becomes.
Watching a single target settle makes the same point over time rather than across targets.
n = (1.96 × 2)² × 0.495 × 0.505 ÷ (precision)²
±2.0 pp → 9,603 rounds → 2.3 days of continuous play
±1.0 pp → 38,412 rounds → 9.3 days
±0.5 pp → 153,648 rounds → 37.4 days
Thirty-seven days of uninterrupted play to settle one cash-out target on the simplest possible game. That is the honest scale of the thing, and it is why the two-minute check on the info screen is not a poor substitute for measuring RTP yourself. It is the only practical method there is.
What certification does and does not tell you
Before a remote game goes live in Great Britain it has to be tested by a Commission-approved test house. The current list includes eCOGRA, several Gaming Laboratories International entities, BMM, Gaming Associates and Quinel, among others, and those labs typically test interactive games against standards such as GLI-19. eCOGRA describes its RTP testing service as checking whether the theoretical return is achievable — including how it splits between base game, features and jackpots — and whether payouts sit inside a statistically acceptable band for that game’s volatility.
- This configuration’s maths match the RTP it advertises.
- The RNG is behaving, so spins are independent of one another.
- The return is achievable across base game, features and any jackpot.
- That the casino chose the highest certified configuration.
- That the build you are playing is the one the review page quoted.
- That the game is good value — an 88% build can be fully certified.
Switching a live game from 96% to 92% is not a hidden slider. It is a different tested build, and the player-facing rules have to show the figure that is actually running — which is precisely why the info screen is worth opening.
Demo play, bonus buys and floors
Demo play is closer to the real game than a review blurb, but only on the same operator. Malta tells licensees that demo RTP must match the real-money setting, and the UK’s RTS 6 requires play-for-free games on the same website to use the same rules and represent the real-money prize distribution. The same standard warns that a supplier’s own demonstration site, built to show off features, can run higher-than-normal returns where that is labelled. A studio demo is not a reading of the casino you will play at.
Bonus buy and extra-bet modes are a second trap. Hacksaw publishes a bonus-buy menu on both Wanted Dead or a Wild and Densho, and those modes often carry their own certified RTP because the player is buying a different slice of the paytable. Where a help screen lists more than one percentage, the relevant one is the line matching the bet you are about to place.
Finally, a licence floor is a legal backstop and not a quality mark. Malta held online games to a 92% minimum until 28 May 2021, when the Malta Gaming Authority cut the floor to 85% so that remote and land-based rules matched. That did not push every slot down to 85%; it made lower certified builds legal on MGA-facing product. New Jersey is more specific: N.J.A.C. 13:69E-1.28A requires a theoretical return of at least 83%, and the Casino Control Commission’s FAQ says the Division’s Technical Services Bureau tests machines against that mark. The UK sets no statutory machine minimum at all — an 85% label is lawful provided it is displayed. An 88% build can be fully legal and still be a worse price than the 96% build of the same slot in the next lobby.
You cannot audit a slot from the outside, but you can audit the class of game used as the control in this article. The crash board carries every settled round we record, with the hit rate at each cash-out target against the 99% model — watch a target drift and settle in real time instead of taking the convergence chart on trust.
Open the Stake Crash tracker →
Replay a real seed pair and watch its RTP settle in the simulator →
- We have not measured any slot’s RTP, and neither can you. No slot publishes its round outcomes, so the only figures in this article that were measured come from a crash game used as a control. Everything said about slots rests on the studios’ and regulators’ own published documents.
- Crash is not a slot. Its return sits on one published formula and a single cash-out decision; a slot spreads the same idea across a paytable with far rarer top outcomes. The control case is the easy case, which is the point — a slot needs more rounds to measure, not fewer.
- 2,500 rounds over about 20 hours on one casino’s game, against a store of 21,499. That describes those rounds, not Stake Crash’s lifetime and not any other operator’s crash game.
