To choose a casino originals provider, score every shortlisted supplier on the same evidence, not on demos and sales decks. Five areas decide most outcomes: the game math and its limits, fairness and certification, wallet integration and failure handling, product fit for your markets, and the rights, security and support terms in the contract. Ask for documents, test the games in a sandbox before you sign, and drop any supplier that fails a criterion you cannot live without, whatever its total score.
This checklist is for founders, Heads of Product, CTOs and procurement leads adding crash, dice, plinko, mines and similar games to a casino lobby. It covers game suppliers only. Choosing the casino platform itself, or the right sourcing model, are separate decisions, and this is not a comparison of named suppliers. If you are new to the format, start with what casino originals are and how the four types differ.
The short answer
- Score evidence, not claims. A criterion scores full points only when you have seen the document, report or test result that proves it.
- Weight what can hurt you. Math documentation, RNG integrity, exposure limits, certificate scope, wallet integrity, failure handling and the rights you receive carry the most weight.
- Use knockouts. Some criteria are pass-or-fail for your situation. A supplier that fails one is out, even with a high total.
- Test before you sign. Duplicate wallet calls, disconnects mid-round and late cash-outs show more about a supplier than any feature list.
- Compare price last. Price only matters among suppliers that pass.
Before you score: set your context
The same supplier can be a good fit for one casino and a poor fit for another. Settle three questions before you send the checklist:
- Where will the games run? A crypto-first lobby where players expect to verify rounds puts provably fair verification at the top of the list. A regulated market puts certificates, supplier licensing and local product rules there instead. Many operators need both.
- Which delivery model are you buying? With a hosted game, the supplier runs the game server, so uptime, support and escrow matter. With a source-code license, the code runs on your infrastructure, so the completeness of the code package and your own ability to operate it matter. The difference between these models is covered in build, license or buy.
- Who enforces limits? Maximum wins, payout caps and stake limits can sit in the game server, in your platform or in both. Agree on this before scoring, or two suppliers will answer the same question about different systems.
How the scoring works
Each of the 25 criteria gets a score and a weight.
Score.
- 0: not available, or the supplier cannot show any evidence.
- 1: available but only claimed, partial or undocumented.
- 2: available and proven: you have seen the document, report or test result.
Weight.
- 3: critical. A weakness here can cost money, players or a license.
- 2: important.
- 1: useful, but rarely decisive on its own.
Points for a criterion are the score multiplied by the weight. A critical criterion (weight 3) scored 1 gives 3 points out of a possible 6. The 25 weights add up to 50, and the top score is 2, so the maximum total is 50 x 2 = 100 points. That makes the total easy to read as a percentage.
Knockouts. Some criteria must score 2, whatever the total:
| Context | Knockout criteria |
|---|---|
| Every evaluation | 1 Math documentation, 6 RNG implementation, 11 Wallet integration integrity, 12 Failure and recovery handling |
| Regulated markets | Add 7 Certificate scope and 10 Supplier licensing |
| Crypto-first lobbies that promise provably fair games | Add 8 Provably fair verification |
| Crash and other shared-round games | Add 5 Max win and exposure controls |
Reading the total. These bands are a recommendation, not an industry standard:
- 80 to 100: strong fit. Move to contract.
- 60 to 79: workable. Turn every criterion scored 1 into a contract obligation with a date.
- Below 60, or any failed knockout: do not proceed with this supplier for this use.
Worked example. Supplier A totals 78 points but scores 1 on wallet integration integrity, because its rollback behavior is described in a sales email and not in the API documentation. Wallet integrity is a knockout in every evaluation, so Supplier A stops there until it documents rollback and passes the sandbox test. Supplier B totals 71 points with no failed knockouts. B goes to contract, with its weaker criteria written in as obligations.
Price is deliberately not in the sheet. Compare price and commercial terms only among the suppliers that pass.
The 25-point scoring sheet
A. Game math and limits
| # | Criterion | Weight | Evidence that earns a 2 |
|---|---|---|---|
| 1 | Math documentation | 3 | Written rules, paytable or multiplier tables, RTP, hit frequency, volatility and maximum win for every game, in a document your team can review |
| 2 | Simulation results | 2 | A simulation report comparing observed and theoretical RTP for each mode over a stated number of rounds, with the method described |
| 3 | RTP for every mode | 2 | RTP stated for each risk level, board size, profile or exit point, not one headline figure |
| 4 | Math configurability | 1 | A list of operator-selectable math profiles or RTP variants, each with its own approved model, and a clear split between operator and player choices |
| 5 | Max win and exposure controls | 3 | Documented maximum win per bet, maximum total payout per round for shared-round games, and where each limit is enforced |
A single RTP figure is rarely enough for originals. In many games the return depends on what the player or operator chooses: a risk level, a board size or how far the player goes before cashing out. Ask for the figure for every option and for the simulation behind it.
