Earlier this year The Drop published results of a research paper by a professor at UNLV who claimed players couldn’t detect a change to the machine hold percentage. There were some limitations in the research, for example the machines that were chosen were not the strongest on performance and this may have impacted play and certainly were of a game type that would have been chosen by players who were inexperienced and low-value gamers. Machines were positioned together but performance would have been affected by the position in the overall bank configuration. However, new research using AI appears to confirm that this is the case and players are unable to detect changes. Below is an article by CDC Gaming reporting on the new research
Artificial intelligence has weighed in on the ongoing debate over whether regular slot players can tell the difference in house edge from one game to another.
The verdict so far? They can’t.
According to Stasi Baran, co-founder and Chief Operating Officer of nQube Data Science, current AI shows that players don’t recognize differences in house edge. Baran presented the results of an ongoing AI case study on slot floor optimization at the Elwha River Casino in Washington state during a talk on AI and Guest Engagement at the TribalNet Conference in Nashville Tuesday.
The controversy gained new life this year after Anthony Lucas, a professor at UNLV’s William F. Harrah College of Hospitality, released a series of studies that challenged the belief of casino operators and consultants about a player’s ability to detect differences in how much and how often a slot machine pays. Lucas said some casinos were even using his study to increase the house edge and boost their revenue.
“We have seen nothing in the AI that suggests (players) can determine hold or that there’s been adjustments on the hold,” Baran said. “Hold affects the performance of the machine, but in terms of player response, I say no.”
Baran said Lucas “does amazing work” but that he has not been allowed to survey entire slot floors when conducting his studies.
“There’s more work to be done, and he’s leading it,” Baran said. “I know there are two different schools of thought on it trying to come to an agreement.”
Baran said she understands why slot directors have a different view. Some regular slot players note the outcome of every spin while they play.
“If you have data, you can perhaps detect a hold,” Baran said.
Baran, who previously worked with artificial intelligence to detect breast cancer, said casino operators have become more interested in AI over the past year and are now starting to embrace it.
It’s best to look at data over a three-month or six-month period to gain insights on the slot floor, she said. Data measured over that type of timeframe can tell a casino operator whether to remove machines from the floor, which slots to buy and where to put them. It can also provide theoretical win percentages.
“We’re finding the slot mix is more important (than) location,” Baran said. “When players enter the floor, you want them to find exactly what they want as quickly as possible. You don’t want them wandering around the floor. By segmenting the machines and matching them, we (can) ensure that happens.”
Data have also shown that a floor can be more profitable after removing several machines, since, if a popular medium-performing machine is taken away, there are people who will go play a higher performing one.
“Operators are shocked to hear that,” she said. “Win-per-unit might go down on the floor, but that’s all right if the entire floor is making more money.”
Baran said some of the other takeaways are that the lowest performing slots aren’t always a priority for retirement, because the worst they’re doing is taking up real estate, and that as more machines of a particular type are added, the performance of the individual machines tends to decrease due to market saturation or cannibalization.
“If you add more product on the floor that’s similar, you’re going to dilute interest,” Baran said. “It’s hard for humans to balance that supply and demand. AI gives the optimal layout for the slot floor. The slot floor is an ecosystem of players and machines and complex interactions happening at all levels. Every change on the slot floor affects every other part of the floor. If the performance of a particular machine goes up, the question is what is going to happen to the rest of the floor.”
Elwha River Casino, a small locals casino in Port Angeles, Washington, has 138 slot machines. The casino’s study showed that by changing 8.8 percent of their floor, they were able to find a 3.8 percent potential increase in theoretical win. By changing 18.3 percent of the floor, it could increase theoretical win by 5.2 percent.
The bigger the casino, the better AI can work, Baran said.
“As floors get larger, there are more inefficiencies because it’s harder for humans to track,” she said. “AI (can do) millions of slot floor possibilities.”