A recent study of the behaviour of slot players has caused some controversy. Anthony F Lucas, Ph.D., and Katherine A. Spilde, Ph.D, furthered the research into whether gamblers would be able to tell what was the relative Return to Player (RTP) of a particular slot machine game and whether they would migrate to a machine with a higher RTP.
Why is this important? Well, some casinos have seen a significant decline in slot revenues, and it has been argued that decreasing RTP is a major factor because a lower RTP will lead to a shorter time on device; players want more time for their money. The received wisdom is that low RTPs will lead customers to stop visiting their local casino and could, in extreme circumstances, damage the casino company’s brand.
The initial research carried out by Lucas and R.K. Singh in 2011 investigated whether, on a single trip basis of varying lengths (measured by number of bets), a player would be able to detect a difference in RTP. Their methodology used pairings of the same slot machine game but with different RTPs and different levels of volatility. The results of the outcome of 10,000 virtual players was recorded and analysed to see if the different RTPs would result in notable differences in actual (as opposed to theoretical) winning spins. Volatility is measured using a standard deviation; the higher the standard deviation the more volatile the game. In other words, the short and medium term revenues of high volatile games will more often deviate significantly from the expected revenues.
All slot machine games have an expected theoretical RTP; the game structure determines it for each game. But some games are so volatile that it could require ten million spins or more for the actual RTP to get close to the theoretical RTP. At one spin every seven seconds or so, ten million spins equates to 20,000 hours or 830 days of continuous play.
The RTP in the initial study was varied between 88% and 97%. The results showed that on a single trip basis, players were unlikely to see much difference in winning spins between machines with comparatively low and high RTP. Not surprisingly the characteristic they would likely be able to detect would be the difference in volatility between the paired machines.
Two of the criticisms of the initial research were that it only investigated the performance of game simulations and not slot games in live casinos, and, also, that it was based on a single trip. The most recent piece of research, published in July of this year, looked at actual machine performance in live high frequency (that is, locals) casinos in the US, Australia, and Mexico.
The researchers’ methodology was to compare the performance of two identical slot machines, positioned in similar places on a casino floor, but with different RTPs. Over a nine-month period, they compared the actual win per device per day of pairs of machines with the games “Tokyo Rose” and “Dragon’s Fortune X”. The RTPs of the pairs varied between 92% and 85%, or a PAR (theoretical house win percentage) of between 8% and 15%, a significant difference (almost double). That’s enough, you would have thought, to persuade players to migrate to the higher RTP machines.
Contrary to expectations, although not to the researchers, the revenues for the low RTP games were substantially higher than the paired games with the higher RTP. It would appear that players did not migrate to the games with the higher RTPs, although the study only looked at machine data and not player data, so it is only safe to assume that the earnings of the machines was improved by reducing the RTP.
If slot directors read this and think it means that they can reduce the RTP on their games and all will be good and customers will be happy, they are not looking at the full picture.
Let me take a step back. The casino business is the entertainment business; we package time and sell it to our customers. Win or lose, as long as the player sees value in it, they will be willing to return. The key here is value. What is value? Well, some people see value in a McDonalds Happy Meal. Others see value, regardless of cost, in a meal at a Michelin-starred restaurant with thick white tablecloths and plenty of waiters hovering around. A big part of value is expectation – what do I expect to get in return for my money or my time?
Let’s use an absurd example to negate the researchers’ argument. Suppose I put a slot machine on the floor of a locals’ casino that had an RTP of just 2%. I will guarantee you that quickly the daily revenues of that machine will be extremely low, and quite possibly nothing at all. So clearly players are sensitive to RTP.
Another example, this one in support of their hypothesis. In Europe, at a single zero roulette game, the even-chance bets (red/black, odd/even, and high/low) have an RTP of 98.65%, while all other bets have an RTP of 97.3%. Compared to slot machine games, roulette is not very volatile. If players were solely concerned with RTP, the only bets they would make would be the even chance bets. Clearly that is not the case; in fact, comparatively few even chance bets are placed. Thus a higher RTP does not appear to drive betting volumes. It’s true that there are some customers, not many, who like nothing better than to sit at a roulette table for a few hours betting the minimum amount on the even chance bets. There are others who would rather poke a stick in their eye before playing that way.