Exposure deserves a separate look in crash and other shared-round games. Many players can cash out at the same high multiplier in the same round, so a maximum multiplier per bet does not cap what you can lose in one round. Ask for the per-round cap, who enforces it and what happens to bets when it is reached. How these settings work in practice is a topic in its own right; for scoring, you only need to know whether they exist, are documented and are enforced where the supplier says.
B. Fairness and certification
| # | Criterion | Weight | Evidence that earns a 2 |
|---|---|---|---|
| 6 | RNG implementation | 3 | The RNG type, where it runs (server-side), how raw output is scaled to game outcomes, and an RNG test report |
| 7 | Certificate scope | 3 | Test reports that name the lab, the standard and market, the game version and digital signature, the platform and the channels covered |
| 8 | Provably fair verification | 2 | A written algorithm, a seed commitment shown before the bet, player control of the client seed where the design allows it, and rounds your team has verified with its own code |
| 9 | Change management and re-testing | 2 | Versioned releases with change logs, a stated rule for which changes need external re-testing, and a clear answer on who pays for it |
| 10 | Supplier licensing in your markets | 2 | Licenses or registrations held where your markets require them, or a documented position on why the delivery model needs none, confirmed by your counsel |
Keep these criteria apart. An RNG test report covers the random number generator, not the whole game. A game test report usually covers one build against one market’s standard. Provably fair verification lets players check rounds, but it does not replace independent testing where a license requires it. Supplier licensing is about the company, not the game: in Great Britain, for example, the law covers businesses that manufacture, supply, install or adapt gambling software, not only those that host it, according to the Gambling Commission’s guidance on what counts as gambling software. The two sections after the sheet show how to check criteria 7 and 8.
C. Integration and operations
| # | Criterion | Weight | Evidence that earns a 2 |
|---|---|---|---|
| 11 | Wallet integration integrity | 3 | Documented debit, credit and rollback calls with idempotency keys and round IDs, retry rules and reconciliation reports |
| 12 | Failure and recovery handling | 3 | Round state held on the server, documented reconnect behavior, auto-cash-out executed on the server, and settlement rules for abandoned rounds |
| 13 | Round logs and dispute support | 2 | Any round can be rebuilt from logs: bets, results, seeds or RNG output, wallet transaction IDs and timestamps, with a stated retention period |
| 14 | Back office and reporting | 1 | Configuration controls, per-game reports by currency, and real-time monitoring of exposure |
| 15 | Performance and sandbox | 1 | Load test results for real-time games, mobile client performance figures, and a sandbox you can use before signing |
Originals are fast, and many players use autobet, so even one popular game sends a steady stream of wallet calls. If a retried debit is processed twice, or a lost credit is never resent, you either overcharge players or overpay them, and both show up as disputes. Idempotency (a repeated request with the same transaction ID has no further effect) and documented rollback are what prevent it.
Multi-step and shared-round games add a second risk: the player’s connection can drop while a round is open. The supplier should hold the round state on the server, restore it on reconnect and execute auto-cash-out on the server so that it still happens when the player’s device is offline.
D. Product fit
| # | Criterion | Weight | Evidence that earns a 2 |
|---|---|---|---|
| 16 | Branding depth | 1 | A list of what can be changed (name, logo, colors, art, sounds, interface) and the turnaround time |
| 17 | Currencies, languages and promotions | 1 | Supported fiat and crypto currencies with decimal precision, supported languages, and how free bets or tournaments work with your bonus engine |
| 18 | Responsible gambling and market controls | 2 | Settings for autoplay, speed of play, stake limits, RTP display and reality checks that can differ by market |
| 19 | Catalog depth and release cadence | 1 | The number of production-ready titles, releases over the last 12 months and a roadmap |
| 20 | Track record and references | 1 | Games live with other operators, reference contacts you can call and an incident history |
Market controls matter more for originals than for most content, because many originals are built to resolve almost instantly. Since January 17, 2025, for example, the UK Gambling Commission’s remote technical standards have set a minimum spin speed of 5 seconds for casino games, excluding peer-to-peer poker (RTS 14G), according to the Commission’s summary of the RTS updates. Ask whether pace, autoplay and display can be set per market without a new build.
