How a similarity score is built here, weight by weight

Seven checks, 100 points on paper and 85 reachable in practice, applied to 50 operators drawn from a pool of 78 records. Of the 350 cells that produces, 54 stayed empty, and an empty cell is filed under its own heading instead of being paid a zero.

What each operator publishes about itself: the company named on the licence, the permit number and where it was read, and how many countries the operator excludes in its own terms. Similarity here is a count with a denominator printed beside it — BitStarz names 74 countries, and an operator's overlap is stated as a share of those 74. An operator whose own list was not read carries nothing to compare rather than a low score, because an unread list is not evidence of difference. An empty cell means the document was not read — it is not a zero, and it is not a permission.
Casino	Operator	Licence	Countries barred	Withdrawal limit	Visit
VaveOur pick	Latcas B.V.	Curaçao Gaming Authority	list not read	from 50,000 USDT in instalments cl. 8.8	Visit
BitStarz	Gareton B.V.	Curaçao Gaming Authority	74	not published
Oshi Casino	Novatrix SRL	Anjouan Gaming Board	224	not published
PlayAmo	Novatrix SRL	Anjouan Gaming Board	217	not published
Bitdreams	Hollycorn N.V.	Antillephone N.V.	216	not published
7Bit Casino	not published	Curaçao Gaming Authority	216	not published
Bitkingz	Novatrix SRL	Anjouan Gaming Board	218	not published
Mirax	Scores55 Tech B.V.	Curaçao Gaming Authority	220	not published
Slotum	Novatrix SRL	Anjouan Gaming Board	216	not published
Metaspins	Metaspins N.V.	Curaçao Gaming Authority	58	not published
CryptoLeo	Uno Digital Media B.V.	Curaçao Gaming Authority	74	not published
Bets.io	Bets Entertainment N.V.	Curaçao Gaming Authority revoked	42	not published

Seven checks, and where each weight came from

The score behind every ranking here adds seven weighted properties together, totalling 100 on paper. Weights follow one rule: the harder a property is to imitate, and the more documents it takes to settle, the heavier it is.

Territory overlap takes 30 because it is the only check that compares a long list against a long list. An operator can copy a bonus shape or a lobby in an afternoon; a restricted-territory clause reflects which payment rails, which supplier contracts and which regulator a business is actually working under, and it takes two full documents to compare. The licensing authority takes 15, and the company named on the permit takes another 15, because a shared company is the strongest form of resemblance there is and the rarest to establish honestly.

The remaining 40 splits four ways at 10 each: original games built for the operator, a published VIP tier, a casino with the sportsbook left out, and a verification posture matching the BitStarz one, which describes the check and publishes no amount. Sports betting subtracts rather than adds. A reader looking for something resembling BitStarz is looking for a casino, and an operator whose front page opens on football markets is offering a different product, whatever else it matches on.

The Wild.io lobby with the coin selector open — Eighth on this page at 58 points, and one of the twenty operators that publish a countable coin list: twelve currencies, named on the cashier screen itself.

The ceiling is 85, and the arithmetic says so

Across all 50 rows, zero operators name Gareton B.V., the company printed on the BitStarz permit read on 26 August 2026. The 15 points reserved for a sister brand were therefore available to everybody and claimed by nobody, so the practical ceiling on this dataset is 85. Mirax at 73 stands at 73 of 85, and any reader converting these totals into percentages of 100 would be measuring against a maximum the dataset cannot reach. Those unclaimed points are recorded as unclaimed across the scored sites like BitStarz.

That is disclosed rather than repaired. Dropping the check would flatter the scores; renormalising to 85 would hide the fact that the strongest form of similarity turned up zero times in 50 attempts. The zero is itself a finding, and it survives here in the shape of a gap.

Why the denominator is printed beside the number

Similarity here is a count with its denominator attached: 72 of 74, 57 of 74, 38 of 74, all measured against the reference list. Both halves travel together for two reasons.

The first is direction. The fraction asks what share of the reference clause an operator reproduces, and it stays silent about the rest of that operator’s own document. Oshi Casino scores 72 of 74 while excluding 224 countries in total, so 152 entries of its own clause sit outside the comparison entirely. A bare percentage would let a reader believe the two lists agree, when what was measured is that one list contains most of the other.

The second is arithmetic hygiene. Scoring over filled fields, rather than over a fixed set of checks, promotes the operators who disclose least: an operator publishing one readable thing scores full marks on one check out of one. Every row here is divided by the same seven checks and the same 74 countries, whether it filled those cells or left them empty.

An empty cell has its own word

Four states can produce a low number, and only two of them are about the operator. The clause may say something that fails to match. The clause may say something that partly matches. The document may exist and stay shut to us. Or the property may be one nobody publishes at all.