- The ±49 percentage point figure for 500 plays is derived from the Commission’s own worked example, which uses one game at 91.68% with a volatility of 5.6. A different game with different volatility gives a different band. The formula is theirs; the extension to session-sized samples is ours.
- The rounds-needed figures assume you could play one fixed cash-out target continuously at our measured pace. Nobody plays that way, and stake size, table limits and bankroll would end the experiment long before the sample did.
- Settled rounds, not players. We record what a game paid, not what any individual staked, cashed out or finished with.
- RTP configurations change. The Hacksaw and Pragmatic figures here are what those studios published at the time of writing, and the build at your casino can differ from all of them — which is the reason to read the info screen rather than this table.
FAQs
Open the game at the casino you will actually play, then open its info, help, paytable or game-rules screen. The theoretical RTP is listed there, usually on the last page of the paytable. UK remote rules require RTP, house edge or winning probabilities to be easily available before you commit. If the figure is missing, ask support for that instance’s RTP rather than assuming the default.
Studios certify several configurations of one title and the operator picks which to run. Hacksaw Gaming publishes Wanted Dead or a Wild at 96.38%, 94.55%, 92.33% and 88.42%, and Densho follows the same four-band pattern. The artwork, features and max win stay identical; only the hold changes. The 96.38% build keeps 3.62% of stakes and the 88.42% build keeps 11.58%.
No, and not by a small margin. On the Gambling Commission’s worked example of a 91.68% game with volatility 5.6, the 95% band after 500 spins is about plus or minus 49 percentage points. CasinoDrop measured the equivalent on a crash game with a known 99.00% design: across 2,500 rounds the apparent return read anywhere from 97.60% to 108.00% depending on which cash-out target you measured.
More than anyone plays. On the simplest case CasinoDrop could measure — a crash game’s 2x cash-out target, which lands about half the time — pinning the return to within half a percentage point takes 153,648 rounds, or roughly 37 days of continuous play at the pace we record. A slot spreads its return across much rarer outcomes, so it needs more rounds than that, not fewer.
Not as a statement about the casino you are using. Review pages and studio marketing usually quote the default or highest published build. Pragmatic Play lists Sweet Bonanza at 96.48% on its own site, which is the branded default rather than a promise that every operator runs it. The only figure that describes what you will be charged is the one on the in-game screen at your casino.
No. They are separate properties tested for separate reasons. RNG testing checks that each spin is independent of the last. RTP is the share of stakes the maths is built to return over the long run. A perfectly fair random-number generator and an 88% return sit in the same certified game without any contradiction.
It depends on the licence, and a floor is a backstop rather than a quality mark. Malta cut its online minimum from 92% to 85% on 28 May 2021. New Jersey requires a theoretical return of at least 83% under N.J.A.C. 13:69E-1.28A. Great Britain sets no statutory minimum for gaming machines at all — a low RTP is lawful there provided the figure is displayed.
On the same operator’s site, it should. Malta tells licensees that demo RTP must match the real-money setting, and the UK’s RTS 6 requires free-play games on the same website to use the same rules and represent the real-money prize distribution. A supplier’s own demonstration site is different: RTS 6 warns those can run higher-than-normal returns where that is labelled.
Often not. Bonus buy and extra-bet modes usually carry their own certified figure, because the player is buying a different slice of the paytable. Hacksaw publishes a bonus-buy menu on both Wanted Dead or a Wild and Densho. Where a help screen lists more than one percentage, read the line that matches the bet you are about to place.
No. What people describe that way is volatility, which is how lumpy the path to a game’s average is rather than a change in that average. A high-volatility 96% game can empty a bankroll faster than a low-volatility 94% game because the 96% is concentrated in rare features. A certified RNG keeps each spin independent, so no run of results makes the next spin more or less likely to pay.
That this configuration’s maths match the RTP it advertises, that the return is achievable across base game and features, and that the RNG is behaving. It does not say the casino picked the highest certified build, and it is not a judgement of value. An 88% configuration can be fully certified and legally offered.
