Slot Volatility vs RTP: Same 96%, Very Different Sessions

Slot volatility is how unevenly a game spreads its return: small amounts often, or large amounts rarely. With the RTP held at exactly 96%, it decides almost everything you feel in a session. In three illustrative pay tables we built, a bankroll of 100 bets went bust within 1,000 spins in 35%, 79% or 93% of sessions, depending only on volatility.

The RTP, the share of all money wagered that a game pays back over its lifetime, is the price. Volatility (also called variance) is the ride. This page measures the ride with exact maths and 10,000 simulated sessions per setting. If you want the price side first, how RTP works and where to read it covers that.

High-volatility sessions that went bust from 100 bets within 1,000 spins (low 35%, medium 79%)

93%

High-volatility sessions that finished ahead after 1,000 spins (low 28%, medium 17%)

5.6%

Average longest run of empty spins in 1,000, high-volatility model (low 10, medium 18)

26

Spins before observed RTP is reliably within 5 points of 96%, high model (low model: 9,660)

1.9m

Source: Our study S6: three made-up pay tables with exactly 96.0% RTP, exact maths plus 10,000 sessions per model and length

Three made-up pay tables with the same 96% RTP

Real slot maths is confidential, so we wrote our own. Each model below is a list of possible results per spin, with a probability for each, set so that the RTP is exactly 96.0%. They are illustrations, not real games: real slots have many more outcomes, bonus rounds and often larger top prizes. What they isolate is the one thing that differs between them, the shape of the payouts.

Low volatilityhits 47.4% of spins, top prize 100x

39.5515.5

Medium volatilityhits 28.7% of spins, top prize 1,000x

253536

High volatilityhits 19.6% of spins, top prize 5,000x

1719.559.5

• Returns up to 2x (includes 0.5x, which still loses half the bet)
• Prizes of 3x to 25x
• Prizes of 50x and above

Source: Our study S6 pay tables (illustrative models, not real games). Numbers under each bar are points of RTP; bars end at 96 of 100 and the gap is the 4% house edge.

In the low-volatility model, 39.5 of the 96 points of RTP come from returns of 2x the bet or less, and only 5.5 points from prizes of 50x or more. The high-volatility model flips that: 59.5 points come from prizes of 50x and above, and 18.5 points from its single top prize of 5,000x, which lands about once in 27,027 spins.

Model	RTP	Spins returning anything	Spins returning more than the bet	Standard deviation per spin (bets)	Variance per spin	Top prize (x bet)
Low volatility	96.0%	47.4%	20.4%	2.51	6.29	100
Medium volatility	96.0%	28.7%	10.7%	10.71	114.67	1,000
High volatility	96.0%	19.6%	6.6%	35.41	1,253.53	5,000

A 'hit' includes 0.5x returns, which still lose half the bet. Variance and standard deviation are exact, in bets.

Source: Our study S6, models[].payTable

The standard deviation per spin is the single number that best sums up volatility. It runs from 2.51 bets in the low model to 10.71 and 35.41 bets in the others. That is a factor of about 14 between the calmest and the wildest game, at an identical price. Studios sometimes publish a rating for this. One game page we checked shows a volatility score of “4 / 5” next to its RTP figures, but does not say how the scale is defined.

Where the spread comes from

Volatility is not a separate setting bolted onto a game. It falls straight out of the pay table. Take one prize from each model. In the low-volatility table, a 2x return lands one spin in ten and contributes 20 of the 96 points of RTP. In the high-volatility table, the top prize of 5,000x lands about once in 27,027 spins and contributes 18.5 points on its own. Both deliver a similar slice of the return, but one arrives every few minutes and the other may never arrive in your lifetime of play.

The variance adds up those differences. Each outcome contributes its probability times the square of its distance from the average return of 0.96 bets. A 2x return sits about one bet from the average; a 5,000x prize sits almost 5,000 bets away, and squaring that distance makes even a tiny probability count for a lot. That is why the high-volatility model’s variance is about 200 times the low model’s (1,253.5 against 6.29), even though both return exactly the same amount on average.

