Razor Shark Slot RTP: Eleven Builds, One Game

Two on the Page, Eleven in the Code

The studio does not publish a return for this slot. It publishes two. On its own game page the field is labelled RTPs, in the plural, and under that label sit 96.70% and 94.06% with nothing to say which build is which, or where either one runs.

Hover the field and Push Gaming explains itself: "RTP's may vary per casino." Read that twice. The studio is telling you the figure belongs to the casino rather than to the game.

Then we opened the client bundle the demo loads, and the two became eleven. A variant table inside the code lists six returns on the base line and five more on a second line the file flags as isSMM. Top to bottom, that is a spread of 11.0 percentage points on one title.

The Eleven Returns Carried in the Client Bundle

Figures current as of 9 Oct 2026; the operator's live terms prevail.

Six of the base builds answer on launch addresses of their own, numbered in sequence from razorshark through razorshark-06. Invented identifiers return 404, which is the control that makes the live ones mean something. One address we did not expect, razorshark-07, answers normally yet appears nowhere in that table, so its return is simply unknown and we are not going to guess it.

Where the Game Shows Its Own Figure

We loaded the studio's public demo on 9 October 2026 and went looking for the number that tooltip promises. It sits at the foot of the GAME INFO screen in one flat sentence: "The theoretical average return to player (RTP) is 96.70%." The same screen stamps the build, Server 2.8.0 and Client 13.2.6, which matters more than it sounds, because a return belongs to a build rather than to a name.

That is the top of the eleven. It tells you what Push Gaming ships by default and nothing at all about which variant your lobby has loaded.

The Lower Number Is Italy's

94.06% is not a theoretical twin that nobody uses. Sisal, a concessionaire of Italy's ADM, must publish a rules document for every title it runs, and the closing line of the one for this game reads "Il valore dell'RTP teorico calcolato su base statistica e pari al 94,06 %". An Italian player is on that build.

Two details deserve keeping. The comma is Italian punctuation rather than a different number, and the document is a regulator-mandated filing rather than marketing copy, which makes it the strongest single piece of evidence on this page.

It also lines up with the code. 94.06% is build 03 in the variant table, so Sisal is not contradicting the studio at all; it is naming which of the eleven its concession runs, which is exactly what every other operator declines to do.

Who Printed Which Return

Britain: Two Prints, Three Silences

Here the pattern inverts, and it surprised us. British licensees do not disagree about the figure; they disagree about whether to show it. MrQ, on Gambling Commission account 60629, prints 96.70%. Grosvenor Casinos, account 57924, prints 96.70%.

Two separate companies, the same decimal.

Then the quiet half. Virgin Games and Jackpotjoy both publish nothing, and that is one silence rather than two, because both brands sit on Gamesys Operations account 38905. PlayOJO, account 39326, does keep a return field, and what the field contains is the literal string "undefined%". Its neighbours are broken in the same way: the key-facts block also prints a minimum bet and a maximum withdrawal as "€NaN.undefined", with only the British regulatory stake cap filled in properly. Two broken money fields and a broken return, on a live page.

We hunted for 94.06% at every British licensee we could read and found it at none of them. MrQ is the fullest witness here and also the most fragile one: 15 of the 38 MrQ game pages with a readable grid publish a maximum stake above Britain's five-pound cap, so its 96.70% comes quoted with that caveat attached.

The Affiliate Figures Were Real

Affiliate pages circulate two different lists. One is a four-way set, 96.70, 94.70, 94.06 and 90.52; the other, on slotcatalog and bigwinboard, runs 96.70, 95.05 and 90.52. We went in expecting to knock the odd figures down, and the code knocked us down instead. 95.05% and 90.52% are genuine builds, numbered 02 and 05 on their own launch addresses.

So the correction runs the other way. Neither list was invented; both were incomplete, and the fullest of them names three of the eleven builds.

One figure in it we still cannot place is 94.70%, which the variant table does not contain at all. Treat it as unverified rather than as debunked.

Nobody Is Forced to Publish an RTP

Britain has no rule compelling a licensee to show a return on a game page, which is why three of five say nothing. Sweden runs further in the same direction: the word RTP appears zero times in the regulator's published requirements for licensees, and so does aterbetalningsgrad. Expekt, a Swedish licensee, publishes a full specification table for this slot, with provider, reel layout and line count, and no return anywhere in it.

Italy is the exception that makes the point. Where a regulator obliges the operator to file the number, the number appears, and it turns out to be the lower one.

How to Count the Builds Yourself

Nothing here needs special access. Open the client, take the bundle name out of its index.html, fetch the bundle and grep it for a return-shaped value. The eleven come straight out of that one command, and anyone can run it against the same public demo we used.

For the rest we read the studio's game page, then each operator's own game page rather than a review of it, and confirmed every licensee account number in the Gambling Commission register instead of in a site footer. Where a page was unreachable from our address, we logged it as unreachable, because an operator we cannot read is not an operator that publishes nothing.

One caveat we cannot engineer away: operators edit these pages without notice, and a figure true on 9 October 2026 can be stale within a month. Read the loading screen of the game you are actually in, which is precisely what the studio recommends.