Same slot, different RTP: 730 of 3,761 games ship in more than one version

What we counted, and what we found

730 games, 15 studios, one frozen dataset

The trade press covered this subject through 2026 without publishing a single figure or a single link to data. So we counted. Of 3,761 games where we hold a confirmed return figure taken from the studio's own pages, 730 (19.41%) are published in more than one certified version. That is roughly one game in five.

How far apart the versions sit

A second version is only interesting if it is meaningfully worse, and it usually is. For 676 of these games the worst available build sits below 94 percent. For 434 it sits below 90 percent. A player who meets the low build of a title is not playing a slightly different game, they are playing a materially worse one under the same name.

Why we will not give you one industry average

The breakdown by studio

5 studios do this, 10 do not

The practice is concentrated, not universal. 10 of the 15 studios we track publish exactly one return figure for every title they make. Anyone writing that the whole industry is quietly cutting returns should read this table first.

Play'n GO publishes a tier list

The clearest case is not hidden at all. Play'n GO's own content layer carries named fields, rtp84, rtp87, rtp91, rtp94 and rtp96, labelled "RTP - 84" and so on, alongside a defaultRtp. 373 of their 387 titles carry more than one, with a mean spread of 11.97 pts.

The ladder that repeats 116 times

The same tiers recur across the catalogue, which is what a configuration menu looks like rather than a per game accident. The most common ladder, 84.2 / 87.2 / 91.2 / 94.2 / 96.2, appears on 116 separate games.

For all 373 of their multi version games, the studio's own defaultRtp field equals the highest tier. That settles how to read any advertised figure from them. The marketing number is the best case, and every rung below it is a build an operator is free to deploy instead.

Yggdrasil, Hacksaw Gaming and Push Gaming

The other three look nothing alike. Yggdrasil publishes several versions for 91.18% of the titles where they publish anything at all, with a median spread of 5.50 pts. Hacksaw Gaming does it on every title we hold and their spread is wide, 8.41 pts on average. Push Gaming also does it on every title, but their versions sit close together at 2.12 pts, which is a very different proposition for a player.

How many versions does a game get?

Can casinos adjust the RTP on slot machines?

What the operator actually chooses

This is the question people actually search, and the honest answer has two halves. No operator can reach into a running game and change the outcome of your spin. What an operator can do is decide which certified build to integrate in the first place. That decision happens in a commercial conversation months before you open the lobby, and the table above shows how much room that conversation has.

What the regulator requires

Certification covers each build separately, so a 91 percent version of a game is exactly as licensed and as audited as the 96 percent version. Nothing irregular is happening when an operator runs the lower one. What most rulebooks do not require is a clear label at the point of play telling you which build is loaded. That distance between what is legal and what is visible is the whole subject of this study.

What you can see from the player seat

Very little, without looking for it. The lobby tile will not tell you. Review sites usually repeat the studio headline figure, which we now know is the top rung. The one number tied to your session lives inside the game, and even that only helps if you know what the alternatives are, which is what our biggest RTP variants list is for.

Where do you find RTP on a slot machine?

The info panel is the only session specific number

Open the game, then open its info or paytable screen. Most builds print the return figure there, and that figure belongs to the build you are playing rather than to the title in general. It is worth checking the same game at a second operator, because a difference between the two is direct evidence that the title ships in more than one version.

Why the marketing figure is the ceiling

Studios advertise their best build. We can show that rather than assert it. Across every Play'n GO title with multiple versions, the studio's own default field points at the highest rung, without exception. Treat any single advertised percentage as the best case available, not as a promise about the copy in front of you. Our RTP calculator and house edge calculator show what the difference between two builds costs over a session.

What is the average RTP for slots?

What the ToolsGambling registry covers

We hold 3,761 games with a confirmed figure across 15 studios, collected from the studios' own pages rather than copied from other aggregators. The full registry, filterable by studio and by return, sits at the Slot RTP Registry, and the highest figures we hold are ranked on best RTP slots.

Why an average hides the tail

Headline figures cluster in the mid nineties, which makes any average look reassuring. The average is not where the risk lives. The risk lives in the gap between the best and worst build of a single title, and in the 434 games whose worst build drops under 90 percent. An average across titles cannot show you that, which is why this study reports spreads instead.

The games with the widest gap

Every row links to its proof

Each receipt returns the source URL, the time we fetched the page, the HTTP status and a safe text excerpt of what was there. You do not have to take the table on trust.

Tivoli Bonanza tops out at 88.52%

The extremes are worth sitting with. The best published build of Tivoli Bonanza returns 88.52 percent. Not its worst build, its best. A player who finds that game at 75.92 percent is not being cheated by anyone's definition. They are playing a different, equally certified product that carries the same name and the same artwork as the one they read about.

The other half of the story is silence

Yggdrasil leaves 250 of 557 blank

A game missing from this dataset has not been shown to have a single return figure. It may simply be that nobody published one. Yggdrasil leaves the return field empty for 250 of 557 titles, 44.88% of their catalogue, in their own public API.

Hacksaw pages carry no specification block

For 100 of 134 Hacksaw Gaming titles the game page has no specification section at all. We checked that against stored snapshots and against live pages in a browser before saying so. This is non publication rather than a hole in our collection, and it is worth reporting as a finding in its own right.

How ToolsGambling verifies every number

A receipt for every row

When we read a provider page we store the raw response and hash it. The hash travels with the value into the dataset, and anyone can exchange it for a receipt describing that document. We do not republish the raw HTML, because it carries third party scripts and somebody else's copyright, but the receipt proves what we read and when. The full procedure is written up in our methodology, and changes over time land in the change feed.

One page can prove several games

These 730 games rest on 454 distinct snapshots rather than 730, because a single provider page can embed a content layer covering more than one title. We would rather state that than let anyone assume a one to one relationship that does not exist.

We correct ourselves in public

While preparing this study we found 20 of our own Play'n GO records carrying tiers spliced in from a neighbouring game's record, and fixed them before publishing anything. BlackJack MH publishes a single figure of 99.54 percent, and we had been showing four lower rungs underneath it that Play'n GO never published for that game. Three further titles turned out to publish no return figure at all and were removed rather than filled in.

We document our own error on the same page that reports the finding, because a study about undisclosed numbers has no business hiding its own corrections.

RTP terms used on this page

RTP

Return to player. The share of total stakes a game returns over a very long run of play. It says nothing about a single session.

Certified version, or build

One configuration of a game that has passed testing on its own. A title can have several, each with a different return.

Spread

The distance between the highest and lowest published version of one title, measured in percentage points.

House edge

The mirror of RTP. A 94 percent build carries a 6 percent house edge over the long run.

Volatility

How lumpy the returns are. Two builds can share a return figure and feel completely different to play. Our volatility calculator covers this.

Evidence receipt

A public record of one fetched document: its source URL, time, HTTP status, size and a safe excerpt, addressed by the sha256 of the response.

Non publication

The provider holds the figure but does not put it on a public page. Different from a collection failure, and reported separately.

Take the data

Download and licence

The full set of 730 games is published under CC BY 4.0, which means you may use it commercially provided you credit the source. Every row carries the source URL, the snapshot hash and a link to its receipt.

How to reproduce the tables

Download the CSV, group by studio, count the rows and you will get the studio table above. Then take any hash from the snapshot_sha256 column and open /api/evidence/ followed by that hash to see the document the value came from. If you are working on a story and need a cut we have not published, ask.

Free RTP tools on ToolsGambling.com

Everything below is free to use and built on the same registry as this study.