Bust risk: what 1,000 spins does to 100 bets

All three models charge the same 4% of turnover. Yet the chance of losing the whole 100-bet bankroll within 1,000 spins is 35.1% for the low-volatility model, 79.4% for the medium and 93.5% for the high one. These are exact figures from a dynamic programme over every possible balance, and 10,000 simulated sessions per model agree within sampling error.

Source: Our study S6, sessions[1,000 spins].probabilityBustExact

See the numbers as a table

Chance of going bust from 100 bets within 1,000 spins

	Sessions that went bust
Low volatility	35.1%
Medium volatility	79.4%
High volatility	93.5%

Session length matters as much as volatility. No session can go bust within 100 spins, because losing 100 bets takes at least 100 losing spins and play would have ended. After 500 spins, the picture has already split: 8.8% of low-volatility sessions are bust against 65.2% and 88.8% for the others.

Source: Our study S6

See the numbers as a table

Bust probability by volatility and session length

	100 spins	500 spins	1,000 spins
Low volatility	0.0%	8.8%	35.1%
Medium volatility	0.0%	65.2%	79.4%
High volatility	0.0%	88.8%	93.5%

Who finishes ahead, and by how much

Because every model pays back 96% in the long run, the chance of finishing ahead falls the longer you play. It also falls faster when volatility is high. After 1,000 spins, 28.1% of low-volatility sessions were ahead, 16.7% of medium and 5.6% of high.

Model and session	Bust	Finished ahead	25th percentile (bets)	Median final (bets)	75th percentile (bets)	Mean final (bets)	Spins played, average
Low volatility, 100 spins	0.0%	35.0%	78.5	91.0	107.0	96.0	100
Low volatility, 500 spins	8.8%	32.7%	39.0	73.5	112.5	80.3	492
Low volatility, 1,000 spins	35.1%	28.1%	0.5	48.5	109.0	64.9	878
Medium volatility, 100 spins	0.0%	24.8%	50.0	67.5	99.0	96.0	100
Medium volatility, 500 spins	65.2%	22.4%	0.0	0.5	81.0	86.0	351
Medium volatility, 1,000 spins	79.4%	16.7%	0.0	0.5	0.5	80.8	480
High volatility, 100 spins	0.0%	12.8%	27.0	38.5	65.0	96.0	100
High volatility, 500 spins	88.8%	8.4%	0.0	0.5	0.5	91.3	218
High volatility, 1,000 spins	93.5%	5.6%	0.0	0.5	0.5	89.6	259

Bust and ahead are exact; percentiles are from 10,000 simulated sessions per row; mean final and spins played are exact. A median of 0.5 means at least half of sessions went bust.

Source: Our study S6

The spread chart below shows the middle half of results (25th to 75th percentile) with the median as a line. The low-volatility band drifts down slowly. The other two collapse onto the floor by 500 spins, which is what a median of 0.5 bets means: more than half of those sessions are already over.

Shaded band: 25th to 75th percentile of 10,000 sessions

Source: Our study S6, sessions[].finalBalancePercentiles

See the numbers as a table

Middle half of session results, by volatility

Spins	Low volatility	Low volatility (low)	Low volatility (high)	Medium volatility	Medium volatility (low)	Medium volatility (high)	High volatility	High volatility (low)	High volatility (high)
0	100	100	100	100	100	100	100	100	100
100	91	78.5	107	67.5	50	99	38.5	27	65
500	73.5	39	113	0.50	0	81	0.50	0	0.50
1,000	48.5	0.50	109	0.50	0	0.50	0.50	0	0.50

Five sessions from each end

Percentiles hide what a single session looks like. The study stores 40 sample sessions per model. Here are the first five stored sessions of the low-volatility model, unselected:

Source: Our study S6, models[low].samplePaths (sessions 1 to 5 of 40)

