Built to Scale: How MSU Athletics Kept Refining an AI-Powered Stats System Until It Worked for Every Sport

Mississippi State Athletics Communications didn't set out to build an automated statistics system. It set out to fix one page. Then another. Then another — using advanced Excel skills, a deep working knowledge of how the NCAA reports game data, and a determination to close a real efficiency gap that was eating up hours every game night. A year and a half later, that steady problem-solving has grown, sport by sport, into a system that turns a full slate of player statistics around in minutes instead of hours, and is now being built out to run across every sport in the department.

"It's kind of been a little play as we go and experiment as we go sort of thing over the last year and a half," said Brian Ogden, director of communications for MSU Athletics. "We're not coders, and that's the easiest way we can make it work."

Ogden started small, using ChatGPT to help write Excel and Power Query formulas that could pull game data off the NCAA website and clean it up automatically. It worked, but it also hit a wall. Refreshing every player's stats still took the better part of an hour, and requesting too much data too quickly risked getting their account locked out entirely.

Rather than settle for a partial fix, the team went looking for a better one. They connected with Josh King, a doctoral candidate who runs the athletics department's new AI & Data Science Lab, and asked a simple question: if we already have the underlying data file, can you help us route it straight into our spreadsheet instead? King and his students jumped in and built exactly that, eliminating the need to pull anything from the website at all. With that bottleneck solved, the improvements kept stacking: Will Simmons, another athletics staffer, realized they didn't need two separate data refreshes, one for game stats and one for career totals, since the career numbers could simply be calculated from the game log. So, they cut it down to one.

That kind of steady refinement — noticing a redundant step, a manual workaround, a formula that almost worked, and going back to fix it — is what took the project from a rough first draft to something dependable enough to run in real time during games. Every version got tested against sample data before it was trusted with the final product.

Now that the core system is proven, scaling it to new sports is turning out to be the easy part. Because every Mississippi State team's data flows through the same underlying files, a formula built for one sport can largely be adapted for the next. The project began with stats for the soccer team, was tweaked and applied to basketball, with continual improvements to those sports and a new application to softball on the horizon. The improvements rely on an innate knowledge of the quirks of each sport's scoring, like getting Excel to understand that in baseball, "0.1" innings means a third of an inning, not a tenth, or that in golf, the lowest score wins instead of the highest.

"Because all of these sports are run off those same XML files, we have them," Ogden said. "It's very easy to, once we get it set up, make it work for everything."

Ogden is quick to credit the collaboration for getting the system this far. "I want to make sure they get the credit for what they did there," he said of Simmons and King. "Those two really helped elevate it and take it to the next level." With the groundwork now in place across sports, the team is turning its attention to teaching the next generation of student workers to build the same kind of fixes themselves.