What does the lesson “How an Online Slot Is Made: From Maths Model to Live Game” cover?

By Vadym Zelinskyi·Practitioner·11 min read·verified 2026-11-04

How an Online Slot Is Made: From Maths Model to Live Game

This lesson follows an online slot from its maths model to a player's screen, using what public rules require and what research papers show: the PAR sheet, the random number generator, the rules shown to players, design limits, independent testing, licensing, distribution and monitoring. No primary public source shows how studios generally work inside, so it describes the requirements they work under, mainly in Great Britain, with Malta and Germany for comparison. It is written for adults (18+) and does not promote gambling or suggest it is a way to make money.

Key takeaways

A slot's return to player is set by its reel strips, paytable and feature rules, and a PAR sheet records it. Identical-looking versions of one Ontario land-based game paid back from 85% to 96.2%.
In Great Britain a new game can only be released once an approved test house has finished testing and its report has reached the Gambling Commission.
Design limits depend on the country. Great Britain sets 2.5 seconds per slot game, bans autoplay and caps stakes at £5, or £2 for under-25s. Germany requires at least five seconds per game on average.
Checks continue after release: updates that may affect fairness go back to a test house, and licensees compare live RTP with the advertised figure.
No primary public source shows how studios generally work inside. This lesson rests on published rules, a regulator's case record, a competition decision, a peer-reviewed paper and calculations you can repeat.

By the end you can

Explain what a PAR sheet records and how reel strips, the paytable and feature rules together set a slot's theoretical RTP.
Describe the difference between calculating an RTP exactly and estimating it by simulation, and why volatility affects how many simulated games are needed.
List what Great Britain requires of a slot before release: acceptable randomness, disclosed rules, design limits, approved test house checks and a report to the regulator.
Trace how a tested game reaches an operator and what follows release: licences, hosting and aggregation, change control and live RTP monitoring.

Step 1: the maths model and the PAR sheet

No primary public source shows how studios choose themes or run their projects, so this lesson follows what public rules require and published papers show. A slot's maths has three main parts: reel strips (the ordered symbols on each reel), a paytable (what each combination pays) and feature rules. Great Britain's Gambling Commission says that together they 'determine the overall game RTP', the return to player. GLI-19, a technical standard published by the test lab Gaming Laboratories International, describes the specification as a PAR sheet: it gives the theoretical RTP (the designed figure), hit frequency and reel strips. Researchers Harrigan and Dixon obtained real PAR sheets for land-based slot machines in Ontario under freedom of information law and published their analysis in 2009. One game's five reels held 47, 46, 48, 50 and 50 symbols: 259,440,000 combinations. Its approved versions looked identical yet paid back from 85% to 96.2%. The designers changed the symbols on the reels and kept the number per reel.

Quick check

All versions of the Ontario game in Harrigan and Dixon's paper look identical to the player. What made one pay back 85% and another 96.2%?

Step 2: checking the sums and adding randomness

GLI-19 says a theoretical RTP may rest on 'mathematical calculations or simulations'. Calculation works when every outcome can be counted. A toy game invented for this lesson has three reels of ten stops, so 1,000 outcomes (the fact-checks list its reels and paytable). Counting them gives exactly 95.3%. Swap one cherry for a blank on the third reel, keep the paytable, and it falls to 81.2%. Simulation plays the game many times instead. Five simulated runs of 1,000 spins returned between 94.5% and 126.4%; five runs of 10 million returned between 95.24% and 95.72%. Live outcomes come from a random number generator (RNG). In Great Britain, the Commission's technical standard RTS 7 requires results that are 'acceptably random', bans a 'compensated game' whose probabilities are changed during play, and bans swapping one losing result for a near-miss. GLI-19 has a lab test the RNG's output after scaling, the step that maps raw numbers onto the range a game uses, at a 99% confidence level.

Quick check

The toy game's exact RTP is 95.3%, yet five simulated runs of 1,000 spins returned between 94.5% and 126.4%. What does that show?

Step 3: what the game must show and what design may not do

Rules differ by country. In Great Britain, RTS 3 requires the game's rules, and its RTP, house edge or probability of winning, to be easily available before you commit to gamble. RTS 14D sets a minimum of 2.5 seconds from starting a slot game until the next game cycle can begin. RTS 14E rules out turbo, quick spin and slam stop. RTS 14F bans celebrating a return less than or equal to the stake, which in the toy game covers the 18.9% of outcomes returning exactly the stake. RTS 8A requires you to commit to each game cycle individually, which rules out autoplay. A 2025 regulation, SI 2025/215, caps an online slot stake at £5 per game cycle, or £2 if you are under 25. Germany's gambling treaty requires at least five seconds per game on average. GLI-19 leaves autoplay to each jurisdiction. In June 2026 the supplier Stakelogic BV settled with the Commission for £122,835 after its slots ran under 2.5 seconds, one at 1.97.