See the numbers as a table

First five stored sessions, low-volatility model

Spins	Session 1	Session 2	Session 3	Session 4	Session 5
0	100	100	100	100	100
10	95.5	97	91.5	101	97.5
20	91.5	92	88.5	113	96.5
30	84.5	89	89.5	111	89
40	82.5	133	86.5	115	107
50	82	126	84.5	119	102
60	75	119	78	117	94
70	73.5	113	75	131	91
80	71	110	74.5	137	90
90	64.5	107	72.5	135	89
100	55.5	104	70	138	82.5
110	106	103	63.5	136	86
120	101	98.5	58.5	130	84
130	92	89.5	55	128	83.5
140	97.5	81	45.5	121	91
150	96.5	74	43.5	118	89
160	95.5	76	37	113	82
170	100	69	36.5	109	85.5
180	96.5	64	42	108	83
190	99.5	77.5	42	113	80
200	94.5	83	39	110	75
210	91	75	36.5	108	70.5
220	85	80.5	29.5	117	64
230	84	76.5	26	111	59.5
240	86.5	92.5	25	109	64.5
250	84	91.5	19	106	64.5
260	76	86	12	113	64
270	76.5	84	7	106	67
280	73.5	89	3.50	100	71
290	66.5	90	5	98.5	77
300	64	82	4.50	92.5	77.5
310	58.5	77	0.50	94	70.5
320	57	78	0.50	88	104
330	64.5	74	0.50	85.5	111
340	58.5	69.5	0.50	79.5	107
350	56	62.5	0.50	79	105
360	49	53	0.50	78.5	106
370	51.5	47	0.50	123	100
380	56	52	0.50	123	94
390	49	48.5	0.50	120	97.5
400	41	46.5	0.50	124	95
410	45.5	45.5	0.50	120	96
420	44	47.5	0.50	128	89.5
430	47.5	44	0.50	132	87.5
440	49	48.5	0.50	135	88.5
450	53	46	0.50	131	82.5
460	54	43.5	0.50	123	76.5
470	53.5	41	0.50	126	80
480	49.5	64	0.50	122	80.5
490	46	66	0.50	118	76.5
500	40	69.5	0.50	116	85
510	34	80	0.50	109	79
520	31.5	82.5	0.50	108	86
530	23	89.5	0.50	102	89
540	15.5	83	0.50	116	83.5
550	16	81	0.50	116	84.5
560	9.50	118	0.50	116	78
570	4.50	125	0.50	107	74.5
580	11	136	0.50	103	69.5
590	7	180	0.50	104	75
600	2	181	0.50	102	75
610	0.50	186	0.50	96	69.5
620	0.50	187	0.50	89	65.5
630	0.50	180	0.50	85.5	72.5
640	0.50	177	0.50	80	70.5
650	0.50	173	0.50	74.5	72
660	0.50	171	0.50	76	73
670	0.50	170	0.50	73.5	76.5
680	0.50	174	0.50	74.5	79
690	0.50	168	0.50	92	83.5
700	0.50	169	0.50	113	76.5
710	0.50	164	0.50	110	76
720	0.50	167	0.50	108	71.5
730	0.50	180	0.50	107	69.5
740	0.50	175	0.50	126	62.5
750	0.50	171	0.50	124	58.5
760	0.50	166	0.50	127	53.5
770	0.50	173	0.50	126	54
780	0.50	168	0.50	135	48.5
790	0.50	160	0.50	141	45
800	0.50	162	0.50	145	39
810	0.50	161	0.50	145	35
820	0.50	170	0.50	146	31.5
830	0.50	160	0.50	172	28
840	0.50	159	0.50	175	23.5
850	0.50	163	0.50	174	28.5
860	0.50	158	0.50	167	27.5
870	0.50	159	0.50	161	21
880	0.50	153	0.50	166	25.5
890	0.50	145	0.50	161	20
900	0.50	145	0.50	156	17
910	0.50	136	0.50	157	13
920	0.50	129	0.50	154	10
930	0.50	125	0.50	152	8.50
940	0.50	122	0.50	154	4.50
950	0.50	126	0.50	148	7
960	0.50	123	0.50	149	8.50
970	0.50	119	0.50	141	22
980	0.50	118	0.50	137	21.5
990	0.50	116	0.50	154	14.5
1,000	0.50	106	0.50	152	13