Merging the third state into the first is the failure this site is built to avoid, because it converts a gap in our reading into a verdict about a business. So a cell prints the word unread, and the reader is told which situation produced the total. The counts are these: 54 of 350 cells stayed empty; the territory check is empty on 21 rows and the company check on 16; the verification posture is empty on 16; the authority resolved on 49 rows of 50; and 25 rows are filled in on every one of the seven. Fifteen rows share a total of 25 points, and that cluster is a picture of our own reading rather than a family resemblance among 15 businesses.

Our pick How the top row was read Termsread 25 August 2026Registersearched on the domainChecksthe same seven as the fifty Open the casino 18+

The block for our pick on this page, drawn by us: the reels pass once and stop, nothing spins on a click, and the three lines are the reading behind that row rather than a promotion, which the operator rewrites on its own schedule.

Five places where this ruler bends

Every measure has edges, and printing them costs less than having a reader find them.

Names are matched as strings. The authority check compares the words an operator prints against the words BitStarz prints. Bitdreams prints Antillephone N.V., Rocketpot writes Curacao without its cedilla, and Wolf.bet writes out the government of the autonomous island of Anjouan in full. Each of those scores zero on that check, and a human reader comparing islands would score at least one of them differently.

Two checks read a tag rather than a clause. Original games and the VIP tier come from a short list of properties attached to each operator record, so a missing tag produces a zero where an unread document would have produced the word unread. Those two cells are the softest on the page, worth 20 points between them, and no reader should treat a zero there as a finding.

The comparison runs one way. It is anchored on the reference clause and never asks the reverse question. Mirax excludes 220 territories of its own, of which 70 fall inside the reference list, meaning 150 entries of Mirax’s clause take no part in the score that puts it first. A fraction of this kind reports containment, and a reader who hears equality in it has heard something the arithmetic never said.

A measured zero looks like an empty cell and differs from one. Four operators had their clauses read and came out at the floor: Bitsler and Bitcasino.io repeat 0 of the 74, Stake and Cloudbet repeat 1. Those four zeros are findings, arrived at by opening a document and comparing two lists, and they sit in the table next to 21 blanks that mean the opposite thing. Printing both as a bare 0 would have destroyed the distinction the site exists to keep.

Resemblance is separate from standing. The permit number Bets.io publishes, OGL/2024/1178/0479, resolved in the register on 26 August 2026 showing the status Revoked, and its similarity total was left exactly where the seven checks put it. One question asks how much an operator resembles the reference brand; a different question asks whether a business is in good order with its regulator. A single figure answering both would answer neither, so the status is reported beside the score rather than inside it.

What the editorial choice moves is position, never wording, and the easiest way to confirm that is the row itself: an operator picked out for visibility would hardly have chosen to publish a jurisdiction with zero permit numbers attached to it, and yet that is what the cell reports, because that is what the document showed on 25 August 2026.

When these numbers expire

Terms change on the operator’s schedule and with no notice to anybody. What binds a player is the document at the operator, rather than the summary of it here, which is why a date sits beside every figure on this site: a dated clause number can be opened, compared and seen to have shifted, and an undated one cannot. Corrections are made or explained, and the person who makes them is described on who writes this, with the scored table itself on the fifty operators in order.

Questions people actually type

Why is the denominator printed on every row?

Because a share computed over filled fields rewards the operators who publish least. If a score were divided by the cells an operator happened to fill, a business that publishes one readable thing and withholds the rest would finish above a business that publishes everything. The denominator here is fixed at seven checks for all 50 rows, and the territory figure is written as a fraction of 74 so a reader sees both halves at once.

Why is an unread clause a category rather than a zero?

A zero on the territory check would say the operator repeats none of the BitStarz list. An unread clause says our reading stopped, which is a statement about us. Those two situations look identical inside a total and mean opposite things, so the cell prints the word unread and the row keeps whatever the other six checks earned it. Twenty-one rows of 50 sit in that state on the strongest of the seven.

Does our pick get a better score?

Vave lands in the lower half of the ranking, 27th, on 35 points, and the cells that produced that number were filled on the same seven checks as the rest. Its territory clause went unread, which leaves the heaviest of its seven cells blank, and it prints a jurisdiction with zero permit numbers attached — its permit OGL/2024/1676/0905 was found in the Curaçao register on 2 September 2026, not on its own pages, and the cell records what the operator publishes. Its place at the head of the table and the link out of this page are an editorial choice and are marked as one; the wording of a cell follows the documents, not that choice.

How often do these figures go stale?

Operators revise terms whenever they choose and owe nobody an announcement. Every figure here therefore travels with the date it was read, most of them 26 August 2026, and a clause that has moved since can be seen to have moved. A rescore takes one pass of the same seven checks over the same documents, and it changes the order whenever the documents have changed.