Step 4: independent testing before release

The Commission's testing strategy (last updated 31 October 2025) sets out what Great Britain requires. New games must be tested by a test house from its approved list. For the RNG, the lab reviews documents, known weaknesses and source code, and runs statistical tests. For the game, it checks the design: 'maths, artwork or rules as displayed to players, and theoretical RTP'. It then checks the software in an environment reflecting live use, by simulation (automated play, for a number of games that depends on volatility), emulation (reproducing rare outcomes such as jackpot triggers) and manual play. The report identifies the game by name, RTP, software number and digital signature. 'All new games and RNGs can only be released once the testing has been completed and the report provided to the Commission.' Limits such as RTS 14 fall to the licensee's own pre-release testing where appropriate. Elsewhere, a regulator can adopt GLI-19 'in whole or in part'; Malta requires a game to be 'approved or otherwise recognised' by its regulator.

Quick check

Under Great Britain's testing strategy, when may a new slot be released?

Step 5: licences, and how the game reaches a casino

In England, Wales and Scotland, section 41 of the Gambling Act 2005 makes it an offence to manufacture, supply, install or adapt gambling software in the course of a business without an operating licence. Licence condition 2.2.1 requires remote casino, bingo and betting licensees to use software made and supplied by holders of a gambling software licence. That licence does not cover running the gambling: a holder that 'hosts a poker network or a games platform' also needs a remote casino licence. Malta has a business-to-business critical gaming supply licence, covering 'supply and management of material elements of a game', for supply from Malta, to anyone in Malta or through a Maltese entity. GLI-19 calls the system that hosts games or aggregates content a gaming platform, also known as a remote gaming server (RGS) or game aggregator. Reviewing live-casino games in 2020, the UK's Competition and Markets Authority said aggregators 'act as intermediaries' between suppliers and operators, and suppliers 'earn a commission only when consumers play their games'.

Step 6: changes, monitoring and audit after release

In Great Britain, a major update is any software change that may affect a game's fairness, including to the RNG or game rules. It goes back to an approved test house, with the report filed before release. An RGS or RNG update means retesting a representative sample of games. A minor update, such as a fix to a sound file, can rely on internal testing, though all updates must be recorded. RTS 7D bars changing rules, payouts or probabilities while a game is available, unless its rules provide for it. GLI-19 version 4.0 (3 November 2026) asks operators to record changes affecting theoretical RTP and to set 'a new reporting baseline' when a change is material. On live games, licensees must compare actual RTP (win divided by turnover) with the advertised figure. One acceptable method measures daily over the last 30 days of play. An approved test house audits these controls yearly for holders of both a software licence and a casino or game-host licence. RTS breaches are reportable to the Commission.

Quick check

A supplier corrects a bonus-round win calculation that was wrong in a rarely seen scenario. How does Great Britain's testing strategy treat that update?

Key terms

PAR Sheet Reel Strip Paytable RTP (Return to Player)RNG (Random Number Generator)Game Certification Remote Gaming Server (RGS)Aggregator

Check yourself

Check yourself · 1/4

According to GLI-19's definition, what does a PAR sheet provide?

Check yourself · 2/4

A slot in Great Britain pays back exactly what you staked on a spin. What does RTS 14F say about that?

Check yourself · 3/4

Which activities does section 41 of the Gambling Act 2005 require an operating licence for, when done in the course of a business?

Check yourself · 4/4

What is the main form of live RTP monitoring that the Gambling Commission expects?

Next → Now the professional layer: take “KYC & AML in Practice: Verifying and Monitoring Players”.

Go deeper

Sources (14)

Verified 2026-11-04 · next review 2027-11-04 · high confidence

Fact-checks (15)