E. Rights, security and support
| # | Criterion | Weight | Evidence that earns a 2 |
|---|---|---|---|
| 21 | Rights level and exclusivity | 3 | A contract clause that states the rights level (hosted license, non-exclusive code license, exclusive license or IP assignment) and its scope |
| 22 | Code delivery or escrow | 2 | For a code license, a complete package list: client, server, math model, build scripts, documentation and third-party licenses. For a hosted game, escrow terms with clear release triggers |
| 23 | Security | 2 | Evidence of an information security management system, such as ISO/IEC 27001 certification, a recent penetration test summary and a process for patching dependencies |
| 24 | SLA and support | 2 | Uptime commitment for hosted games, incident response times by severity, maintenance notice and a 24/7 support channel |
| 25 | Exit and termination | 2 | What happens to open rounds, player balances, logs and code rights when the contract ends, with notice periods |
“Buy” rarely means “own” in game deals. Most source-code deals grant a non-exclusive license, and most white-label deals keep the code with the supplier. Score criterion 21 on what the contract says, not on what the sales deck calls the deal.
How to check a provably fair claim
A provably fair game commits to a result before the player bets and lets anyone check afterwards that the result was not changed. “Provably fair” on a feature list proves nothing until your own team has checked it. Six checks turn the claim into evidence for criterion 8:
- Get the algorithm in writing. Ask which hash function is used, how the server seed, client seed and nonce are combined, and exactly how the resulting bytes become an outcome such as a dice roll, a crash point or a mine layout.
- Check that the commitment comes first. The hashed server seed must be shown to the player before the bet, and the seed itself revealed only after it is rotated.
- Verify rounds with your own code. Do not rely only on the supplier’s verification page. Take 20 or more rounds from the sandbox, including large wins and losses, and recompute them with code your team wrote or an independent open-source verifier.
- Check player control. Where the design allows it, the player should be able to set or change the client seed, and the nonce should behave as documented after a seed change.
- Ask how shared rounds work. In crash and other multiplayer rounds, everyone shares one result, so it cannot depend on one player’s client seed. Ask how round results are committed in advance, for example with a pre-generated hash chain, and where the commitment is published.
- Check the mapping against the math. Converting random bytes into outcomes can introduce small biases, such as modulo bias. The mapping in the algorithm should produce the RTP stated in the math documentation, and the simulation should confirm it.
Provably fair and lab testing answer different questions. Where a license requires independent testing, provably fair does not remove it: in Great Britain, the Gambling Commission’s testing strategy requires an approved third-party test house to perform statistical analysis of the RNG and its outputs. Standards are starting to cover provably fair systems directly: the draft GLI-19 version 4.0, which GLI issued for public comment on July 7, 2026, adds a framework for provably fair RNGs based on commit-reveal logic. At the time of review, it was still a draft. How provably fair and certified RNG work together in one game is a separate subject.
Five checks on any test certificate
A supplier that says its games are “certified” should be able to show a test report for each game and market. Read it with five questions, which together decide criterion 7:
- Who issued it, and is that lab approved for your market? Regulators keep their own lists. The UK Gambling Commission publishes a list of approved test houses and notes that not every test house is approved in every area, so it is the licensee’s responsibility to appoint one approved for the gambling facilities it provides, according to its test houses guidance.
- What exactly was tested? An RNG report, a game math verification, a full game test and a platform or system certification are different documents. A report on the RNG alone does not certify the game.
- Does the build match? Under the UK Gambling Commission’s testing procedure, a game test report includes the game name, RTP, software number and digital signature, the platform supplier and platform version, and the game clients covered. Compare the version and signature with the build you will actually run.
- Which market and standard does it cover? A report against one market’s technical standard does not certify the game for another market.
- What has changed since? Ask for the change log after the report date. The Commission’s Annex A defines a major update, which requires external retesting by an approved test house, as any software change that may affect the fairness of a game. If a game moves to a different RNG or platform, the testing procedure also expects additional integration testing.
Keep one more distinction in mind: a test report is not the end of the process. In Great Britain, license holders must submit their games and RNG test results to the Commission through its games register, and the Commission’s information requirements say that satisfactory testing must be completed and the report submitted before a game is released for play. Other markets add their own steps, including regulator approval in some, so confirm the process for each of yours. The certification process itself, and which changes trigger re-testing, deserve their own guides.
