Can AI Replace Casino Analysts? A Real-World Test of ChatGPT’s Market Forecasting Skills

Life moves pretty fast, but sometimes we need to slow it down a bit.  You’re doing that right now by reading this blog post.  The world of AI, or the world which AI is creating, is developing at break-neck speed.  The capabilities of AI are growing and increasing so fast that even those actively trying to keep up just can’t.  If we can’t keep up, what do we do? 

Recently someone told me that AI would replace me professionally.  I was told that all analytics would be done by AI, and that people like me would be out of a job.  I laughed it off, as my previous experiences with AI have shown its results to be dreadful, full of errors, and completely unreliable.  But as I am continually discovering, it is learning and getting better. So, how can we best use AI in casino market analysis?

The Challenge: Can AI Replace Human Casino Analysts?

I input the following prompt to ChatGPT: Please build me a gravity model (the tried and true method for regional gaming market analysis) and forecast the gaming revenue potential for a new casino in (location X) with 200 tables and 3,000 slots.  Please be sure to include all competition within a 3 hour drive. 

The Results and Where AI Fell Short

I watched my screen as ChatGPT thought about this and pulled in data from the web and it’s own databases.  I watched as it got initial market research steps right.  Next, it attempted to do the math.  OK – the math wasn’t entirely bad.  Where it fell flat on its face was on the revenue generation potential.  Rough estimates and assumptions were applied, seemingly pulled from public copies of market studies like the ones our company undertakes.  But each study is different, as each market is different.

The AI model assigned a win per unit per day, seemingly based on a very wide range of facilities in the region, and none of which could or should be indicative of this market.  Then the market share derived from the “gravity model” was applied to generate an estimate of the gaming revenues for the facility.  There was no definition of the market, how many people it included, what their income levels were, where they lived in relation to the facilities, what their spending patterns were, or whether they currently visit casinos.  No consideration was given to access to a wider region, none to visitors or tourists. 

Additionally, AI couldn’t take into account what analysts learn from being on the ground – from the site visits that often greatly influence our team’s forecasts and recommendations. While site visits typically take no more than a few days, they are invaluable for the information and insights they provide to market analysis. This can’t be done from a desktop or through AI.

Why Human Expertise Still Matters

The current problem is how to check these models.  In our work, we meticulously document what we do, and we cite every source.  We know from experience that not all sources are created equal – some are reliable, and some are just B.S.  Chat GPT will provide source documentation, but checking them is laborious, and often results in the conclusion that Chat GPT didn’t use the best source, but rather the one that was easiest to find or repeated more often. 

Now, this is fine if you’re just curious about something.  It is not if you are trying to make real-world decisions.  Would you invest your money in a project that Chat GPT says has the potential to give you a return on your investment?  How confident are you in its results?  It takes an experienced professional to detect the mistakes, and to illuminate the possibilities that aren’t already out there on the web or in datacenters. 

The Future: Partnering with AI for Better Insights

That said, AI is helpful.  In another experiment, I’ve input lots of data from a model I created, and am asking for feedback. So far, it has been akin to working with a colleague – someone who asks the right questions, and helps me find the answer on my own.

Perhaps we don’t need to keep up with AI, but simply figure out the many ways in which it can make us and our work product better.  That’s what I’m focusing on, and I’m excited.

• Suzanne Leckert, Co-Founder & Managing Partner

Learn more about Convergence Strategy Group’s evolution with AI: Accept AI for its imperfections and use it to pull ahead, Will AI Re-Invent Research Soon?, Cautiously Curious and Learning More about AI, and Forever Learning, and Apparently Quicker and Better than AI

What are others saying? The Projected Impact of Generative AI on Future Productivity Growth and How will AI Affect the Global Workforce?