One year, and 458,492 characters that were not there before
Two captures, fifteen months apart
The operator's own text bundle held 444,165 characters when an archive took it on 24 April 2025 and 902,657 characters when it took it again on 9 August 2026. The document did not grow by that much. It was replaced, and the replacement is 2.03 times the size of what it replaced.
Most write-ups of an operator treat its terms as a single settled object: the terms say this, the terms allow that. Two dated copies of the same address show why that habit is wrong.
The bundle that feeds this operator's site with its own text was archived on 24 April 2025 at 444,165 characters across 8,919 strings, and archived again on 9 August 2026 at 902,657 characters across 16,077 strings. Both were opened and read on 2 September 2026.
Fifteen months, and the document is 2.03 times its former size.
Did anything survive the rewrite?
Not the sentences. Searching the later file for the distinctive wording of the earlier Terms of Service returns nothing: that document has left the bundle entirely rather than being amended inside it.
Not the numbering either. The earlier text is a Terms of Service in eight numbered sections with unnumbered paragraphs beneath each heading. The later one is a Terms and Conditions in seventeen sections with decimal sub-clauses — 2.2, 3.4, 4.1, 5.2, 5.3. A reader who noted a clause number in 2025 is pointing at a paragraph that has been replaced and renumbered underneath the pointer.
That matters more than it sounds. A contract you can cite is a contract you can argue with, and the renumbering quietly disposes of every citation anybody had made.
The one substantive count that can be run across both copies is the restricted-territory list, which stood at 24 named places in the earlier file and 37 in the later one. The names, and the direction of that movement, are set out on the territory clause itself.
What does a year of edits look like from the reader's side?
It looks like nothing, which is the problem.
Neither copy contains an undertaking to tell an existing account holder that the document beneath them has been rewritten, and no notice of the rewrite appears anywhere in either bundle. The two files exist because a public archive happened to take them on two particular days. Had it not, the comparison on this page could not be made by anyone outside the company.
So the practical asymmetry is worth stating plainly. The operator can replace the whole text at a moment of its choosing, and the reader's only defence is a copy of their own with a date on it.
Which of the three dates can be trusted?
The current bundle carries three “last updated” strings, and each one behaves in a different way.
The Refund Policy states its date plainly: last updated 1st October 2025. That is a document dating itself, and it is the strongest of the three.
The Terms and Conditions carry a 1st October 2025 string that arrived in the export separated from the body. It is the only unattached date of its kind in the bundle and it reads as the date of the terms, but nothing printed beside it confirms the attachment. Our record flags it as a reading rather than as a fact, which is the same rule applied to every deduced clause number on the reading method.
The Privacy Policy is the third, and it reads “Last updated [27.11.2019]”. The square brackets are the operator's own. They are the placeholder marks of the template the document was drafted from, still standing in a copy captured in August 2026, seven years after the date they enclose.
A date nobody removed the brackets from is a date nobody checked.
What is a policy dated October 2025 still describing?
The Refund Policy is the sharpest illustration of what a stamp on a document is worth, because it is current text rather than an old copy.
In the file captured on 9 August 2026, carrying its own October 2025 date, that policy still describes trades in Steam skins, bots that deliver them, commission taken in skins, and withdrawals to QIWI, WebMoney and Yandex Money wallets.
Its substantive rules deserve reading on their own account, whatever the surrounding text describes. A refund is considered only within twenty-four hours of the deposit, or within thirty days on a sworn statement that another person or a minor used the account. Funds already played cannot be refunded. Where notarised identification is demanded and goes unsupplied for thirty days, the policy states that the account shall be closed and the balance forfeited, with the decision final and beyond appeal.
That last sentence is the one to carry away. A document whose payment methods are years out of date still contains a live forfeiture rule with a thirty-day fuse on it.
Which sentences left the document, and what replaced them?
Two protective sentences present in the April 2025 copy are absent from the edition in force, and this site declines to carry them forward as though they still applied.
One closed a minor's account immediately on identification. The other set a minimum self-exclusion period of six months during which reopening was impossible; the current text names durations of one, five and eight days in a help-centre answer instead, with anything longer routed through support.
Both are recorded here as a difference between two dated files, without any claim about why the change was made, because neither file explains itself.
Which clause did the year leave standing?
Clause 4.1 of the edition in force is the single most consequential sentence in the whole document, and it is not a promotional term.
It requires the value of every deposit to be wagered at least once before a withdrawal is requested, states in its own words that this applies to all deposits including sports bets rather than only to promotional funds, and permits the operator to delay or deny a request that arrives before the wagering is complete.
One times the deposit, promotion or no promotion.
That multiple is the only hard arithmetic commitment anywhere in the general terms, which is worth holding against the bonus section, where clause 6 says each promotion carries its own requirements disclosed with the offer and publishes no multiple at all. What the payout section around clause 4.1 does and does not commit to is taken apart on the payout clauses.
What can a reader do about a document that moves?
Three habits, none of them expensive.
Save a dated copy of the terms on the day you open an account, because a comparison needs two files and the operator supplies only one at a time. Note the clause numbers you rely on, and expect them to move rather than to hold. And treat the absence of a change notice as normal rather than as evidence that nothing changed.
The full comparison between the two captures, section by section, is set out on the two editions, and the twelve operators read on the same rules sit on the front page.