RTP and volatility

A slot's information panel prints two numbers that are often read as one, and they answer different questions: the "RTP" (return to player) says what a very long run of play returns, while "volatility" says how far a short run can wander from that average, which is the only run a player ever actually plays. This page is about the statistics behind both.

We use two imaginary machines rather than real games, because their arithmetic fits in a line, and a hypothetical player who stakes $2 a spin for 1,000 spins on each of them. The page on how a slot works, from the random number generator to the information panel, is the place to start for anyone who has not read it.

Two machines with the same RTP

The first, "machine A", pays 96 times the stake on one spin in a hundred and nothing on the others. The second, "machine B", pays 1.92 times the stake on one spin in two. Both return 96 per cent of what is staked (0.01 times 96 for the first, 0.5 times 1.92 for the second), so both keep four cents per dollar in the long run, and a table of the kind we print below would list them on the same row.

What separates them is the "variance" (the average of the squared distances between each result and the mean), or more usefully its square root, the "standard deviation", which is the typical distance of one spin from the average. Per dollar staked it is about $9.55 on machine A and $0.96 on machine B; the first machine is ten times as volatile as the second.

Now our player. On 1,000 spins at $2 the total staked is $2,000 on either machine and the expected loss is $80 on both (four per cent of $2,000). Machine B's player ends the session, nineteen times out of twenty, somewhere between a loss of about $200 and a gain of about $40, while machine A's player, in the same share of sessions, ends anywhere between a loss of $1,232 and a gain of $1,264.

The more volatile machine is also the one that sends the player home ahead more often, about 42 sessions in 100 (eleven wins or more out of a thousand spins) against fewer than 10 in 100 on machine B, which is worth knowing before choosing between them. The price is paid in the other sessions, where the losses on machine A run to several hundred dollars!

The swing grows more slowly than the loss

The typical "swing" of a session grows with the square root of the number of spins, not with the number itself. A session a hundred times longer swings only ten times wider. The expected loss, on the other hand, grows in proportion: the player who spins a hundred times as much loses, on average, a hundred times as much.

This is the whole mechanism of the "long run" in two lines. Over a short session the swing dominates and the RTP is invisible, which is why you can win on a 92 per cent game and lose on a 98 per cent one on the same evening; over a long enough run the expected loss outgrows the swing, and in our view this "law of large numbers" (the rule by which an average settles as the trials pile up) is the only sense in which an RTP is ever paid.

How many spins before the average means something

A practical threshold is the point at which the expected loss becomes twice the standard deviation of the session, since from there on a profit is an unusual outcome rather than an ordinary one. On machine B that point arrives after about 2,300 spins, which at a hypothetical pace of 600 spins an hour is less than four hours of your time.

On machine A it arrives after about 228,000 spins, that is 380 hours at the same pace! For a volatile slot the RTP in the panel describes the game's whole audience (every spin by every player, added together), not any one player's evenings.

What a given RTP costs

Because the long-run cost of play is the turnover multiplied by the edge, the RTP converts directly into dollars, and the table we give below does it for $1,000 of total bets, whether staked in ten spins or in five thousand (the size of each bet does not enter the calculation).

Expected loss per $1,000 wagered, by RTP

RTP	House edge	Expected loss per $1,000 wagered	Example
98.94%	1.06%	$10.60	Baccarat, banker bet, eight decks
97.30%	2.70%	$27.03	Roulette with a single zero
96.00%	4.00%	$40.00	A slot set at 96 per cent
94.74%	5.26%	$52.63	Roulette with a double zero
92.00%	8.00%	$80.00	A slot set at 92 per cent
88.00%	12.00%	$120.00	A slot set at 88 per cent

The table says nothing about volatility, and we think it should not, since two games on the same row cost the same over $1,000 of bets whatever their swings. What volatility changes is the "bankroll" (the sum a player brings to the session) that you need to reach that $1,000 of bets without running dry first, and the "risk of ruin" (the chance of losing all of it before the end) falls when the stake is kept small and constant, which is the method of flat betting.

The same idea as the house edge

A slot's RTP and a table game's house edge are one number seen from two sides, since the edge is simply what the RTP leaves out (100 minus 97.30 gives the 2.70 per cent of single-zero roulette). The difference lies in how the figure is known: a table game publishes its rules, from which anyone can calculate the edge, while a slot publishes the RTP because the weights of its reels cannot be seen.

Blackjack is the exception that proves the arithmetic, because its edge depends on the player's decisions; no single RTP describes every hand that is played. Video poker is similar, and a figure that assumes "perfect play" (the best decision on every hand) overstates the return of a player who plays by instinct, which is worth remembering when a panel prints one.

Who checks the figures

Neither the "RNG" (the generator of random numbers) nor the RTP is taken on trust in a regulated market. The Gambling Commission, which regulates gambling in Great Britain, requires its licensees to appoint an approved "test house" when its "testing strategy" calls for third-party testing, and the list we read on its site (approved as of 2 March 2026) has fourteen entries, among them eCOGRA, BMM Testlabs and several companies of the GLI group (Gaming Laboratories International).

eCOGRA describes its RTP testing as simulation on large samples, with a report on whether a game's payout falls within a "statistically acceptable interval" given its volatility, which is, to us, this page in the language of a laboratory. Other laboratories, such as iTech Labs (which is not on the British list), describe similar work on their own sites; what a certificate covers is the "theoretical" game, not the luck of any casino's month.

For a progressive slot the certified RTP includes the "contribution" to the jackpot, which the base game does not pay back (see our page on progressive jackpots). We would read the two figures of a panel together, then, whichever game you open: the RTP to know the price of the long run, the volatility to know how rough the short one will be.

Questions

By the Rand Casinos editorial team · Last updated October 2026 · Online since 2007