Editorial · Method
How we test online casinos
A test here starts with a real registration and ends with real money arriving in a real bank account. In between sit 42 single criteria in six groups whose weights are written down before the first click. This page is the procedure rather than the catalogue: the order of the steps, what is put on a clock and what is only observed, the four things that cannot be judged from outside, how often a report is looked at again, and what happens when we get something wrong.
By the Casinohafen deskUpdated: 20 October 2026Checked: September 2026
In brief
- Only operators listed on the whitelist of the Gemeinsame Glücksspielbehörde der Länder (GGL) at the editorial deadline enter the test. No German permission, no test — not even as a negative example in the same table.
- Testing is done as an ordinary customer, with our own money, without telling the operator in advance and without a test account.
- Every test begins with a real deposit and ends with a real withdrawal. Durations are taken from two timestamps, never from a figure the operator states about itself.
- Four areas cannot be assessed from outside — random number generators, the early-detection algorithm under § 6i(1) GlüStV, the individual conditions attached to a permission, and unpublished game payout rates. They are marked unassessable and left out of the grade rather than guessed at.
- Weights are fixed before the measurements and not moved afterwards; a group grade below 4.0 withdraws the recommendation whatever the weighted mean says.
- The site is currently run as a content-only publication: there are no ranking tables, no score badges, no bonus blocks and no links to play anywhere on it.
- Single criteria
- 42 in six weighted groups
- Timestamps per duration
- 2 order sent, money credited
- Unassessable areas
- 4 marked, not scored
- Whitelist check
- monthly independent of the test cycle
The question that comes before the method
Nothing is measured until one thing is settled: is the operator listed on the GGL whitelist at the editorial deadline? The Gemeinsame Glücksspielbehörde der Länder is the joint gambling authority of the sixteen federal states, and the whitelist it keeps under § 9(8) of the Glücksspielstaatsvertrag — the interstate treaty on gambling, usually shortened to GlüStV — is the only place that answers the question.
This is not a judgement about quality, it is a precondition. Without a German permission there is no competent supervisory authority, no described complaints route, and no set of conditions to test compliance against. A grade in that situation would suggest a reliability that does not exist, and a row in a table is a form of attention we are not willing to hand to an unlicensed offer. How the list is read, and the four ways of misreading it, is set out under the German gambling licence.
The check runs monthly rather than once per cycle. If a permission falls away, the report goes with it, mid-cycle if necessary. It is the only single fact that can remove a page from the catalogue on its own.
Why the order of the steps is the method
The sequence in the table below is not administrative tidiness. It reproduces what an ordinary customer does, and it puts the uncomfortable steps at the end on purpose, because an operator's real character shows when money leaves the house rather than when it arrives.
Registration comes first, and what is recorded there is not only how long it takes but what is asked: which fields are mandatory, how age is checked, and at what point the deposit limit is put in front of the reader. Then the limit itself is set, lowered and raised — three separate actions, because § 6c GlüStV treats them differently and an operator's handling of that asymmetry is one of the more revealing things about it. A reduction has to take effect at once; an increase has to sit through a seven-day waiting period. An interface that blurs either of those is a finding.
Only then comes the deposit, with our own money, and play in the ordinary catalogue. Protective functions are opened and counted next: the panic button, the break function, the pointer to counselling, the route to an OASIS self-exclusion. Support is contacted, identity verification is completed, and the withdrawal comes last, because everything before it can be pleasant while that one step is not.
Operators are not told any of this is happening. An announced test measures how a company behaves towards testers, which is not the question anyone has.
What is measured with a clock and what is only counted
Duration is taken from two timestamps, never one. For a withdrawal the first is the moment the request is sent and the second is the moment the money is credited to the destination account. What lies between them is what a person actually waits, which is a different quantity from the internal processing time an operator quotes. For a support enquiry the first stamp is on sending and the second on the first substantive reply; an automated acknowledgement is not a reply.
Everything else is counted or observed and written into the log: which documents were demanded and at what moment, how many steps lie between the home page and the deposit form, how many between the home page and the limit page, whether the limit reduction was immediate, whether the increase genuinely waited, and whether the panic button is the clearly recognisable and unambiguously labelled control that § 6i(3) GlüStV requires or a grey line in a submenu.
