Inspiration
The general manager of a fictional Las Vegas casino planned to replace the ten machines with the lowest house win from January through June 2025. We wanted to challenge that decision. Were these machines actually underperforming, or were other factors like luck, location, and downtime responsible?
What it does
The Gilded Ace is an interactive casino analytics dashboard that helps management make data-driven decisions. It evaluates machine performance while accounting for exposure, availability, and luck. It also explores floor demand, downtime, player outcomes, and loyalty program costs before recommending how to allocate a $100,000 budget.
Our analysis found that most underperforming machines were concentrated in Back Forty, where occupancy was just 25%, compared with 45% in Center Pit. We also identified approximately $94,000 in potential lost theoretical win from downtime.
How we built it
We used Python, DuckDB, Pandas, and PyArrow to process the parquet datasets, build analytical queries, and generate reproducible results. We exported the results as JSON and integrated them into a dashboard built with HTML, CSS, and JavaScript.
For forecasting, we developed a weekday-mean model trained on January–April and tested it on May–June. Its MAE was approximately $68,699, compared with $218,258 for a constant-mean baseline.
We used Cursor as our development environment and Grok as an AI coding assistant. Grok helped us develop and refine the Python and DuckDB pipeline, adapt the provided example dashboard, connect analytical results through JSON, and draft our write-up and Claims & Evidence appendix. We used Cursor to navigate the repository, edit and debug code, and integrate the different components.
Challenges we ran into
One of our biggest challenges was separating genuine machine underperformance from luck, limited exposure, and downtime. We also had to account for machines changing locations and ensure that our joins and aggregations did not double-count financial data.
Accomplishments that we're proud of
We turned our analysis into three actionable recommendations: a Saturday promotion ($15,000), six machine moves ($10,800), and an additional grave-shift technician ($11,520).
Rather than spending the entire budget or replacing machines immediately, we recommend spending $37,320 and keeping the remaining $62,680.
What we learned
We learned that low historical earnings do not necessarily mean a machine is performing poorly. Understanding the context behind the numbers is just as important as the numbers themselves. We also learned the importance of time-based validation, transparent assumptions, and measuring whether recommendations actually work.
What's next for The Gilded Ace
Our next step would be to track the results of our proposed actions over the July period. We would compare machine occupancy, repair times, and promotion performance against our established baselines, using the results to refine future recommendations.
Built With
- css
- cursor
- duckdb
- grok
- html
- javascript
- pandas
- pyarrow
- python