Licensed online slots are not rigged in the way most players mean: a game that watches your session and changes its odds to make you lose is banned by the rules regulators enforce, and independent labs test the random number generator and the advertised return before a game goes live. What is real is less dramatic but matters more: one game can be offered in several RTP versions, every slot has a disclosed house edge, and an unlicensed site is outside all of these checks.
This page sets out what “rigged” would mean under the rules, who tests what, where the genuine risks are, and why a fair game can still feel fixed. Every rule quoted comes from the regulator’s or lab’s own page.
- Confidence level GLI-19 sets for the statistical tests on a game’s RNG output
- 99%
- GLI-19 3.2.2
- Certified RTP versions one studio lists for a single game, 88.42% to 96.38%
- 4
- Hacksaw Gaming
- Fair 1,000-spin sessions with at least one run of 20 empty spins, illustrative high-volatility 96% game
- 93%
- S6
- Time a UK-licensed business has to resolve your complaint before free ADR
- 8 weeks
- UK Gambling Commission
Source: Sources at the end of the page; S6 is our own study with made-up pay tables
Sources: GLI-19 v3.0 sections 3.2 and A.6.1; UK Gambling Commission RTS 7, FOI responses on slot testing and RTP, and "How to complain" (all fetched 9 October 2026).
What “rigged” would mean under the rules
A rigged slot would be one whose results do not follow its published rules and probabilities. The UK Gambling Commission’s Remote Gambling and Software Technical Standards (RTS) spell out what that rules out:
- RTS 7A requires that “random number generation and game results must be ‘acceptably random’” and states that “adaptive behaviour (that is, a compensated game) is not permitted”. The guidance adds that this “prohibits automatic or manual interventions that change the probabilities”.
- RTS 7B requires that “games and events must be implemented fairly and in accordance with the rules and prevailing payouts”. Random numbers are “to be used in the order in which they are received”.
- RTS 7C bans designs that may reasonably be expected to mislead a player about the likelihood of winning, and says “games may not falsely display near-miss results”.
GLI-19, the interactive-gaming standard from Gaming Laboratories International that many regulators use as a base, says the same in its own words. Section 3.2.4 says a game “shall not adapt its theoretical return to the player based on past payouts” and “shall not adjust the likelihood of a bonus/feature occurring, based on the history of awards obtained in previous games”. It also bans a game from swapping the losing outcome the RNG picked for a different losing outcome that looks closer to a win.
| Rule | What it requires | Source |
|---|---|---|
| RTS 7A | Results acceptably random; compensated (adaptive) games not permitted | UK Gambling Commission |
| RTS 7B | Game follows its rules and payouts; random numbers used in the order received | UK Gambling Commission |
| RTS 7C | No misleading designs; no false near-miss displays | UK Gambling Commission |
| RTS 3C | RTP or probability of winning easily available before you commit to play | UK Gambling Commission |
| GLI-19 3.2.1 | Lab reviews RNG source code, including for undisclosed switches or parameters | GLI-19 v3.0 |
| GLI-19 3.2.2 | Statistical tests on final results, evaluated at a 99% confidence level | GLI-19 v3.0 |
| GLI-19 4.7.1 | Theoretical payout of at least 75%, unless the regulator sets otherwise | GLI-19 v3.0 |
| GLI-19 A.6.1 | Operator records every change to theoretical RTP and compares theoretical with actual RTP | GLI-19 v3.0 |
Source: UKGC RTS 3 and RTS 7 pages; GLI-19 v3.0 PDF (all fetched 9 October 2026)
Physical machines are a different case
“Are slot machines rigged” often means the machines in a venue rather than online games, and there the answer has a twist. The UK Gambling Commission’s consumer guide to gaming machines says “compensated machines are also random, but they vary the chance of winning a prize” depending on results from previous games, while on “random machines” the odds “are not affected by wins or losses in previous games”. Both kinds are legal in that guide, and machines must show “their category, %RTP and whether they are compensated or random”. For remote games, RTS 7A does not permit compensated games.
Slot RNG explained: what the labs actually test
A random number generator (RNG) produces numbers that cannot be predicted from the ones before. The game maps each number to a result according to its pay table. That mapping is where volatility and RTP come from, and it is fixed when the game is certified.
GLI-19 sets out what an independent lab must do:
- Read the code. Section 3.2.1 requires the lab to review the source code of the randomness, scaling and shuffling algorithms, looking for “sources of bias, errors in implementation, malicious code” and “undisclosed switches or parameters having a possible influence on randomness and fair play”.