Playing within a limit
A 96% slot is still a 4% house game, and checking the RTP is price comparison rather than an edge. Reading the info screen tells you what a game charges; it does not change the direction the arithmetic runs. Set deposit, loss and time limits before you start. CasinoDrop’s responsible gambling guide covers the tools, and How We Rate sets out how we treat published game data across the site.
- Read it, do not measure it. The info screen on the game at your casino is the only practical check. Measuring the same number yourself would take 153,648 rounds on the easiest game we can measure, and longer on a slot.
- The instance is the product. One title can ship at 96.38%, 94.55%, 92.33% and 88.42%. The operator chooses. A review page’s figure describes the studio’s default, not your lobby.
- A session tells you nothing about RTP. On the Commission’s own example the 95% band after 500 plays is about ±49 percentage points. Our own tape put a 99.00% game anywhere from 97.60% to 108.00% across 2,500 rounds.
- Certification is not endorsement. A certificate says the maths match the advertised figure and the RNG behaves. An 88% build passes that test exactly as a 96% build does.
- A floor is not a typical value. Malta’s online minimum is 85%, New Jersey’s slot floor is 83%, and Great Britain sets no statutory machine minimum. Legal and good value are different questions.
- CasinoDrop Stake Crash tracker — first-party, 21,499 rounds stored from 11 Aug 2026; the 2,500-round detail window used here covers 2026-08-17 to 2026-08-18
- Stake Crash round feed — first-party, the JSON behind every measured figure: settled multiplier per round plus the published per-target probability model
- CasinoDrop Stake Originals simulator — first-party, replays a real seed pair with settlement, running RTP and a profit-and-loss curve
- How to calculate return to player (RTP) — primary, Gambling Commission: the theoretical-versus-actual split and the quoted RTS 3C wording
- Live RTP performance monitoring of games of chance — primary, Gambling Commission: the 91.68% / £1,200,000 turnover worked example
- What volume of play should be achieved before measuring actual RTP? — primary, Gambling Commission: volatility 5.6 and the ±4.91 / ±1.74 / ±1.10 points this article’s band curve reproduces
- RTS 3 — Rules, game descriptions and the likelihood of winning — primary, Gambling Commission: the disclosure requirement and where the information must sit
- RTS 6 — Result determination for play-for-free games — primary, Gambling Commission: demo games must represent the real-money prize distribution
- Return to player: how much gaming machines payout — primary, Gambling Commission player guide: machine labelling, and no statutory minimum RTP
- Approved test houses — primary, Gambling Commission: the current list of testing laboratories
- Testing strategy for compliance with remote gambling and software technical standards — primary, Gambling Commission
- The MGA streamlines the return to player percentage — primary, Malta Gaming Authority, 28 May 2021: the 92% to 85% floor change
- Player protection — primary, Malta Gaming Authority: game rules one click from play, and demo RTP matching real money
- N.J.A.C. 13:69E-1.28A — primary, New Jersey Administrative Code: the 83% theoretical payout standard for slot machines
- Frequently asked questions — primary, New Jersey Casino Control Commission: Division testing against the payout standard
- Wanted Dead or a Wild — game data — primary, Hacksaw Gaming: the four certified RTP configurations and the bonus-buy menu
- Densho — game data — primary, Hacksaw Gaming: the same four-band pattern
- Sweet Bonanza — primary, Pragmatic Play: the single published 96.48% default
- RTP percentage testing — primary, eCOGRA: what an RTP test covers
- Ensuring fair play with RNG testing and eCOGRA certification — primary, eCOGRA: RNG testing as a separate job from RTP
- Cutting RTP in the UK: a margin fix or a market misstep? — reported, iGaming Business, April 2026: the Helen Walton and Eyal Loz remarks on RTP as a pricing lever
- Viking Runecraft 1000 — reported, iGaming Business, August 2026: multiple RTP configurations listed as a product feature