Four things are not done. Values are not derived from the operator's own statements. Single observations are not extrapolated into averages. Measurements are not carried over from an earlier cycle into a new one. And a missing measurement is not filled with a plausible number. Where only one observation exists, the report says so next to it.
Three labels, and one of them is an empty field
Every figure in a report carries one of three labels: measured, stated by the operator, or not assessable. The second is never converted into a score. The third stays empty, and the empty field is itself information — an invented number is not.
The four things that cannot be judged from outside
These are named rather than quietly omitted, because their absence would otherwise look like carelessness.
The first is the random number generator behind each game. Every virtual slot machine needs its own prior official approval under § 22a(1) GlüStV, and that examination happens at the authority and at accredited testing bodies. An editorial desk cannot repeat it and should not pretend to.
The second is the early-detection system required by § 6i(1) GlüStV. It is algorithmic, its thresholds are not public, and a test from outside could at best record that nothing happened — which proves nothing at all. We record that the obligation exists and note any visible intervention.
The third is the content of the permission itself. A German permission is an administrative act with individual ancillary conditions, and those conditions are not published. We can check whether the generally known obligations are visibly implemented; we cannot check what a particular operator was additionally told to do.
The fourth is the payout percentage of an individual game where the operator does not publish it. A few hundred rounds in a test session is statistically worthless, and presenting it as a measurement would be the most convincing kind of fabrication. Separately, a criterion that does not apply at all — poker criteria at a slots-only house — is taken out of the group calculation rather than scored zero, which is not the same thing as a bad mark.
From 42 scores to one verdict, without a thumb on the scale
Each criterion is scored from 0 to 10 against a written expectation. A group grade is the unweighted arithmetic mean of its criteria; the verdict is the weighted mean of the six group grades. The weights — 25 per cent for licence and supervision, 20 each for payouts and player protection, 15 for the game range, 12 for costs and limits, 8 for service and transparency — are fixed before testing begins and are not moved afterwards. The reasoning behind that distribution is set out under our test criteria.
One rule overrides the arithmetic. If a single group falls below 4.0, the recommendation is withdrawn regardless of the weighted total. Good service does not compensate for an unclear permission situation, and a broad game range does not compensate for a payout process that fails.
There is no rounding upwards, no discretionary adjustment and no bonus point at the end. When a result looks wrong, that is a reason to repeat the measurement, not to adjust the figure.
One consequence of the site's current shape belongs here plainly. Casinohafen is published as a content-only site: it carries no ranking table, no score badge, no bonus block and no link that leads to a deposit. The method and the catalogue are published in full, and individual reports describe what was found, but nothing on these pages is arranged to move a reader towards an operator. That is a deliberate restriction on the output, not on the measurement.
Rechecks, corrections and what we do with a complaint
Every report is rechecked at least once per edition. The recheck does not repeat the whole catalogue; it repeats the part that can have changed — permission status, payment routes, fees, withdrawal duration, limit handling, terms.
On top of that come rechecks for cause. A change in the law is a cause, and 2026 supplied one: since 1 July the GGL has allowed staged increases of the maximum stake under the opening clause in § 22a(7) sentence 2, to 3 euros from the age of 21 and to 5 euros after a 90-day qualifying period without problem-gambling indicators and with behavioural monitoring, an arrangement limited to the end of 2027. A change in an operator's conditions is a cause. So is an entry appearing in or dropping out of the whitelist, and so is a cluster of readers reporting the same thing.
Language corrections are made silently. A factual error — a wrong figure, a wrong deadline, a wrong section number — is corrected and noted at the foot of the page with the date and the previous wording. An error that carried a judgement leads to the judgement being withdrawn: the grade is suspended until the affected group has been measured again.
Notices are accepted from anyone, readers, operators and competitors alike, and are treated the same way. What helps is anything that evidences the point: a link, a date, a timestamped screenshot, a file reference. A request aimed at removing criticism is handled like any other notice — if the criticism is wrong the text changes, and if it is right the text stays.
No operator sees a report before publication and none receives an advance copy. There is at present no monetisation of any kind on this site, which is described in detail under about this site.
Two limits on all of this are worth stating. A test describes a moment and a perspective — one account, one payment route, one amount, one week — so a different figure measured by somebody else is not a contradiction. And a test is not advice to play: what is assessed here is the quality of the handling. Anyone asking the prior question will find more useful instruments under responsible gambling; counselling is free and anonymous from the Bundesinstitut für Öffentliche Gesundheit (BIÖG) on 0800 1 37 27 00, Monday to Thursday 10:00–22:00 and Friday to Sunday 10:00–18:00, with self-tests at check-dein-spiel.de. The helpline is answered in German.
