Feature article
How can gaming machines meet their %RTP if they are random?
Introduction
Regulation under section 240 of the Gambling Act requires that all nonremote gaming machines display information about:
a) the proportion of amounts paid to use the machine that is returned by way of prizes; or
b) the odds of winning prizes from use of the machine.
This requirement is typically met by including a statement of %RTP (return to player percentage) on the face of the gaming machine or within a software help page. There is no statutory minimum %RTP figure, and for gaming machines which include multiple games, each separate game must state (and meet) its own individual %RTP figure.
Gaming machines achieve these percentage return figures in one of two ways:
 either by using a form of feedback compensation within the game software (compensated games)
 or by utilising software which is fully random in operation (fully random games).
Note that compensated games are also random, but they have an additional control element within them to provide the compensation.
It is important to state that “random” usually means “pseudo random” with regard to the method used to generate gaming machine outcomes. A random number generator (RNG) is used to generate outcomes for both the random and compensated types of game.
The number of game cycles required to achieve the stated %RTP differs for compensated and fully random games. Compensated games are designed to achieve their target %RTP in typically tens or hundreds of thousands of game cycles, whereas games which are fully random take considerably longer to achieve their %RTP figure, typically in the order of a million or more game cycles.
Compensated game control
This is a complex topic but in simple terms:
The game software constantly checks to see how much has been awarded in winnings in the recent past; and, knowing what the current %RTP figure is and what the target %RTP figure should be, the software will either slightly vary the odds of achieving a win, or slightly vary the value of any win, for the subsequent series of games. Thus, the game will pay out more or less often, or pay out greater or lesser amounts, than it might otherwise do; depending upon whether the actual %RTP is currently below or above the target %RTP of the game.
The reason for using compensation at all is twofold: 
 Firstly, to give a less volatile game experience to the player. This gives a somewhat more even spread of wins and losses over time than might be given by a fully random game. It also results in a more evenly spread cashbox take, which operators generally prefer.
 Secondly, in technical terms, the difficulty in attempting to model the gameplay for very complex ‘hightech’ style games (found in, for example, pubs) has meant that games of that style would potentially struggle to meet their %RTP figures due to the multiple play options and game strategies a player can adopt. Compensation is often the only way to make those types of game work with regards to %RTP targets.
Random game control
Random games do not control the %RTP in the same way as compensated games but rely purely on the statistical chance of a random event offering a win. The way in which fully random games control to a specific %RTP can seem confusing, the following calculations may help to clarify:
This simplified example shows the processes involved.
A reel based “fruit machine” game uses 3 reels, each with 20 positions on them. Each reel consists of 2 sets of each of the following symbols, Cherry, Lemon, Orange, Melon, Bell, Star, Bar, which between them take up 14 reel positions, the remaining 6 positions on each reel are filled with XXXs.
Assume the stake is 50p for one game and the game awards a jackpot of £250 and has only one winline. The %RTP is required to be 90% and the game is fully random – that is, no compensation at all.
Thus, 3 reels with 20 positions on each means there are 20 x 20 x 20 possible display combinations, 8000 in total, a few will equate to wins and many more to losses.
Statistically speaking, and following the law of large numbers, in 8000 games it would be highly unlikely that all possible reel combinations would be displayed on the winline, but if a very much larger number of games were played then it would be much more likely that all possible combinations would be shown on the winline an equal number of times. The probable number of game cycles needed to achieve this averaging effect can be determined statistically and it could be in the order of millions, this is the reason we need to allow random games a longer time to achieve their %RTP figures than compensated games.
Returning to the example, the following is a simplification based on an average series of 8000 games where it is assumed for calculation purposes that all possible outcomes are shown once during the series.
Let us say that the possible winning combinations are made up as follows:
Reel 1

Reel 2

Reel 3

Win Value (£)

No. of chances

Total value (£)

Bar

Bar

Bar

250

2 x 2 x 2 = 8

250 x 8 = 2000

Star

Star

Star

75

8

600

Bell

Bell

Bell

45

8

360

Melon

Melon

Melon

20

8

160

Orange

Orange

Orange

20

8

160

Lemon

Lemon

Lemon

20

8

160

Cherry

Cherry

Cherry

20

8

160





Total

£3600

To explain further, using the Bar symbol as an example, there are two bar symbols on each reel, so there are two chances that reel 1 can show a bar on the winline, similarly two chances of reel 2 showing a bar, thus the combined chance is 2 x 2 or 4 chances, and if reel 3 is included, then this becomes a total of 8 chances (in the 8000 possible outcomes) of 3 bars on the winline.
The win values have been carefully chosen to provide the required overall total value when multiplied by the appropriate number of chances. If different win values are needed for the game, then to arrive at the correct total value, the number of chances can be finetuned as explained below.
If 3 bars award a win of £250, then during the 8000 games, this outcome will be achieved 8 times awarding a total of £2000 to the player.
For 8000 games at 50p per game, the player stakes a total of 8000 x 50p = £4000. This is the cost of play.
Assuming each possible display combination occurs only once, then the winning combinations as shown above would yield £3600 over the 8000 games.
Thus %RTP = Wins awarded = £3600 = 0.90 = 90%
Cost of play £4000
It must be stressed again that 8000 is too low a number of games to expect a fully random game to achieve its target %RTP. The actual %RTP after this number of games would most likely be considerably above or below the target percentage depending upon the random allocation of wins awarded. It is only after very many games that the averaging effect evens out the volatility of wins and losses and the target %RTP is achieved.
Fine Tuning The %RTP
Assume for example that there were 3 cherries on the third reel instead of two and hence one less XXX symbol. Then the cherries would contribute 2 x 2 x 3 chances = 12 x £20 = £240 and not £160, thus the %RTP calculation would become:
New %RTP = Wins awarded = £3680 = 0.92 = 92%
Cost of play £4000
In this way the game designer can fine tune the %RTP. Other methods used would be to give several smaller wins, or to offer say 19 or 21 positions on one or more reels. All of these affect the calculation by differing amounts.
The whole calculation is of course based on the statistical probability that everything will average out over a sufficiently large number of games, hence the need to allow random games a much longer time to “settle down” than compensated games and hopefully this explanation will show why %RTP must always be quoted as an ‘average’ figure – over a large number of games.
Next chapter: LLEP Assessment templates