Home/Casino Economics/Randomness and Provably Fair
Casino Economics - Long read14 min

How casino randomness actually works

There are two kinds of randomness in online casinos. Here is how each works, what you can verify, and what you cannot.

Published
May 2026
Reviewed by
Wager Warriors editorial team
Updates
Tracked monthly. Last verified: 12 May 2026.
Verdict
Provably fair is more verifiable. It is not more random.
Section 01The two kinds of randomness

Two ways to generate a random number

One locks its numbers in ahead of time and lets you check them. The other runs non-stop and asks you to trust an independent lab. Both are genuinely random. Only one lets you verify it.

A casino slot or dice game needs a number every time you press a button. The difference between a provably fair game and a certified-PRNG game is not whether that number is random. It is who you have to trust that it is.

Provably fair

You can check it yourself

What goes inA server seed, your client seed, and a nonce (a counter that ticks up with every bet)
The recipeA fixed cryptographic formula (HMAC-SHA256) that anyone can run
Locked in before playThe casino publishes a scrambled version of the server seed before you bet, so it cannot change it later
Shown to you afterThe casino reveals the full server seed once it switches to a new one
Can you re-check it?Yes - you can re-run any past bet yourself and confirm the result was honest
Who you trustThe math, not the operator
Certified PRNG

An independent lab checks it for you

What goes inA random source inside the game server, topped up continuously
The recipeA certified software random-number generator, firing thousands of times a second
Locked in before playNothing - results are pulled the instant you spin
Shown to you afterNothing - the underlying math is kept private by the game studio
Can you re-check it?No - you cannot re-run a past bet to confirm it yourself
Who you trustAn independent certifier (GLI, eCOGRA, iTech Labs)

The provably fair side is fully transparent. The casino locks in a server seed up front, you and the operator add a client seed and a nonce (a counter that ticks up with every bet), and a fixed formula turns those into a result anyone can re-run later. The operator cannot tamper with a single outcome without breaking every earlier bet that was already locked in.

The certified-PRNG side is closed to the player but checked by outsiders. A licensed provider runs a continuous random-number generator, and a regulator-approved certifier tests it for bias. The result you see is genuinely unpredictable. It is also not something you can re-run and check on your own.

Statistically, neither is more random than the other. The difference is who you have to trust.
Section 02How much results can swing

One slot. Ten thousand spins. A hundred players, each with a different night.

Modern slots do not roll dice live. They pick from a huge pre-built list of possible outcomes - that list is called the book. This demo uses the public Waylanders Forge structure, paytable and 97.70% RTP to build a Stake Engine-style book. It is not the real Waylanders Forge math package.

Game
6x4 all-ways slot
Bet size
$1.00
Target RTP
97.70%
Spins
10,000
The run we are followingrun 27 of 100 - spins 1-10,000
Estimated high-volatility book - 100,000 possible outcomes
Assumed hit rate
27.1%
Bonus frequency
1 in 220
Max-win odds
1 in 100,000,000
Model RTP
97.70%
21%228%435%02,5005,0007,50010,000spinsbudget
99 other runs in grey - the run we are following in amber - up the side: money left as % of the 10,000-spin stake budgetGenerated from public structure, not live production math
Actual RTP this run
91.4%
target: 97.70%
Money left at the end
91.4%
of the 10,000-spin stake budget
Longest losing streak
21
spins in a row returning nothing
Longest winning streak
5
spins in a row that paid something
How this run ranked
Beat 62
of the 99 other runs - middle of the pack
Where all 100 runs ended up

How much money each run had left after 10,000 spins

100 runs
50%75%100%125%150%final bankroll as % of start
How often the money runs out

What if your bankroll is not big enough to last 10,000 spins?

The chart above assumes every player has unlimited money to keep spinning. Real bankrolls run out. With $1 spins and a fixed starting balance, this is how often the model hit zero before 10,000 spins were up - and roughly when.

Starting bankrollRan out of moneyTypical spin they ran out on
$100100 of 100 playersspin 132
$25098 of 100 playersspin 364
$50096 of 100 playersspin 741
$1,00092 of 100 playersspin 1,629
$2,50077 of 100 playersspin 4,227
$5,00043 of 100 playersspin 8,170
Honesty note

This is not the production Waylanders Forge math package. Hit rate, bonus frequency, tail frequency and max-win odds are explicit assumptions chosen to demonstrate high-volatility slot behaviour around the public 97.70% RTP and 80,085x max-win claim. The model copies the way Stake Engine lays its games out: a pre-built book of outcomes, payout multipliers, and how often each outcome comes up. Use it to understand how wildly results can swing, not as the exact Waylanders Forge odds.

Section 03What actually decides your night

RTP and volatility do the work. Provably fair or not does not.

If two games have the same RTP and the same volatility, your session will play out the same way - whether the game is provably fair or runs on a certified PRNG makes no difference to the spread of outcomes.

Return to Player is the average return over very long play. It is what the numbers settle toward over hundreds of thousands of spins, not a promise for tonight. Even at 99% RTP across 10,000 spins, individual sessions still land anywhere from roughly 85% to 115% of your starting bankroll. At 96%, that spread is wider.

