Slot RTP (return to player) is the percentage of all the money wagered on a slot that the game is built to pay back as prizes over its lifetime. A 96% RTP game keeps 4% of everything staked on it on average, so $1,000 of $1 spins costs $40 in expectation, and that 4% is the house edge.
That is the whole definition. The rest of this page covers where to find the figure, why a single game can list more than one, what counts as a good number, and why RTP tells you so little about the next hour of play.
- Expected cost of 1,000 spins at $1 with a 96% RTP ($80 at 92%)
- $40
- S6 arithmetic
- Slot house edge range, so an RTP of 85% to 98%, from one major manufacturer
- 2% to 15%
- Wizard of Odds
- RTP versions listed for one title on a game studio’s own page (88.42% to 96.38%)
- 4
- Hacksaw Gaming
- Percentage points the observed RTP can sit from 96% after 1,000 spins, even in our calmest model
- ±15.5
- S6, 95% range
Source: Sources listed at the end of the page; S6 is our own study with made-up pay tables
- 96% RTPhouse edge 4%$40 kept per $1,000
- 94% RTPhouse edge 6%$60 kept per $1,000
- 92% RTPhouse edge 8%$80 kept per $1,000
Source: Our study S6, rtpComparison (expected loss = (1 - RTP) x total wagered).
RTP in one formula
The Gaming Laboratories International standard for online games, GLI-19, defines return to player as “a ratio of the ‘total amount won’ to the ‘total amount wagered’ by a player”. It adds that the figure may be “theoretical” (from mathematical calculations or simulations) or “actual” (from what the game has metered in real play). The number in a slot’s help screen is the theoretical one.
From that definition, two identities follow:
house edge = 100% − RTP
expected loss = (100% − RTP) × total amount wagered
The second line matters most. The cost of playing a slot does not depend on your deposit. It depends on how much you put through the reels. Fifty $2 spins and one hundred $1 spins carry the same expected loss, because both wager $100. A $50 deposit that is won, re-staked and lost again can easily turn over $500, and at 96% that turnover costs $20 on average.
Source: Our study S6, rtpComparison
See the numbers as a table
| Average loss on $1,000 wagered | |
|---|---|
| 96% RTP | $40 |
| 94% RTP | $60 |
| 92% RTP | $80 |
| RTP | House edge | Wagered | Paid back on average | Expected loss | Extra cost vs 96% |
|---|---|---|---|---|---|
| 96% | 4% | $1,000 | $960 | $0 | |
| 94% | 6% | $1,000 | $940 | $20 | |
| 92% | 8% | $1,000 | $920 | $40 |
1,000 spins at $1. These are expected values: the average over a very large number of identical sessions, not a forecast for yours.
Source: Our study S6 (validated: 1,000 x $1 at 92% loses exactly twice what 96% loses)
Over what horizon?
RTP is a lifetime average, and the regulators say so. The UK Gambling Commission’s guide to gaming machines puts it plainly: the %RTP “is an average achieved over a significant number of game plays and not each time the gaming machine is played”. It gives the example that a machine showing 85% does not mean you should expect to win back 85 pence for every £1 you stake during a playing session. For physical machines that vary their odds based on earlier results (so-called compensated machines), the same page says the average is generally measured over 10,000 or 100,000 games, and over more games than that for random machines.
We put a number on “significant” further down: even a very calm game needs thousands of spins before its observed return settles anywhere near its RTP.
How to find slot RTP
On most online slots the figure sits inside the game rather than on the casino’s lobby page. The usual route:
- Open the game’s help or information screen. It is usually behind an “i”, a question mark, or a menu icon with three lines.
- Look for the words “RTP”, “return to player”, “theoretical return” or “expected return”. It is often near the end of the rules, after the paytable and feature descriptions.
- Check how the figure is described. Some games state a minimum or a maximum, or give separate figures for the base game, a bought feature or a jackpot.
- If you see several figures, read which one applies to the version you are playing. Some games list a range or more than one RTP for different settings (more on that below).
What the rules require
In Great Britain, the Gambling Commission’s Remote Gambling and Software Technical Standards (RTS) set this out. RTS 3C says that “information that may reasonably be expected to enable the customer to make an informed decision about his or her chances of winning must be easily available before the customer commits to gamble”. The listed information includes the house edge, “the return to player (RTP) percentage” or “the probability (likelihood) of winning events occurring”. The implementation guidance names “the average theoretical return to player percentage” as a way to meet it.