So what is going on here? Are players sensitive to RTP or are they not? There must be a point somewhere at which low RTP puts players off. Is it the same for all games?
It is difficult to (eventually) win many times your stake by playing the even chance bets on roulette, and playing this way to most people is extremely boring. Blackjack is also mostly a 50:50 proposition game, but differs in that it allows the player to decide when to draw a card and when to double (buy a card), which adds an element of fun or entertainment. Blackjack, in most countries, Asia excluded, is the most popular of the table games. It would appear that the entertainment component is an important deciding factor for some players.
If you watch casino players for a length of time, you will notice that they will get up from a table or slot machine when they lose their money too quickly. This would suggest that the rate at which a player loses money, or the velocity of loss, is a consideration for a player in deciding if the experience is offering value. In fact, Harrah’s developed its player tracking system to analyse the live player data and determine the maximum rate of loss that each player would accept before moving to another machine. The company’s analysts believed the data showed that at the point when a player decides to change machines for this reason there is a risk they will leave the casino altogether. They proposed a system to identify any player that was approaching the rate of loss at which they would be likely to move and then automatically bonus their playing account. Unfortunately, the regulators would not allow this, on the basis that they were not treating all players equally. So instead the system flags the player and the slot host physically intervenes, possibly offering a comp of some kind.
I believe that there are many factors that interact to determine whether a player will play a particular game and how long they will continue to do so. To say that RTP is the only factor is simplistic at best. In my view, it is the relationship between several factors – stake to prize ratio, RTP, frequency of bet/outcome, and entertainment value – that determines the long-term revenue earning potential of a game. Games with low RTP and high frequency of outcome will not remain popular unless the stake to prize ratio is high and/or the entertainment factor, which could include volatility, is also high.
Lottery directors know it is hard to sustain lottery revenues over the long term with a 50% RTP game but with a relatively low frequency of only one drawer per week. Revenues are only sustained in the long term with ever-higher prizes, more games, etc., because higher frequency will reduce the life of the lottery unless the other factors are altered. Scratch cards suffer a similar fate – compare the size of the prizes available today to what was on offer thirty or even twenty years ago. Lotteries have very little entertainment value, because players do not participate in a game, but prizes are large relative to the stake and the frequency of outcome is low. Imagine if every five minutes there was a state lottery draw with an RTP of 50% or less. Within a few weeks the lottery would lose its ability to generate sales.
AWPs in the UK have an average RTP of 75%. The maximum prize is 250 times the maximum stake, and the machines employ reflexive software to smooth out the volatility. These games have a high entertainment component, employing features such as nudges, skill runs, etc., so much so that I find them confusing. Consequently, the combination of low RTP, low stake to maximum prize ratio, and high frequency of outcome means that many games’ popularity (high daily revenue) lasts between 6 and 9 months. Operators will place a machine with a new game in a venue until its popularity (revenues) wanes and then move the machine to a secondary location and then to a third, a fourth and so on. If you offer these games, you are also in the furniture-moving business.
Slot floors do not work in the abstract. The characteristics of each machine impacts the revenues of the machines around it. One slot floor is not necessarily like another. There is not one type of player; preferences and expectations of value differ. Some players like low volatility and a long time on device; others value volatility at the expense of RTP. I do not believe that the study of one machine characteristic will give any meaningful answers.
I think Lucas and Spilde’s research is an important first step in understanding the impact of changing RTP within certain bounds, but these problems are too difficult to be solved with simple linear equations (holding all but one characteristic static and seeing what happens to the revenue of that machine when you vary that characteristic). To get meaningful answers, large data sets need to be employed and the impact of any change(s) measured across the entire slot floor. It is only through this type of analysis that can you safely say that reducing RTP, within certain bounds, has no impact on the revenue earning potential of the entire floor. After all, it is the total revenue of the floor that should be of interest to the slot director.