And the first five of the high-volatility model:

Source: Our study S6, models[high].samplePaths (sessions 1 to 5 of 40)

See the numbers as a table

First five stored sessions, high-volatility model

Spins	Session 1	Session 2	Session 3	Session 4	Session 5
0	100	100	100	100	100
10	93	91	92	92.5	110
20	94	84	82.5	83.5	102
30	84	76.5	79	74.5	95.5
40	74	67	73.5	65	86
50	64.5	60	64.5	69	78.5
60	57	51	56	68.5	72
70	57.5	43	46.5	59.5	62.5
80	50	35	36.5	52.5	52.5
90	41.5	26	27.5	51	44
100	32.5	18	22.5	47	35
110	23	9	21.5	37	28.5
120	18.5	0	12	28	21.5
130	11	0	3	20	26.5
140	3	0	0	16.5	20
150	0	0	0	9	20
160	0	0	0	1	12
170	0	0	0	0	3
180	0	0	0	0	0.50
190	0	0	0	0	0.50
200	0	0	0	0	0.50
210	0	0	0	0	0.50
220	0	0	0	0	0.50
230	0	0	0	0	0.50
240	0	0	0	0	0.50
250	0	0	0	0	0.50
260	0	0	0	0	0.50
270	0	0	0	0	0.50
280	0	0	0	0	0.50
290	0	0	0	0	0.50
300	0	0	0	0	0.50
310	0	0	0	0	0.50
320	0	0	0	0	0.50
330	0	0	0	0	0.50
340	0	0	0	0	0.50
350	0	0	0	0	0.50
360	0	0	0	0	0.50
370	0	0	0	0	0.50
380	0	0	0	0	0.50
390	0	0	0	0	0.50
400	0	0	0	0	0.50
410	0	0	0	0	0.50
420	0	0	0	0	0.50
430	0	0	0	0	0.50
440	0	0	0	0	0.50
450	0	0	0	0	0.50
460	0	0	0	0	0.50
470	0	0	0	0	0.50
480	0	0	0	0	0.50
490	0	0	0	0	0.50
500	0	0	0	0	0.50
510	0	0	0	0	0.50
520	0	0	0	0	0.50
530	0	0	0	0	0.50
540	0	0	0	0	0.50
550	0	0	0	0	0.50
560	0	0	0	0	0.50
570	0	0	0	0	0.50
580	0	0	0	0	0.50
590	0	0	0	0	0.50
600	0	0	0	0	0.50
610	0	0	0	0	0.50
620	0	0	0	0	0.50
630	0	0	0	0	0.50
640	0	0	0	0	0.50
650	0	0	0	0	0.50
660	0	0	0	0	0.50
670	0	0	0	0	0.50
680	0	0	0	0	0.50
690	0	0	0	0	0.50
700	0	0	0	0	0.50
710	0	0	0	0	0.50
720	0	0	0	0	0.50
730	0	0	0	0	0.50
740	0	0	0	0	0.50
750	0	0	0	0	0.50
760	0	0	0	0	0.50
770	0	0	0	0	0.50
780	0	0	0	0	0.50
790	0	0	0	0	0.50
800	0	0	0	0	0.50
810	0	0	0	0	0.50
820	0	0	0	0	0.50
830	0	0	0	0	0.50
840	0	0	0	0	0.50
850	0	0	0	0	0.50
860	0	0	0	0	0.50
870	0	0	0	0	0.50
880	0	0	0	0	0.50
890	0	0	0	0	0.50
900	0	0	0	0	0.50
910	0	0	0	0	0.50
920	0	0	0	0	0.50
930	0	0	0	0	0.50
940	0	0	0	0	0.50
950	0	0	0	0	0.50
960	0	0	0	0	0.50
970	0	0	0	0	0.50
980	0	0	0	0	0.50
990	0	0	0	0	0.50
1,000	0	0	0	0	0.50

The high-volatility lines look like staircases going down: long runs of empty spins with the occasional small return. The sixth stored session is the exception. A big hit lifted it to 674.5 bets by spin 120, and it ended at 692 bets. Across all 40 stored sessions per model, 14, 35 and 37 went bust, and 12, 5 and 2 finished ahead. The best stored finishes were 221.5, 802 and 1,295 bets.