The figure is also checked. In its answer to a Freedom of Information request dated 11 February 2026, the Commission said that testing against its technical standards includes “verifying that the advertised RTP is correct”. For machines in physical venues, its consumer guide says machines “must make information available about their category, %RTP and whether they are compensated or random”.
GLI-19, which many regulators use as a base for their own rules, does not force a display everywhere. Section 4.7.2 leaves it to the regulator, but says that if the RTP is displayed, “the artwork shall fully explain how the displayed RTP was determined (i.e., minimum, maximum, average, etc.)”. It also says that for bonus features you pay extra to enter, the displayed RTP must assume that extra wager was placed unless the game says otherwise.
The help screen shows the theoretical RTP
The figure in the game rules comes from the pay table maths, not from recent results. GLI-19 requires operators to keep both numbers and to “periodically compare the theoretical and actual RTP percentage” so that large gaps get investigated. A game’s recent payouts never change its theoretical RTP.
Why one game can list more than one RTP
A slot’s RTP is a property of its pay table and reel weighting, and a studio can build more than one version of the same game. We found direct evidence of this on a studio’s own website. Hacksaw Gaming’s page for its game Wanted Dead or a Wild lists four RTP values in its game data: 96.38%, 94.55%, 92.33% and 88.42%. The page does not say which operators use which version.
The difference is not cosmetic. Applying the same arithmetic as above to each listed figure:
Source: RTP values: Hacksaw Gaming game page (fetched 9 October 2026); arithmetic ours
See the numbers as a table
| Average loss on $1,000 wagered | |
|---|---|
| 96.38% version | $36.20 |
| 94.55% version | $54.50 |
| 92.33% version | $76.70 |
| 88.42% version | $115.80 |
Two rules from GLI-19 frame how this works in regulated markets. Section 4.7.1 sets a floor: each house-banked game “shall theoretically payout a minimum of seventy-five percent (75%)” over its expected lifetime “unless otherwise specified by the regulatory body”. And appendix A.6.1 requires operators to keep records of “dates and type of changes made affecting the game’s theoretical RTP percentage”, with each such change treating the game “as new for all reports and records”.
So the practical habit is simple: read the RTP inside the game you are actually playing, not a figure quoted on a review site for the same title. Whether operators may switch versions, and what would count as rigging, is covered in our page on whether slots are fixed.
What is a good RTP for slots?
There is no official “good” number, and we could not find a regulator that publishes a market-wide average for online slots. What we can say from sources we checked:
- Wizard of Odds puts slot house edges at 2% to 15%, a range it says is “based on available returns from a major manufacturer”. That is an RTP of 85% to 98%. It is one manufacturer’s spread, not a survey of every game.
- GLI-19’s default floor is 75%, unless the regulator sets its own minimum. A legal game can sit far below the figures you see quoted most often.
- Every RTP below 100% loses money in the long run. A higher RTP is a cheaper price, not a way to win.
The useful comparison is the price per dollar wagered. A 96% slot costs $4 per $100 staked on average. That is cheaper than many slots and far more expensive than a well-played blackjack hand or a single-zero roulette bet. Slots are also fast, so the turnover per hour, and with it the hourly cost, climbs quickly. Our ranking of every common bet by house edge and cost per hour puts slots next to table games on the same scale.
A 98% slot will give me back 98% of my deposit
RTP applies to the money wagered, not the money deposited. If a $100 deposit is spun through ten times, $1,000 has been wagered and the expected loss at 98% is $20, not $2. Re-staking wins is how most of a deposit gets turned over.
RTP, hit frequency and volatility are different numbers
Players often read RTP as “how often it pays”. It is not. Three separate properties describe a slot:
- RTP: how much of the money wagered comes back in total.
- Hit frequency: how often a spin returns anything at all.
- Volatility: how unevenly the return is spread, from frequent small amounts to rare large ones.
To show that the first number does not fix the other two, our study S6 built three made-up pay tables with exactly the same 96.0% RTP. They are illustrations, not real games. The low-volatility model returns something on 47.4% of spins. The high-volatility model returns something on only 19.6% of spins, and saves its value for prizes of up to 5,000 times the bet.
