basics
Why volatility is a different measurement from RTP, what high variance does to a bankroll across a single session, why bonus buys purchase variance rather than value, and how much a provider's own volatility rating is worth.
Two slots can both return 96% and feel nothing alike. One drips small wins for an hour. The other takes your balance in twenty minutes or hands you six hundred times your stake. The number they share is the average. The thing that separates them is volatility.
Volatility is the less discussed of the two figures and the one that decides what your session actually looks like. This guide sets out what it measures, how to see its effect in arithmetic rather than adjectives, and how much weight a provider's own rating deserves.
Return to player is a single point: the long-run average proportion of turnover a game pays back. It is covered in full in RTP explained.
Volatility describes how far individual results scatter around that point. Statisticians call it variance or standard deviation. Slot marketing calls it low, medium or high.
The relationship is worth stating precisely, because it is the source of most confusion. RTP tells you where the distribution is centred. Volatility tells you how wide it is. Neither constrains the other. A 96% game can be built with almost any spread the designer wants, and a 99% game can be built to be brutal.
Volatility does not change the house edge. It changes how long the edge takes to express itself, and how wide the range of outcomes is on the way.
The cleanest way to see this is with a toy example where every number is checkable.
Game A pays 2x your stake on 48% of spins and nothing otherwise. Expected return is 0.48 × 2 = 0.96, so 96% RTP.
Game B pays 96x your stake on 1% of spins and nothing otherwise. Expected return is 0.01 × 96 = 0.96, so 96% RTP.
Identical RTP. Identical house edge of 4%. Now play 100 spins of $1 at each.
Game A produces about 48 wins. The balance wobbles and drifts slowly downward. After 100 spins you are typically holding somewhere near $96, and you almost never hold nothing.
Game B is a different object entirely. The chance of no win at all across 100 spins is 0.99 raised to the power 100, which is about 36.6%.
| Outcome over 100 spins of Game B | Probability | Balance from $100 staked |
|---|---|---|
| No win | about 36.6% | $0 |
| Exactly one win | about 37.0% | $96 |
| Two or more wins | about 26.4% | $192 or more |
Read the first row again. More than a third of sessions end with nothing, on a game with the same 96% RTP as the one that felt gentle. The average across all those sessions is still $96. No individual session resembles the average, because the average falls in a gap between outcomes that actually occur.
That gap is volatility. Real slots are far more complex than Game B, but the mechanism is exactly this one.
The practical consequence is that the median result and the mean result separate.
A $100 bankroll played in $1 spins at 96% RTP loses an expected 4 cents per spin. Divide and you get a mean survival of about 2,500 spins, since winnings are restaked and recycled.
On a low volatility game, real sessions cluster reasonably close to that figure. On a high volatility game, most sessions end far sooner, and the average is held up by a minority of sessions that run very long after a large hit. The typical experience is worse than the average experience. Both statements are true at once.
This is why the same deposit feels generous on one game and vanishes on another. Nothing about the operator or the maths has changed. The width of the distribution has.
Three practical effects follow.
Bankroll requirements scale with volatility, not with RTP. A high variance game needs a far deeper bankroll to give the same chance of surviving a set number of spins.
Short sessions tell you nothing. The smaller the sample, the more the outcome is dominated by variance rather than by the underlying return.
Wagering requirements are harder to clear on high volatility games. The turnover figure is the same, but the chance of the balance reaching zero before you get there is higher.
Volatility is usually driven by two design levers you can often see in the game information screen.
Hit frequency is the proportion of spins that return anything at all. Maximum win is the largest multiple of stake the game can pay, commonly stated as something like 5,000x or 10,000x.
A high maximum win must be paid for out of the same RTP budget. If a game reserves a slice of its return for an outcome that occurs once in hundreds of thousands of spins, that slice is not available for the small wins that keep a session alive. Large top prizes and frequent small wins are in direct competition, and the paytable is where the trade is made.
So a game advertising a very large maximum win is telling you something about its volatility whether or not it prints a rating.
The clearest demonstration available is a game where you choose the volatility yourself and the distribution is published.
Plinko with 16 rows offers low, medium and high risk settings. The maximum multipliers are 16x on low, 110x on medium and 1000x on high. The buckets follow a binomial distribution, and the outermost bucket occurs about once in 32,768 drops.
| Setting | Maximum multiplier | Character of the distribution |
|---|---|---|
| Low | 16x | Most drops land near the centre for small returns |
| Medium | 110x | Wider spread, centre pays less |
| High | 1000x | Centre buckets pay well under stake, value concentrated in the rare outer lanes |
The RTP does not move between these settings. The shape does. On the high setting the central buckets return less than you staked, so the frequent outcome is a small loss, funding a payout you will probably never see. That is high volatility stated as arithmetic rather than as a label. You can work through the full bucket table in the Plinko odds tool and the game page at /games/plinko.
The same logic runs through the other originals. In Crash, the chance of reaching a multiplier of x is approximately one minus the house edge, divided by x. On a 99% RTP version, reaching 10x happens about 9.9% of the time and reaching 100x about 0.99% of the time. Cashing out early is a low volatility strategy. Holding for 100x is a high volatility one. The house edge is identical in both cases.
A bonus buy lets you pay a fixed multiple of your stake, often 100x, to trigger the feature round immediately instead of waiting for it.
The price is not arbitrary. It is set so that the expected return of the purchased feature is close to the buy price multiplied by the game's RTP. Pay $100, expect roughly $96 back, lose an expected $4. That is the same edge you were paying anyway.
So the buy does not change your expected return per dollar wagered. It changes how fast dollars are wagered.
| Play pattern | Turnover per hour | Expected loss per hour at 4% edge |
|---|---|---|
| $1 spins, 600 per hour | $600 | $24 |
| $2 spins, 600 per hour | $1,200 | $48 |
| $100 buys, 30 per hour | $3,000 | $120 |
| $100 buys, 60 per hour | $6,000 | $240 |
The edge per dollar never moved. The loss rate per hour multiplied by ten. A bonus buy compresses hours of base play into seconds, which compresses the cost with it.
Two further details. Some games publish a separate RTP for the buy feature, and it is not always the same as the base game, so check it rather than assuming. And a purchased feature is itself high variance, so the distribution of outcomes around that $96 expectation is very wide. You are paying full price for a lottery ticket you would otherwise have received eventually as part of normal play.
Treat it as a rough signal, not a measurement.
There is no industry standard definition of low, medium or high. Each studio sets its own scale, and a five-bar indicator from one provider is not comparable to a five-bar indicator from another. Some publish a numerical volatility index in the help screen. Some publish hit frequency. Most publish neither.
What is worth looking for, in order of usefulness: a stated hit frequency, the maximum win multiplier, the paytable itself, and whether the game ships in multiple RTP configurations. That last point matters, because the same title can be deployed at different returns by different operators. Spribe's Aviator runs at 97% and is configurable to 96% or 94%. The version you are playing is a property of the operator's build, not of the game's reputation. The RTP comparison page tracks where published figures differ.
The honest summary is that a volatility label tells you what the studio's marketing department chose to call the game. The paytable tells you what the game is. When the two disagree, believe the paytable.
None of this changes the direction of the maths. Volatility decides how bumpy the road is. The house edge decides where it leads. If session length or stake size is becoming difficult to control, the responsible gambling page lists support services.