The average hides the typical session

The expected loss is the same 4% of whatever you wager in every model. That does not mean the typical session loses 4%.

In the high-volatility model over 1,000 spins, the mean final balance is 89.6 bets, but the median is 0.5 bets: most sessions bust. The mean is held up by a few sessions with very large wins. That is the general pattern with high-volatility slots: the long-run average is made of many small and medium losses plus rare large wins, and most players never see the rare part.

There is a second twist. A bankroll that goes bust stops wagering, so it stops paying the 4%. The high-volatility sessions played only 259 spins on average out of 1,000, so they lost about 10.4 bets on average. Low-volatility sessions played 878 spins on average and lost about 35.1 bets. The low-volatility game did not cost more per spin; it simply kept you spinning for longer. Over the same number of spins, the cost is identical.

Dry spells: the longest run of nothing

A “dry” spin is one that returns nothing at all. In the low-volatility model 52.6% of spins are dry; in the high one, 80.4%. Over 1,000 spins, the expected longest dry run is 10 spins for the low model, 17.9 for the medium and 26.4 for the high one.

Source: Our study S6, sessions[1,000 spins].longestDryStreak.expectedExact

See the numbers as a table

Average longest run of empty spins in 1,000 spins

	Spins in a row that returned nothing
Low volatility	10.0
Medium volatility	17.9
High volatility	26.4

Model	Average	Median	95th percentile	Chance of 20+ in a row	Chance of 30+ in a row
Low volatility	9.99	10	14	0.12%	0.00%
Medium volatility	17.93	17	25	27.8%	1.08%
High volatility	26.38	25	37	92.7%	24.4%

Exact run-length recurrence; the simulation averages were 9.98, 17.98 and 26.4.

Source: Our study S6

A run of 30 empty spins is a freak event on the low-volatility game (about 2 in a million sessions of 1,000 spins) and a one-in-four event on the high-volatility game (24.4%). Neither tells you anything about the next spin. Each spin of a certified game is independent, and a long drought does not make a prize “due”. Whether long droughts point to a fixed game is covered in our look at RNG testing and fairness.

How many spins before RTP shows up

You can measure a slot’s RTP by watching it, but only over a very large number of spins. The spread of the observed return shrinks with the square root of the number of spins, and it starts much wider for volatile games.

Source: Our study S6, rtpConvergence.ci95HalfWidthPctPoints

See the numbers as a table

95% range of the observed RTP, plus or minus percentage points

	1,000 spins	10,000 spins	100,000 spins	1,000,000 spins	10,000,000 spins
Low volatility	15.5	4.9	1.6	0.5	0.2
Medium volatility	66.4	21.0	6.6	2.1	0.7
High volatility	219.4	69.4	21.9	6.9	2.2

Model	SD per spin (bets)	±5 points	±1 point	±0.5 points
Low volatility	2.51	9,660	241,480	965,917
Medium volatility	10.71	176,212	4,405,294	17,621,174
High volatility	35.41	1,926,230	48,155,740	192,622,957

Spins = (1.96 × sd ÷ margin)². At ten spins a minute, 1,926,230 spins is more than 3,000 hours.

Source: Our study S6

For the low-volatility model, ±5 points takes 9,660 spins. For the medium model it takes 176,212 and for the high model 1,926,230. No player’s personal history comes close to the high figure, which is why a winning or losing streak on a volatile game says almost nothing about its RTP.

Does a bigger bankroll change things?