Eight tests to run in the supplier’s sandbox
Documents describe how a game should behave. A sandbox shows how it does. Ask for access before signing and run at least these tests. They provide the evidence for criteria 5, 8, 11, 12 and 13.
| # | Test | How to run it | What a pass looks like |
|---|---|---|---|
| 1 | Duplicate debit | Send the same debit with the same transaction ID twice | The balance is debited once; the second call returns the original result |
| 2 | Lost credit | Make your wallet time out on a win credit | The game retries with the same transaction ID or rolls back as documented; the player is paid exactly once |
| 3 | Disconnect mid-round | In a multi-step game, close the client after two steps, then reopen it | The round resumes at the last confirmed step; no second round is opened |
| 4 | Offline auto-cash-out | In a crash game, set auto-cash-out at 2.00x and cut the client’s connection | The cash-out executes on the server at 2.00x if the round reaches it |
| 5 | Late cash-out | Send a cash-out request that arrives after the crash point | The request is rejected based on server time |
| 6 | Round payout cap | Have several test accounts cash out at a high multiplier in one round | The documented per-round cap is enforced, with the documented effect on each bet |
| 7 | Provably fair replay | Recompute 20 sandbox rounds with your own verifier | Every result matches |
| 8 | Round reconstruction | Pick any round ID and request its full record | Bets, result, seeds or RNG output, wallet transaction IDs and timestamps are all present |
Run the wallet tests against your own platform’s wallet, not only the supplier’s test harness: that is where most integration problems appear. The integration project itself, from API mapping to go-live, is a separate piece of work.
Red flags
Any one of these is a reason to slow down. Two or more are usually a reason to move on.
- Only a headline RTP, with no figures per mode, profile or exit point.
- “Certified” with no test report that names the lab, the market and the build.
- Provably fair verification only on the supplier’s own page, with no written algorithm.
- A maximum win per bet but no cap per round in a shared-round game.
- Limits “handled by the platform”, with no document saying which system enforces what.
- No sandbox before the contract.
- No round logs on request, or logs that cannot rebuild a round.
- No references from operators running the games today.
- A contract that says nothing about open rounds, balances and logs at termination.
Running the evaluation in five steps
- Set your context and knockouts. Decide where the games will run, which delivery model you are buying and who enforces limits. Mark the knockout criteria that apply.
- Send the checklist as a request for information. Ask each shortlisted supplier to answer all 25 criteria and attach the evidence: math documentation, simulation reports, test reports, API documentation and a sample contract.
- Score the documents. Have product, engineering and compliance score their own groups independently, then compare. Large disagreements usually mean the evidence is unclear.
- Test and meet. Run the sandbox tests, hold a technical session with the supplier’s game and backend engineers, and update the scores.
- Call references and contract. Speak to operators running the games, then write every criterion scored 1 into the contract as an obligation with a date.
How OmiSoft answers the checklist
We supply ready-made originals, so we apply the same checklist to our own games. Here is where we stand on the criteria that most often decide a shortlist, using only what we publish:
- Catalog. Our white-label casino games and originals catalog holds 22 ready-made games in four groups: instant and crash games, card and table games, slots, and multiplayer games.
- Math and configuration. How far the math can be configured depends on the game. In Cross the Lake, operators can set a target RTP between 96% and 98.5%, the maximum multiplier, betting limits and cash-out rules, and each risk profile is verified through simulation before production.
- Fairness. Depending on the title, games use provably fair verification or server-side RNG.
- Round state and wallet calls. In Cross the Lake, for example, the trap layout for each round is generated and validated on the server, debit, cash-out credit and rollback requests use idempotent transaction references, and after a reconnect the game restores the last confirmed step from the server instead of opening a second round.
- Testing. Game math, logs, configuration controls and documentation can be prepared for independent testing. Certification comes only from an accredited laboratory after its evaluation, and supplier obligations in each market should be confirmed with legal counsel.
- Rights. You get a non-exclusive license to each game together with its source code. The code runs on your infrastructure under your brand, so the escrow question for hosted games does not arise, and the same game can also be licensed to other operators. If you need a title no other casino can offer, that is a custom build through our casino game development services.
- Branding and integration. We customize each game’s look for your casino and connect it to your player accounts, wallet and backend.










