This is a great question, and it also reveals a fundamental tension in how we think about data analysis. You have the right instinct: you want to find the signal in the noise, to know whether it's the opponent, the weather, or a national broadcast that really moves ticket prices. But you are asking Excel to do something it was never designed to do well. Building a decision tree from scratch in a spreadsheet is possible in the same way that building a car engine with a hammer and screwdriver is possible. You can do it, but you will spend more time fighting the tools than solving the problem.
The Analysis ToolPak add-in is a useful relic, but it was built for linear regression and t-tests, not for the recursive partitioning that decision trees require. A decision tree works by systematically testing every possible split across all your variables, day of week, opponent, weather, TV broadcast, and then choosing the split that best separates the data by average ticket price. That process is computationally intensive and iterative. Excel can simulate it with nested formulas, pivot tables, and manual logic, but the moment your data has more than a few dozen combinations, the spreadsheet becomes brittle, slow, and error-prone. You will be debugging cell references instead of interpreting the results.
The deeper point here is that your goal is not to build a tree inside Excel. Your goal is to understand which variables matter most and at what thresholds. That is a predictive modeling problem, and there are tools that solve it directly, without the overhead. Modern AI-native platforms can ingest your four columns and your aggregated counts, then output a clean decision tree in seconds, complete with split thresholds, node purity, and variable importance rankings. They do not require you to write a single formula. They let you stay focused on the question: is it the Chiefs game that drives prices up, or is it the combination of a Sunday night broadcast and good weather?
So here is our plain opinion: stop trying to make Excel do machine learning. Use it for what it is good at, data storage, quick summaries, and ad hoc filtering. Then export your four variables and your aggregated ticket data into a purpose-built tool that handles the heavy lifting. The answer you are looking for is not a spreadsheet trick. It is a better workflow. The most significant variable is not opponent or weather. It is the tool you choose to find the answer. Choose one that lets you predict smarter, not work harder.