Our study fixed the bankroll at 100 bets and also ran every session with no bankroll limit at all, so we can compare those two settings. We did not compute sizes in between.

Model	Ahead, 100-bet bankroll	Ahead, no bankroll limit	SD of net result (bets)
Low volatility	28.1%	28.9%	79
Medium volatility	16.7%	33.0%	339
High volatility	5.6%	21.9%	1,120

'100-bet' values are exact; 'no limit' values are from 10,000 simulated sessions. The SD of the net result is for 1,000 spins without a bankroll limit.

Source: Our study S6

For the low-volatility game, the bankroll barely matters: 28.1% versus 28.9%. Its swings over 1,000 spins (a standard deviation of about 79 bets) rarely reach the floor before a recovery. For the high-volatility game, the limit is the whole story: 5.6% finish ahead with 100 bets, but 21.9% would with unlimited money, because most of its winning sessions needed to survive a deep hole before the big prize arrived. Its standard deviation over 1,000 spins is about 1,120 bets, more than 11 times the bankroll.

In practice, your bankroll in bets is your bankroll divided by your stake. Halving the stake doubles the number of bets you hold, which on a volatile game is the only thing that gives the rare prizes time to show up. It does not change the 4% price per dollar wagered, and it does not turn a losing game into a winning one.

Four common misreadings

“High volatility means a higher RTP.” No. All three of our models return exactly 96.0%. A studio can pair any volatility with any RTP. The two numbers are set separately, and a game page may show both, as the studio page we checked does.

“A low hit rate means the game is stingy.” Only if you measure stinginess by frequency. The high-volatility model pays something on 19.6% of spins against 47.4% for the low model, yet both return the same 96 cents per dollar wagered. The missing frequency is paid back as size.

“Lots of wins means I am doing well.” Hits include returns of half the bet. In the low-volatility model, 15% of spins return 0.5x and 12% return exactly the stake. Only 20.4% of spins make money. A screen full of small wins can sit on top of a steady loss.

“A volatile game is due after a long drought.” Each spin is independent. After 30 empty spins, the chance that the next spin returns nothing is still 80.4% in the high-volatility model, exactly as it was on the first spin.

What a short session can and cannot tell you

What you can take from one session:

• How the game feels. Long runs of empty spins and the size of the prizes you see are a fair hint of where a game sits on the volatility scale.
• Your own budget limits. If 100 bets vanished in 150 spins, the game is too volatile for that stake, whatever its RTP.

What you cannot take from it:

• The RTP. Even 1,000 spins on our calmest 96% model only pins the observed return to within about 15.5 points either way. On the high model the range is ±219 points.
• Whether a game is “hot” or “cold”. Spins on a certified game are independent; past results do not change the next one.
• Whether a game is fair. A 93% bust rate is what a perfectly fair 96% game with high volatility produces. Losing streaks on their own are not evidence of rigging.
• That a win is coming. In our high model, half of all 1,000-spin sessions saw a run of at least 25 empty spins.

The honest summary is that volatility is a choice about risk, not about value. If you play at all, pick the RTP from the game’s information screen, pick the volatility that matches the bankroll you are willing to lose, and decide your stopping point before the first spin.

What we could not verify, and our assumptions

• Real games’ volatility. Our three pay tables are invented. We do not know the pay table, standard deviation or hit frequency of any real slot, and we make no claim that any real game behaves like one of our models.
• Volatility ratings. One studio page shows a “4 / 5” volatility score, but we found no published, common definition of such scales, so we do not map them onto our models.
• Bonus rounds and features. Our models have one independent result per spin. Free-spin rounds, multipliers and bought features change the spread in ways our models do not show.
• Bankroll sizes between 100 bets and unlimited. We compare only those two. Other bankroll sizes were not computed.
• Normal approximation. The convergence figures assume the average of many spins is roughly normal. For the high-volatility model that is rough below about 100,000 spins.
• Wizard of Odds on standard deviation. It calls slot standard deviation “very high” and gives one figure “based on just one machine”. We did not use that figure because it describes a single unnamed game.