Two slot games can both be published with a theoretical RTP of 96% and still deliver very different sessions. RTP describes the average amount returned across an enormous number of spins. It says nothing about how that return arrives: in many small payments, in a few huge ones, or in a mix of both. The measure that captures the second part is variance, usually called volatility in player-facing product descriptions. Understanding the gap between the two is the single most useful piece of slot literacy a curious player can have.
What RTP actually measures
Return to player is an expected value. If a game has a theoretical RTP of 96%, then over a very large number of spins, the total paid out is expected to approach 96% of the total staked. The remaining 4% is the house edge. That figure is calculated from the game’s math model, meaning the reel strips or equivalent symbol weights, the paytable and the feature rules, and not from any observed run of results.
It helps to be precise about two terms used throughout this article. Gross return is the total amount a game pays back to you, ignoring what you staked. Net profit is gross return minus total stakes. If you stake 100 units across a session and the game returns 120 units, your gross return is 120 and your net profit is 20. If it returns 70, your net profit is negative 30. RTP is a statement about average gross return relative to stakes, so a 96% RTP corresponds to an expected net result of negative 4 per 100 staked. The word “average” is doing a great deal of work in that sentence.
Expected return versus spread
Think of any single spin as drawing a payout from a distribution. The mean of that distribution is the RTP. The spread is how far individual outcomes typically sit from that mean, and variance and standard deviation are the usual ways to quantify it. Two distributions can share an identical mean and look nothing alike. One might pay something small on a large share of spins. The other might pay nothing on most spins and compensate with rare large wins.
For a player, the mean is invisible in the short run. What you experience is the spread: how long the dry stretches are, how often a spin returns something, how big the occasional hit is, and how far your balance wanders from the straight line the average would predict. That experience is what players are describing when they call a slot “tight,” “streaky” or “swingy,” even though two such games could carry the same published RTP.
Two toy distributions with the same mean
The following examples are hypothetical and deliberately simplified. They are not taken from any real game. Imagine a stake of 1 unit per spin and two made-up games, each with a theoretical RTP of 96%, meaning an average gross return of 0.96 per spin.
Game A: frequent, modest hits
In Game A, assume the following outcomes per spin: 40% of spins pay 1.0 (your stake back), 20% pay 2.0, 2% pay 10.0, and the remaining 38% pay nothing. The average is 0.40 × 1.0 + 0.20 × 2.0 + 0.02 × 10.0 = 0.40 + 0.40 + 0.20 = 1.00. That is slightly too generous, so adjust the top tier: let 1.6% of spins pay 10.0 and 38.4% pay nothing. The average becomes 0.40 + 0.40 + 0.16 = 0.96. Here, 61.6% of spins return something, though a payment of exactly 1.0 only returns the stake and yields no net profit.
Game B: rare, large hits
In Game B, assume 10% of spins pay 2.0, 0.5% pay 100.0 and 89.5% pay nothing. The average is 0.10 × 2.0 + 0.005 × 100.0 = 0.20 + 0.50 = 0.70. To reach 0.96, raise the top tier to 0.78% paying 100.0 and set the 2.0 tier at 9%: 0.09 × 2.0 + 0.0078 × 100.0 = 0.18 + 0.78 = 0.96. Now only about 9.8% of spins return anything at all, and most of the expected value is concentrated in an event that happens fewer than once per hundred spins.
Same mean, different experience
Both games average 0.96 gross per unit staked. Yet Game A returns something on roughly six spins in ten, while Game B returns something on roughly one spin in ten. Hit frequency, the share of spins that pay anything, is therefore very different. A player on Game A sees the balance tick along with frequent small movements. A player on Game B may watch dozens of consecutive spins return nothing, then occasionally see a large jump. Neither pattern signals that the game is broken or “due” for anything; each is simply what its distribution produces.
A subtle point: hit frequency and volatility are related but not identical. A game can have a high hit frequency that is mostly stake-back payments, which feel active but produce no net profit, and still be fairly volatile if a rare large payout carries much of the value. Conversely, a low hit frequency does not by itself guarantee extreme swings. Variance depends on the full shape of the payout distribution, including how large the tail wins are relative to the stake.
Why the same RTP can produce very different short-term results
Over a short session, outcomes cluster loosely around the expected value but with wide scatter. In Game A, the standard deviation of a single spin’s gross return is modest, so a few hundred spins will usually land reasonably near the long-run average, though not exactly on it. In Game B, the standard deviation is much larger because of the 100.0 tier. A few hundred spins could easily contain no top-tier hit at all, leaving the session gross return far below 0.96 per spin, or a single top-tier hit could push it well above. Both results are fully consistent with the same RTP.
This is why RTP alone cannot tell you how much you are likely to win or lose in an evening. It is a long-run average over a sample far larger than any individual will play. Volatility tells you how wide the plausible range of results is around that average over the sample you actually play. A player with a small bankroll and a high-volatility game faces a higher chance of exhausting it before the rare large payouts arrive, even though the theoretical return is unchanged.
The recovery myth
A common intuition holds that after a long losing stretch the game must “owe” you a payout, so that results will swing back toward the average. In a game with independent spins, that is not how probability works. Each spin is drawn from the same distribution regardless of what came before. A run of nothing in Game B does not raise the chance of the 100.0 tier on the next spin; it remains about 0.78% in the toy model.
What is true is subtler. Over an immense number of spins, the average per spin converges toward the mean, because the large sample dilutes early deviations. That convergence happens through dilution, not through compensation. Early losses are not clawed back; they simply become a small fraction of a much larger total. For a player who stops after a few hundred spins, no such convergence has occurred, and the shortfall stays a shortfall.
Reading volatility labels sensibly
Studios and operators often describe games as low, medium or high volatility. These labels are informal and not standardised across the industry, so one provider’s “medium” may resemble another’s “high.” They are best read as rough guidance about distribution shape: low suggests more frequent, smaller payments, and high suggests rarer, larger ones. Combined with a published theoretical RTP, they give a fuller picture than either number alone, but neither predicts a specific session.
The practical takeaway is modest and mathematical. RTP answers “what is returned on average over a vast number of spins?” Volatility answers “how bumpy is the path, and how far from average might my actual session land?” Treating them as separate questions explains why two games with identical headline RTP can feel like entirely different products, and why neither feeling reveals anything about what the next spin will do.