- Test the output. Section 3.2.2 requires statistical tests on the final results after mapping, chosen to confirm the intended distribution and independence between draws, “evaluated, collectively, at a 99% confidence level”. It lists tests such as chi-square, runs, serial correlation and duplicates.
- Check every outcome is possible. Section 3.2.3 says each possible RNG selection “shall be equally likely to be chosen”, or follow the game’s intended distribution where the design says otherwise.
The labs describe the same work on their own sites:
| Lab | What its page says |
|---|---|
| GLI | Publishes the GLI standards, including GLI-19 for interactive gaming systems; notes each jurisdiction can set its own standards |
| iTech Labs | Certifies RNGs (source code evaluation, testing of raw and scaled output), games and RTP; offers an RTP/RNG audit certification from server logs |
| eCOGRA | RNG certification starts with a source code review and "extensive statistical analysis" of output; RTP testing uses simulations on large samples |
| BMM Testlabs | Maths testing works from PAR sheets, simulations and RTP of the base game, features and overall game, validated against the production configuration |
Summaries of the labs' own service pages; we did not audit any lab's work.
Source: gaminglabs.com, itechlabs.com, ecogra.org, bmm.com (fetched 9 October 2026)
In Great Britain, the regulator relies on these test houses rather than approving games itself. In a Freedom of Information answer about online slots, the Commission wrote that “the Gambling Commission do not approve games or test results; we require operators to get the game tested”, and in another answer that testing against its standards includes “verifying that the advertised RTP is correct”. The answer did not say how often games are retested.
Can casinos change slot RTP?
This is the part of the “rigged” worry that has a real basis. A studio can build several versions of the same game with different pay tables, and each version has its own RTP. Hacksaw Gaming’s page for Wanted Dead or a Wild lists four RTP values for that one title: 96.38%, 94.55%, 92.33% and 88.42%. The page does not say which sites use which version.
| RTP version | House edge | Expected loss per $1,000 wagered |
|---|---|---|
| 96.38% | 3.62% | |
| 94.55% | 5.45% | |
| 92.33% | 7.67% | |
| 88.42% | 11.58% |
RTP values from the studio's game page; loss = (100% − RTP) × $1,000, our arithmetic. Expected values, not session forecasts.
Source: Hacksaw Gaming game page (fetched 9 October 2026); arithmetic ours
Offering a lower-RTP version is not rigging under the rules above, provided the version is certified and its RTP is disclosed. What the rules control is how changes happen:
- The game cannot change itself. RTS 7A and GLI-19 3.2.4 ban a game from adapting its return to past payouts.
- Changes are recorded. GLI-19 appendix A.6.1 requires records of “dates and type of changes made affecting the game’s theoretical RTP percentage”, and says each change makes the game “new for all reports and records”.
- The figure must be available. RTS 3C requires RTP or probability information to be easily available before you commit to play.
- There is a floor. GLI-19 4.7.1 sets a 75% theoretical minimum unless the regulator sets its own.
So the practical defence is to read the RTP in the help screen of the game you are about to play. How to find and read slot RTP walks through where it sits and what the number means.
The casino tightens a slot when I start winning
A licensed game cannot react to your session. Changing probabilities based on past results is the “adaptive behaviour” RTS 7A bans and GLI-19 3.2.4 rules out. A different RTP version is a different certified game, chosen in advance, not a switch flipped mid-session.
The real risk: unlicensed sites
Every check above depends on a regulator that licenses the operator and requires tested games. A site without a licence from a regulator you can verify is not bound by any of them, and you would have no regulator or approved dispute body to turn to. That, not a hidden setting in a tested game, is where the genuine risk lies.
How to check a licence:
- Find the licence statement. It is usually in the site’s footer, with the regulator’s name and a licence number.
- Check it on the regulator’s own website, never through a link alone. The UK Gambling Commission says you can check whether a business is licensed by “searching for the company’s name, or the name of the website or app” on its public register, which also shows “if we have taken any regulatory action against the company”. It adds: “If the gambling business doesn’t have a licence it is acting illegally.”
- For sites claiming a Malta licence, use the Malta Gaming Authority’s licensee register, which includes a URL checker. The MGA’s player hub also publishes a list of unauthorised URLs and says it “carries out checks on suspicious websites to ensure that players are protected from misleading claims”.
- Match the details. The company name, the website address and the licence status on the register should match what the site shows.
