How to test casino odds yourself with an RTP simulator

What the tool does

An RTP simulator runs a game against itself. It replays your live settings over a large number of rounds with no animation, and reports the aggregate. In LarpCasino it lives in Settings, runs 10,000, 100,000 or 1,000,000 rounds, and reports five things:

The first two are the pair that matter. Everything below is a way of pulling them apart and watching them come back together.

Experiment 1: watch a small edge do real damage

1. 01Pick a game. Crash or Pinball is the clearest starting point.
2. 02Set the house edge preset to Real Casino, which is 1%.
3. 03Run 1,000,000 rounds.
4. 04Read empirical RTP — it will sit very close to 99% — and then read house profit.

The profit figure is the point of this one. One percent is an abstraction; the number of credits the house kept from a million rounds is not. That is the number a real operator is running its business on, and it came out of a game that looked fair on every individual round.

Experiment 2: prove that cash-out targets do not matter

This is the experiment that kills the most popular strategy in online casinos.

1. 01Open Crash and set the edge to 1%.
2. 02Set an auto cash-out at 1.5× and run 1,000,000 rounds. Note the empirical RTP and the win rate.
3. 03Change the target to 20×. Run 1,000,000 again.

The win rate collapses from around 66% to around 5%. The empirical RTP does not move: both land near 99%. The win chance at any target t is (1 − edge) / t and the payout is t, so the target cancels out. Every exit strategy is the same bet with different volatility, which is worked through in how crash games work.

Experiment 3: separate volatility from return

1. 01Open Pinball, set the edge to 1%, and choose Low risk.
2. 02Run 100,000 rounds. Note empirical RTP and longest losing streak.
3. 03Switch to High risk and run 100,000 again.

Both runs return roughly 99%. The losing streak figure will be dramatically longer on High, and the empirical RTP will sit further from the theoretical one. Same cost, completely different ride — which is the entire content of slot volatility explained, demonstrated in two runs.

Experiment 4: watch convergence happen

This one explains why short sessions tell you nothing.

1. 01Choose a high-volatility configuration — Slots on High, or Pinball on High risk.
2. 02Run 10,000 rounds. Note the gap between empirical and theoretical RTP.
3. 03Run 100,000. Note it again.
4. 04Run 1,000,000. Note it again.

The gap shrinks each time, and not in a straight line — sampling error falls with the square root of the number of rounds, so getting ten times closer takes a hundred times the play. Repeat the 10,000-round run a few times and watch how far it wanders between attempts. That wander is exactly the size of the evidence a real player has after a session.

Experiment 5: run the casino from the other side

1. 01Set the edge to Cruel, which is 10%, and run 100,000 rounds.
2. 02Set it to Player's Edge, which is −3%, and run 100,000 rounds.

The first is roughly what a lottery or a bad slot runs at, and house profit makes it obvious why those products exist. The second inverts the game: the player is favoured, the bankroll grows steadily, and it does so for exactly the same structural reason a real one shrinks. Seeing the same machinery run in reverse is what makes the direction feel like arithmetic rather than bad luck.

Experiment 6: catch a rigged game

The hardest thing to demonstrate is a game whose printed numbers are honest and whose behaviour is not. LarpCasino ships a bias dial specifically so you can build one.

1. 01Open Dice. Its payouts are locked fair, so the printed multiplier carries no edge.
2. 02Run 1,000,000 rounds and confirm empirical and theoretical RTP both sit near 100%.
3. 03Turn on bias and run 1,000,000 again.

The theoretical column does not move, because the paytable did not change. The empirical column does. The lobby badge now reads "Biased" — labelled, because the feature exists to show the failure mode rather than hide it — but a real operator would not label it, and no per-round check would reveal it. That gap is the argument in provably fair explained, reproduced somewhere harmless.

A note on reading the results

• Compare empirical against theoretical, never empirical against your expectations. The theoretical figure is the prediction being tested.
• Any single 10,000-round run is one sample. Run it several times before believing its distance from theory means anything.
• House profit is the number to look at when a percentage stops feeling large. It is denominated in the thing you actually lose.
• Presets clear bias and always-win/lose, so picking one gives you a clean configuration to test.

Why do this at all

Because the alternative is learning the same lessons from a real bankroll at a rate of a few hundred rounds per session, with the sample size that implies. A million simulated rounds costs nothing and settles questions that no amount of real play can.

LarpCasino is a simulated casino for iPhone, rated 18+ for simulated gambling. There is no real money in it, no accounts, no sign-in and no game server. The odds are yours to set, which is the point — it exists to show how the maths behind these games actually works.