Harrigan and Dixon, 'PAR Sheets, probabilities, and slot machine play', Journal of Gambling Issues 23, pages 81 to 110 (June 2009, peer-reviewed), analysed PAR sheets obtained under Ontario's Freedom of Information and Protection of Privacy Act for land-based machines. Lucky Larry's Lobstermania has reels of 47, 46, 48, 50 and 50 symbols, 'yielding a total of 259,440,000 possible combinations'; its versions range 'from a low of 85% to a high of 96.2%'; and the percentages 'are achieved by changing the symbols on the reels while maintaining the number of symbols per reel'. The authors also analysed all 259,440,000 outcomes by computer.
GLI-19 version 4.0 (Gaming Laboratories International, revision date 3 November 2026) replaces version 3.0 of 17 July 2020. Its glossary defines a PAR sheet as giving 'the theoretical return to player, hit frequency, symbol combination, number of reels, number of credits that can be accepted, and reel strip listing as applicable', and says a 'theoretical' RTP is 'based on mathematical calculations or simulations' and an 'actual' one on the game's metering. Section 5.2.2 sets statistical RNG testing after scaling at a 99% confidence level, and section 1.2.3 says the standard 'can be adopted in whole or in part by any regulatory body'.
GLI-19 version 4.0 section 6.11.2 has autoplay allowed or disallowed 'reflective of jurisdictional preference'. Appendix A.5.2(b) requires records of changes affecting theoretical RTP and says that where changes materially affect it 'a new reporting baseline shall be established'. Its glossary says the 'Gaming Platform', also known as 'Remote Gaming Server (RGS)' or 'Game Aggregator', is 'used to host interactive gaming and/or aggregate game content'.
Toy game, invented for this lesson and not a real product: each of three reels has ten stops (one seven, two bars, three cherries, four blanks). Three sevens pay 200 times the stake, three bars 30, three cherries 12, and exactly two cherries pay back the stake. Over the 1,000 outcomes the payouts total 200 + 240 + 324 + 189 = 953, an RTP of 95.3%, and 189 outcomes (18.9%) pay back exactly the stake. With one third-reel cherry swapped for a blank the total is 200 + 240 + 216 + 156 = 812, an RTP of 81.2%. The simulations used five seeded runs at each length.
UK Gambling Commission, testing strategy (published 9 February 2021, last updated 31 October 2025): game rules covers 'pay tables, symbol distribution, feature rules etc', and 'Collectively the game rules determine the overall game RTP'; game testing covers 'verification of game design – maths, artwork or rules as displayed to players, and theoretical RTP' and actual RTP 'using simulation, emulation and manual testing'; 'All new games and RNGs can only be released once the testing has been completed and the report provided to the Commission'.
The same testing strategy defines a major update as 'any software change which may affect the fairness of a game', lists a bonus-round win calculation fix as a major example and a sound format change as a minor one, and says 'Licensees must ensure sufficient RTP monitoring is in place for both under and overpayments', giving 'daily measurements based on the last 30 days of play' as one acceptable method. An update to an RGS or RNG requires a 'representative sample of games' to be retested, and a footnote says non-compliance with the standards is 'reportable as an LCCP event notification'. The annual games testing audit applies to licensees holding a gambling software licence alongside specified licences, including remote casino and remote casino (game host).
UK Gambling Commission, RTS 7 (last updated 7 May 2024): 7A requires 'acceptably random' results and says 'Adaptive behaviour (that is, a compensated game) is not permitted'; 7C guidance says games 'may not falsely display near-miss results'; 7D says rules, payouts and outcome probabilities 'may not be changed while it is available for gambling, except as provided for in the rules'. RTS 3 (last updated 7 May 2024): 3A and 3C require rules and the house edge, RTP or probability of winning to be easily available before the customer commits to gamble.
UK Gambling Commission, RTS 14 (last updated 12 January 2026): 14D requires 'a minimum of 2.5 seconds from the time a game is started until the next game cycle can be commenced' and applies to slots; 14E guidance says 'Features such as turbo, quick spin and slam stop are not permitted'; 14F says the system 'must not celebrate a return which is less than or equal to the total stake gambled'. RTS 8 (last updated 21 January 2025): its aim is 'To make clear that auto-play cannot be offered for online gaming', and 8A requires a customer 'to commit to each game cycle individually'.
The Gambling Act 2005 (Operating Licence Conditions) (Amendment) Regulations 2025, SI 2025/215 (made 25 February 2025, extending to England and Wales and Scotland): 'for an online slots game, the total amount which an individual may stake in relation to any game cycle may not exceed' £2 where the individual is less than 25 years old and £5 where 25 or over.
Germany, State Treaty on Gambling 2021 (GlüStV 2021), section 22a on virtual slot games: paragraph 6 says a game must last at least five seconds on average ('durchschnittlich mindestens fünf Sekunden').
UK Gambling Commission, public statement on Stakelogic BV (25 June 2026): after the licensee reported a fault, the Commission found a breach of licence condition 2.3.1 and RTS 14D. Tiger Temple 88 'operated with 1.97 seconds' between games, and a further 15 games ran between 0.001 and 0.675 seconds under the minimum during periods between 31 October 2021 and 30 October 2025. The settlement includes a payment in lieu of a financial penalty of £122,835.
Gambling Act 2005, section 41(1) and (2), extending to England, Wales and Scotland. UK Gambling Commission guide 'What is gambling software?' (last updated 13 June 2024): it quotes licence condition 2.2.1, and says a software licence 'does not permit the provision of facilities for gambling' and that a holder that 'hosts a poker network or a games platform will also require a remote casino operating licence'. It adds that software only ever supplied to operators licensed solely in other jurisdictions need not meet the British standards.
Malta, Gaming Authorisations Regulations (S.L. 583.05): regulation 3(1) requires a licence to 'provide a critical gaming supply from Malta or to any person in Malta, or through a Maltese legal entity'; regulation 3(2) says a licensable game must be 'approved or otherwise recognised by the Authority'; the First Schedule defines the 'Critical Gaming Supply licence: a business to business licence' and lists 'supply and management of material elements of a game'.
UK Competition and Markets Authority, Evolution/NetEnt decision ME/6894/20 (announced 16 November 2020, full text 8 December 2020), which assessed live-casino games: footnote 50 says 'Aggregators act as intermediaries between direct online live gaming suppliers and gambling operators', and paragraph 17 says suppliers 'earn a commission only when consumers play their games'. Commission rates are replaced by '[]' in the published text for reasons of commercial confidentiality.
What is left out: no primary public source was found that describes how studios generally choose themes, staff projects or schedule development, what a slot typically costs to build, or the commission rates suppliers charge. This lesson does not estimate them.

Back to Learn →

Get the iGamer Briefing

A short, cited email when something in iGaming changes — plus one fundamental worth understanding. Independent, 18+, no hype.