Artificial intelligence can make a big difference in supporting casino slot and marketing decisions and increase profits, according to Steve Bright, vice president of data science for OPTX.
Las Vegas-based OPTX produces a data platform that focuses on ingesting data from a casino’s disparate systems and turning that into operational functionality for the marketing, player development, and the slot floor.
Bright said they rolled out their latest technology at this past G2E and it’s going out to existing casino customers at this time to enhance their operations and decision making.
“A slot director might be faced with several hundred potential changes (in game location and content) every year,” Bright said. “What we have found at OPTX, with data that goes back several years with clients, is that slot directors or properties get it right only about 50% of the time. That means any sort of slot-floor change has been a coin flip. It might make you money or cost you money.”
OPTX approaches the problem by trying to understand what drives player behavior and representing those factors mathematically. “You put the mathematical building blocks together in various ways to assemble predictions about any proposed slot floor changes,” Bright said.
One important factor is capacity constraint. By observing when a slot machine is in use, OPTX can estimate how much game play at a given machine is lost by that slot being busy.
“Capacity constraint is an important, if not the most important, factor. We have session-level gameplay data, so we can tell pretty reliably when any given machine is in use,” Bright said. “We can estimate how much revenue is being lost just by this machine not being available.”
Another important factor is the ability to substitute one game for another. If a player sees that a machine with a favorite game is busy, what might they play instead? OPTX has a lot of observational data and detailed information about how any two games are related with regards to how a player considers the games to be valid substitutes.
“It can have similar characteristics, like game-play style and artwork,” Bright said. “We find games with more similar characteristics tend to be more substitutable. It could be animal-themed, mythology-theme, music-themed or pop culture-themed games.”
Other factors include volatility of the base game, along with bonus features and how frequently they hit.
When all the mathematical building blocks are constructed and put together, OPTX comes up with a model that predicts what will happen after a proposed slot-floor change. They can also forecast how one machine will affect other machines, such as those around it and have a similar theme.
“We don’t restrict our analysis and prediction to what we think will happen when a machine is changed. We also want the model to say what will happen at the other machines on the floor,” Bright said. “We want to get an entire floor-wide view of how we think this change is going to impact you.”
Bright said their model doesn’t get it right 100% of the time for various reasons, but it does give the right answer between 60% and 65% of the time.
“There’s a lot of unpredictability in what can happen with regards to a specific machine over a several-week period. But we’re finding that this model takes what was a 50-50 proposition and makes it a 60-40 proposition in favor of the operator,” Bright said. “We’re not aiming for perfection. Rather, we load the dice a little more in favor of the operator and help them make statistically better decisions in the long run. That’s the whole business model of the casino.”
Bright noted that casinos make money by cashing in on small house edges millions of times and OPTX does something similar with the slot floor: Hundreds and thousands of individual floor changes improve the customer experience and make it more profitable for the company.
OPTX doesn’t merely want to evaluate proposed slot changes that originate from the user. One priority high on the company’s road map is to proactively recommend changes to the operator. It’s the same mathematical model, but it originates from the OPTX end, Bright said.
When it comes to casino marketing directors, the objective is different. Marketing is concerned with getting the right offer to the right person at the right time based on how much that player has spent. By evaluating the entire database, OPTX can generate a machine-learning model that makes a prediction about every player. The model observes individual visits and their frequency, recency, and intensity, along with other characteristics of that player’s gambling.
“We can create a forward-looking forecast of that player’s slot or table win over a future time horizon. We find that the forward-looking value from our model is a better predictor of what happens than just relying on whatever that player spent in the last 28 days,” Bright said. “We have so much data that whenever we make a machine-learning model, we can always run the clock backwards and do this pro-forma what-if analysis.”
By pretending they’re looking at data collected up until a month previously, OPTX can make a prediction, then run the clock forward a month to include the most recent data and see how right or wrong the prediction was. The model turns out to be 30% more accurate than relying on the previous time period.
“We’re building this model that can predict, with reasonable accuracy, how much revenue every player will generate,” Bright said. “If you’re a marketing director putting together a campaign or building an offer, you can create different segments, based not just on whatever the player has done in the past, but on what you think will happen in the future. In a sense, you’re letting the operator skate to where the puck is going to be. It’s a helpful and powerful insight to develop.”
OPTX is always looking to develop more features of marketing artificial intelligence. Bright said that anywhere there’s a report, there’s an AI feature or prediction to be made and OPTX is working hard to push AI everywhere it makes sense to go.