Expected Loss, Turnover and RTP
RTP calculations use turnover, not the amount first deposited. A player can deposit A$100 and generate A$600 of turnover by repeatedly staking returned funds. At 96% RTP, the mathematical expected loss relates to that A$600 wagered, not simply to the original A$100 balance.
The basic expected-loss formula
Expected loss = total turnover × house edge
Because house edge is 100% minus RTP, the same formula can be written as:
Expected loss = total turnover × (1 − RTP)
RTP must be expressed as a decimal in the calculation. For example, 96.5% becomes 0.965 and the corresponding edge is 0.035.
| Scenario | Total turnover | RTP | House edge | Mathematical expected loss |
|---|---|---|---|---|
| 100 spins at A$1 | A$100 | 96.5% | 3.5% | A$3.50 |
| 500 spins at A$1 | A$500 | 96.5% | 3.5% | A$17.50 |
| 500 spins at A$1 | A$500 | 94.5% | 5.5% | A$27.50 |
| 1,000 spins at A$0.50 | A$500 | 94.5% | 5.5% | A$27.50 |
The final two rows have the same expected loss because their turnover and RTP are identical. Spin count alone is not the cost driver; stake multiplied by spins is.
Expected loss is not a session forecast
The figures above are mathematical averages, not promises. After 500 spins at A$1, the actual result might be a large loss, a small loss or a profit. High volatility makes that range particularly wide.
I use expected loss to compare the cost of two configurations under equal turnover. I do not use it to tell a player what will happen tonight. Theoretical RTP converges only across substantial play, and even operator-level samples can require hundreds of thousands of games before tolerances become narrow.
Why deposit and turnover diverge
Suppose a player deposits A$100, stakes A$1 and receives A$0.80 back. Staking that A$0.80 again creates additional turnover. Reinvested awards can therefore make total wagers several times larger than the starting balance.
This is also why a wagering requirement is measured against turnover. A bonus asking for A$2,000 of qualifying wagers exposes the player to the game’s edge across A$2,000, even if the cash initially credited was much smaller.
Turnover makes RTP differences accumulate
Compare a 96.5% build with a 94.5% build:
- At A$100 turnover, the expected-loss difference is A$2.
- At A$1,000 turnover, it is A$20.
- At A$10,000 turnover, it is A$200.
The difference grows linearly with turnover. That does not mean actual losses will follow a straight line; volatility continues to dominate short and medium samples.
Stake size changes money risk, not percentage RTP
If the game rules do not specify otherwise, moving from A$0.20 to A$2 per spin does not improve the RTP percentage. It multiplies the monetary exposure by ten. The same 4% edge represents 0.8 cents per A$0.20 spin in expectation and 8 cents per A$2 spin.
Some jackpot or feature configurations can vary by stake level, so the help screen remains the authority. But simply raising the bet because a bonus “looks close” does not change the random probability of the next result.
A Bonanza example
Sweet Bonanza 1000 is highly volatile and can produce wide session results. If comparing a 96.53% build with a confirmed lower configuration, the correct calculation starts with total wagered and the edge attached to each build. Its 25,000x maximum win is irrelevant to the expected-loss formula unless the full probability distribution is being modelled.
The same principle applies to Bonanza Megaways, Big Bass Bonanza and every other game in the catalogue. Mechanics alter the distribution of awards; RTP and turnover determine the long-run expectation.
How I use expected loss responsibly
- Confirm the RTP in the loaded game.
- Set a maximum stake and spin count before play.
- Calculate maximum planned turnover as stake multiplied by spins.
- Multiply turnover by the house edge to compare configurations.
- Treat the answer as an average cost estimate, never as a stop-loss guarantee.
No staking pattern removes the house edge. Reducing turnover, lowering the stake or not playing reduces monetary exposure; changing spin timing does not.