Why fair games feel rigged
Even a perfectly fair slot produces runs that look deliberate. To show how far, we used the three made-up pay tables from our study S6. All return exactly 96.0%, and they differ only in volatility. These are not real games, but they show what honest randomness does over a session.
Source: Our study S6, sessions[1,000 spins].longestDryStreak.probabilityAtLeast
See the numbers as a table
| Sessions with at least one run of 20+ spins returning nothing | |
|---|---|
| Low volatility | 0.12% |
| Medium volatility | 27.85% |
| High volatility | 92.66% |
On the high-volatility model, a streak of 20 empty spins in a row turns up in 92.7% of 1,000-spin sessions. Nothing about that is unfair. It is the arithmetic of a game that returns nothing on 80.4% of its spins.
The final result feels just as personal. With a 100-bet bankroll, most fair sessions on the volatile models end at zero:
Source: Our study S6, sessions[1,000 spins].probabilityBustExact and probabilityFinishAheadExact
See the numbers as a table
| Went bust | Down, not bust | Finished ahead | Total | |
|---|---|---|---|---|
| Low volatility | 35.1% | 36.8% | 28.1% | 100.0% |
| Medium volatility | 79.4% | 3.9% | 16.7% | 100.0% |
| High volatility | 93.5% | 0.9% | 5.6% | 100.0% |
And the return you observe tells you little about the RTP. After 1,000 spins, the 95% range of observed returns around 96% is about ±15.5 points on the low-volatility model, ±66.4 on the medium and ±219.4 on the high one. A player who sees 60% back on a 96% game has not caught it cheating.
Source: Our study S6, rtpConvergence.ci95HalfWidthPctPoints
See the numbers as a table
| 95% range, plus or minus percentage points | |
|---|---|
| Low volatility | 15.5 pts |
| Medium volatility | 66.4 pts |
| High volatility | 219.4 pts |
Operators are required to watch for real faults
GLI-19 appendix A.6.2 asks operators to monitor game and RNG output so that abnormalities, for example “the actual RTP percentage for the period falling outside the expected range”, are logged and investigated. The ranges above are why that monitoring uses large volumes of play, not one player’s session.
For the full set of session numbers, including how long each game takes before its observed RTP settles, see the three-pay-table volatility study. The same logic applies to other machines; the video poker page covers a game where the pay table is printed in full.
What to do if you suspect unfair play
- Write it down. Note the game name, the date and time, your stake and the amounts involved. Screenshot the game’s information screen, including its RTP.
- Complain to the operator first. For businesses licensed in Great Britain, the Commission’s guide says to use the business’s own complaints process and give “all the information about your complaint, including dates, times and amounts of money”. It also says “the gambling business has 8 weeks to resolve your complaint from when they receive it”.
- Escalate to an ADR provider. If you are unhappy after that process, or after 8 weeks, you can take the complaint to an Alternative Dispute Resolution provider, which the Commission describes as “a free, independent service”. The business should tell you which provider it uses. eCOGRA, for example, says it is an approved ADR entity of the Gambling Commission and that it cannot investigate disputes older than one year after the operator’s own process ends.
- Report an unlicensed site. The Commission says you can report illegal gambling activity to it confidentially.
- For MGA-licensed sites, the Malta Gaming Authority has an online form for complaints against its licensees and strongly recommends trying to resolve the dispute with the operator’s support first.
ADR bodies will not take every complaint: the Commission’s guide says they will not accept complaints about a customer service issue or a refusal to accept your bet or custom.
What we could not verify, and our assumptions
- Which RTP version any site runs. We confirmed four published RTP values for one game on one studio’s page. We did not check any operator, and we name none.
- How common multiple RTP versions are. We found only one primary source (a studio’s own page). Secondary articles say the practice is widespread; we did not rely on them. Another studio page we fetched listed no RTP at all.
- How often games are retested. The Commission’s FOI answer did not say, and we found no other primary source.
- Rules outside Great Britain and Malta. We quote UK standards and GLI-19. Other regulators set their own rules; GLI says each jurisdiction can set its own standards.
- Register search fields. The UK search method is quoted from the Commission’s guide. The MGA register page we fetched names a URL checker but does not describe its search fields.
- ADR changes. The Commission’s ADR page notes that some ADR information was removed after updates linked to the Digital Markets, Competition and Consumers Act 2024. Check the current page before you escalate.
- Our pay tables are invented. The S6 models illustrate fair randomness; they are not the maths of any real slot.
- We tested no real game or site. Nothing on this page is a finding about a particular casino or game.