Step by step
| Step | What happens | Who acts | Time frame |
|---|---|---|---|
| 01 | Check the permission | Comparison against the GGL whitelist under § 9(8) GlüStV. No entry, no test. | Test desk |
| 02 | Open an account | Registration as an ordinary customer, with no advance notice to the operator. Mandatory fields, age check and the moment the deposit limit is requested are recorded. | Test desk |
| 03 | Set and change the limit | The deposit limit is set, lowered and raised. Whether the reduction is immediate and the increase waits out the seven-day period under § 6c(3) is checked. | Test desk |
| 04 | Deposit and play | A real deposit with our own money and play in the ordinary catalogue. Fees, routes and limits are recorded. | Test desk |
| 05 | Open the protective functions | Panic button, break function, pointers to counselling, route to an OASIS self-exclusion. The number of steps to each is counted. | Test desk |
| 06 | Contact support | One identical question to every operator. Timed to the first substantive reply, not to the acknowledgement. | Test desk |
| 07 | Verify identity | Which documents were demanded, when they were demanded and how long the release took. | Test desk |
| 08 | Withdraw | A real withdrawal to a real bank account. Two timestamps: order sent and money credited. | Test desk |
| 09 | Score | 42 criteria on the 0 to 10 scale, group grades, weighted verdict, and the 4.0 threshold applied. | Test desk |
| 10 | Fact-check and release | Every figure checked back against the logs, final edit, release by the editor in charge. | Fact-checking and chief editor |
| 11 | Recheck | At least once per edition and additionally for cause; permission status monthly. | Test and regulation desks |
The catalogue and its weights
| Test group | Weight | Criteria | What it covers |
|---|---|---|---|
| Licence and supervision | 25 % | 9 | GGL whitelist, scope of the licence, seat of the operator, complaints route |
| Payouts and processing | 20 % | 7 | processing times, verification effort, chargebacks, minimum amounts |
| Player protection and OASIS | 20 % | 8 | OASIS connection, limit management, panic button, pointers to counselling |
| Game range | 15 % | 6 | number of approved games, breadth of studios, live offer, table games |
| Costs and limits | 12 % | 6 | fees per payment method, stake limits, cooling-off, wagering conditions |
| Service and transparency | 8 % | 6 | availability, language, imprint, clarity of the terms |
Frequently asked
Are operators told in advance that they are being tested?
How exactly is withdrawal time measured?
Why do you not test the random number generators?
Do bonus offers count towards the grade?
How often is a report tested again?
What happens when you get something wrong?
Can I reproduce your results myself?
Sources
- Glücksspielstaatsvertrag 2021, § 6c — cross-operator deposit limit, immediate reduction and seven-day waiting period before an increase — <a href="https://www.gesetze-bayern.de/Content/Document/StVGlueStV2021-6c" rel="nofollow noopener">gesetze-bayern.de</a>
- Glücksspielstaatsvertrag 2021, § 6i — early-detection system, safe server system and panic button.
- Glücksspielstaatsvertrag 2021, § 22a — approval required for each individual game, maximum stake and the opening clause in subsection 7 sentence 2.
- Gemeinsame Glücksspielbehörde der Länder — whitelist under § 9(8) GlüStV, the admission condition for every test — <a href="https://www.gluecksspiel-behoerde.de/de/fuer-spielende/uebersicht-erlaubter-anbieter-whitelist" rel="nofollow noopener">gluecksspiel-behoerde.de</a>
- Gemeinsame Glücksspielbehörde der Länder — FAQ on the deposit limit and the maximum stake — <a href="https://www.gluecksspiel-behoerde.de/de/fuer-spielende/informationen-fuer-spielende-faqs/faq-einzahlungslimit-und-hoechsteinsatz" rel="nofollow noopener">gluecksspiel-behoerde.de</a>
- Bundesinstitut für Öffentliche Gesundheit (BIÖG), counselling helpline 0800 1 37 27 00 — <a href="https://www.check-dein-spiel.de/" rel="nofollow noopener">check-dein-spiel.de</a>