Volatility is how wildly results swing around that average. Low-volatility games keep most players close to the average. High-volatility games hand most players a worse result than the headline RTP suggests, while paying out big to a small handful.

Low volatility

96% RTP, gentle swings

50%75%100%125%where players end up, as % of starting bankroll

Most sessions land within about 10% of where they started. Few big swings either way.

High volatility

96% RTP, wild swings

50%75%100%125%where players end up, as % of starting bankroll

Same average over the long run. Most players lose more than the 96% suggests. A small handful win big and pull the average up.

The house edge is the same in both charts above. The lived experience is not. A high-volatility game with a generous-looking RTP can still bankrupt the majority of its players over a normal session because the average is held up by a small number of large wins that most players never see.

Bonus buys take this further. You pay up front for the rare event, skipping the long stretches of small wins and small losses that fill a normal session. The RTP does not change. You just get to the answer faster - and the answer is usually that you lost.

The house does not need to cheat. The math does the work over enough players.
Section 04Why checkers differ by casino

Same idea, different math, casino by casino

Every provably-fair casino scrambles the numbers slightly differently. The same seeds produce different results across sites - and that is normal, not a flaw.

There is no central body setting a single standard for how the seeds get combined. Each casino publishes its own method, which means the casino's own checking tool is the only one that can verify a bet placed there. A few practical knobs matter for players: whether you can change your own seed, and how often the casino rotates its server seed.

CasinoHow they combine the seedsCan you set your seed?How often the server seed resetsTheir checker
Stakeserver : client : nonceYesWhen you askStake
BC.Gameserver : client : nonce : cursorYesWhen you askBC.Game
Gamdomserver : nonce : clientYesEvery 24 hoursGamdom
Roobetserver : client : nonceYesWhen you askRoobet
BetFuryserver : nonceNoAutomatic, not disclosedBetFury (limited)

Listed for reference only. Inclusion is not endorsement. See each casino's PIR for the full risk picture.

Section 05What checking a bet shows

What checking a Stake Engine slot bet actually shows you

Checking a slot bet is not as simple as a dice roll. The random draw picks one outcome from the pre-built book, and that outcome already spells out the board you saw, any cascades (winning symbols clearing so new ones drop in), and the payout.

What goes in
Model
Waylanders-inspired Stake Engine example
Server seed (revealed)
7f3e0d92a4b1e58f9c2d6b1a7e8c4d5f0a91c...
Client seed
ww-edu-2026
Nonce
42
Mode
base

A real checker should give you enough information - server seed, client seed, nonce, mode, and the outcome - that you can re-run the bet on your own computer and get the same answer.

How it gets checked
1
The random draw
serverSeed + clientSeed + nonce + mode
2
The book it draws from
100,000 possible outcomes, each with its own odds
3
The outcome it landed on
book #71860 - probability 0.00001
4
What that outcome played out as
3 cascades - tail feature
5
What it paid
16923.20x your bet for that outcome
Result: 16923.20x the bet. This is one row from the example book, not a real bet on Stake.

This example follows the way Stake Engine lays its math files out - the outcome list, the odds on each, and how each one plays out. It does not contain Valkyrie's real Waylanders Forge math.

Section 06What you cannot verify

The honest list of things provably fair does not solve

Provable randomness is a narrow guarantee. Here is what it does not cover.

x

Third-party slots

Most of the games you actually play - Pragmatic, Hacksaw, Nolimit, NetEnt, Push, Print Studios, Relax, Evoplay - run on certified PRNGs. You cannot re-run them to check the result yourself.

x

Live dealer games

No RNG decides the main result, and there is no provable fairness to check (though some game-show titles do use RNG for bonus multipliers). You are trusting the broadcaster, the dealer, the studio and the auditor.

x

Whether the casino will actually pay you

Mathematical fairness does not promise withdrawal. The math being honest and the operator being honest are separate questions.

x

Whether the casino is operating legally in your jurisdiction

A licence in one country is not a licence in yours. Check before depositing - that is what PIRs are for.

x

Whether the game is appropriate for your bankroll

A 99% RTP game can still take everything you brought. Variance does not respect intent.

The point

Fair math does not make gambling safe for you. Being able to check that the dice were honest is not the same as being able to predict tonight.

Section 07Help and support

If gambling has stopped being something you choose, talk to someone

None of the math on this page changes anything if you are not in control of when you play. The lines below are free, confidential, and not run by us.

New ZealandGambling Helpline0800 654 655 - gamblinghelpline.co.nz
AustraliaGambling Help Online1800 858 858 - gamblinghelponline.org.au
United KingdomGamCare0808 8020 133 - gamcare.org.uk
CanadaConnexOntario1-866-531-2600 - connexontario.ca
United StatesNational Council on Problem Gambling1-800-GAMBLER - ncpgambling.org
Editorial

Researched and reviewed by the Wager Warriors editorial team. We do not receive affiliate revenue for the operators discussed on this page. If something here is wrong or out of date, write to editorial@thewagerwarriors.com.

How Casino Randomness Works | Wager Warriors