What is worldcuproi?

WorldCupROI is a Python data platform that tries to answer a business question rather than a sports question: given all the noise around a World Cup tournament, which sponsors are getting good value for their money, and how might that change under different scenarios. It combines match performance data, text signals pulled from real media coverage, sponsor spending, and fan attention into one system, then adds a layer that measures how confident its own predictions are. The project has two main parts. Underneath is a set of machine learning models: one predicts match outcomes, and another predicts sponsor return on investment, meaning how much commercial value a sponsor is likely getting relative to what they spent. Both come with uncertainty estimates, using a technique called conformal prediction, so the platform reports not just a single number but a range and how often that range should be expected to hold true. On top of the models sits an interactive dashboard, built with Streamlit and Plotly, that lets someone explore the numbers visually: ranking sponsors by predicted ROI, viewing which factors are driving that prediction, and running what-if scenarios such as a key player being unavailable, a change in sponsor spending, or cooler media coverage, to see how predicted ROI shifts in response. The README presents a series of results tables directly, including match prediction accuracy, ROI prediction error, and how reliable the uncertainty ranges are, along with a ranking of sponsors like Hyundai, Adidas, Coca-Cola, and Visa by a calculated influence score based on how connected they are within the network of teams, players, and matches. It also shows example scenario comparisons, such as predicted ROI dropping several percent if a core player is absent or if media attention cools off. Running the interactive version requires Streamlit, and the project also ships a static HTML preview and a demo video for people who just want to see the dashboard without installing anything. The codebase is Python 3.11 based and optionally supports additional modeling libraries like XGBoost, LightGBM, and CatBoost for the tabular prediction tasks. The README does not state a license.