Performance and competitor modeling platform for elite cycling

To maintain client confidentiality, the company and industry in this case study have been anonymized. The underlying solution is the same.

The problem

Directors at Stadia Cycling Federation were sitting on a season’s worth of race footage, power-meter telemetry, GPS traces, and event timing data, and could barely use any of it. Post-race analysis happened in spreadsheets, took two weeks to compile, and rarely answered the question directors needed answered: how does my rider stack up against the specific riders they’ll race next week?

Directors were calling race tactics on gut feel and last year’s results, and coaches were prescribing training the same way, with no clear read on which efforts would move a rider’s position when it counted.

What we built

A race analytics platform that captures every rider in the peloton at 20ms resolution: 40 riders per race, ~2.5 hours of competition, 100+ structured metrics per reading. Power output, cadence, heart rate, position within the pack, gradient, wind exposure, GPS-derived velocity, fatigue indicators, head-to-head proximity to specific competitors. That’s roughly 1.8 billion data points per race, ingested live and queryable within minutes of the finish.

A statistical layer cuts the noise and surfaces what’s predictive for that rider on that kind of course, rather than peloton-average correlations that don’t transfer to the individual.

On top of the analysis layer sits a tactical simulator. Directors model a specific upcoming race (riders, course profile, conditions, team tactics), and the system runs the matchups against historical performance distributions. It doesn’t predict the winner. It tells you where the realistic margins are, where a 2% gain in sustained climbing power matters and where it doesn’t, and which riders are statistically beatable on this day on this course.

The platform models the field, not just the rider. Every simulation accounts for how other riders are likely to ride under the same conditions: who closes gaps fast, who fades on the climb in the final hour, who burns matches early. Directors stopped asking “how strong is my rider” and started asking “how strong does my rider need to be to win this race, against this group, on this profile.”

Results

Deployed across Stadia’s full racing roster within the first season. Race analysis that used to take two weeks now ships within 48 hours of a race ending, with 1.5 seasons of cleaned, queryable competitor history distilled into a director-readable brief before the next start line. Directors now use the simulator before every major race to pressure-test tactics against the actual field rather than a generic one.

More importantly: directors stopped relying on instinct and last year’s results to decide which riders to send into breakaways, which to save for the finish, and where the marginal hours of training should go. The data tells them, and the platform makes the data legible without a statistician on the support staff.