Source: Our study S6, models[].hitFrequency, partialReturnFrequency, breakEvenFrequency, netWinFrequency
See the numbers as a table
| Nothing back | Half the bet back (0.5x) | Stake back (1x) | Net win (2x or more) | Total | |
|---|---|---|---|---|---|
| Low volatility | 52.6% | 15.0% | 12.0% | 20.4% | 100.0% |
| Medium volatility | 71.3% | 10.0% | 8.0% | 10.7% | 100.0% |
| High volatility | 80.4% | 8.0% | 5.0% | 6.6% | 100.0% |
The chart also shows why a “hit” flatters a game. In the low-volatility model, 15% of spins return half the bet, and another 12% return exactly the stake. Both light up as wins, and neither makes money. Only 20.4% of its spins pay more than the bet. In the high-volatility model that share is 6.6%.
| Model | RTP | Spins that return anything | Spins that return more than the bet | Spins that return 0.5x | Top prize (x bet) |
|---|---|---|---|---|---|
| Low volatility | 96.0% | 20.4% | 15.0% | 100 | |
| Medium volatility | 96.0% | 10.7% | 10.0% | 1,000 | |
| High volatility | 96.0% | 6.6% | 8.0% | 5,000 |
Made-up pay tables with integer weights out of 1,000,000; RTP checked in integer arithmetic. Not the maths of any real slot.
Source: Our study S6
How that spread plays out over a session, with bust rates and dry spells for each model, is the subject of our study of volatility with the same 96% RTP.
How long before RTP shows up
Because RTP is an average over the game’s lifetime, a few hundred or a few thousand spins can land far from it. Using the spread of each pay table, we computed how wide the range of observed returns is after a given number of spins. For our calmest model, the low-volatility one, the 95% range after 1,000 spins runs from about 80.5% to 111.5%. Only by around 1,000,000 spins does it narrow to within half a percentage point of 96%.
Source: Our study S6, models[low].rtpConvergence.ci95HalfWidthPctPoints
See the numbers as a table
| 95% range, plus or minus percentage points | |
|---|---|
| 1,000 spins | 15.54 pts |
| 10,000 spins | 4.91 pts |
| 100,000 spins | 1.55 pts |
| 1,000,000 spins | 0.49 pts |
For riskier games the range is much wider. In our illustrative high-volatility model, getting the observed return within ±5 points of 96% at 95% confidence takes about 1,926,230 spins, against 9,660 for the low-volatility model. At ten spins a minute, that is more than 3,000 hours of play.
Common mistakes with RTP
- Reading RTP as a session forecast. After 1,000 spins, even our calmest 96% model lands anywhere from about 80.5% to 111.5% in 95% of sessions. The figure only describes the long run.
- Comparing RTP with your deposit. The house edge applies to every dollar wagered, including re-staked wins.
- Trusting a figure quoted off-site. One title can exist in several RTP versions. Read the number inside the game you are playing.
- Treating a higher hit rate as a better game. Hit frequency counts 0.5x returns as hits. Two games with the same RTP can pay on 47.4% or 19.6% of spins.
- Expecting a game to “catch up” to its RTP. Spins on a certified game are independent. A cold run does not make the next spin more generous. Our page on rigging and RNG testing explains why.
If you play video poker as well, its pay table is printed in full and its return is easier to check. The video poker guide compares the two.
What we could not verify, and our assumptions
- Which RTP version a given site uses. We confirmed that one studio publishes four RTP values for one game. We did not check which version any operator runs, and we name no operators.
- Other studios’ practice. We could only verify multiple RTP versions from one studio’s own page. We found secondary articles claiming the practice is widespread, but did not rely on them. Another studio’s game page we fetched listed no RTP at all.
- A market-wide average RTP. We found no regulator source for the average RTP of online slots, so we give none. The 2% to 15% house edge range is Wizard of Odds’ figure from one unnamed manufacturer.
- Display rules outside Great Britain. We quote the UK RTS and GLI-19. Other regulators set their own rules, and GLI-19 leaves RTP display to each regulator.
- Our pay tables are invented. The three S6 models are generic illustrations with an RTP of exactly 96.0%. Real slots have many more outcomes, bonus features and often larger top prizes, so their spreads differ.
- Normal approximation. The convergence ranges use a normal approximation, which understates how lopsided results are for games driven by rare large prizes.