Start with the race, not the winner
Cross-country times do not travel cleanly from one course to another. A fast, flat September course and a cold, rolling championship course can reward the same fitness with very different clock times. Altitude begins with the complete result list so the gaps and shape of the field help describe the day.
Convert every distance to a common baseline
Results at common high-school distances are projected to a 5K equivalent before comparison. That gives the model one unit of time, while preserving the original distance and result in the athlete’s source record.
Estimate a race correction
Runners with reliable performances on other courses act as anchors. The model compares what those athletes were expected to run with what happened in this race. A robust center reduces the power of one breakthrough, one poor race, or one inaccurate record to move the entire field.
Reward repeat evidence
A season index should not be a runner’s single best afternoon. Altitude weights the best two ratings, the middle of recent form, consistency, and the number of credible race lines. One-result returners remain clearly labeled as provisional seeds; deeper profiles earn higher confidence.
Keep uncertainty visible
Course ratings are estimates, not official measurements of an athlete. Small differences should not be treated as absolute. The practical value is comparing performances across different Colorado courses and building transparent virtual fields—not pretending the model can remove every unknown.
Use ratings for a virtual meet
A virtual meet ranks selected athletes by their current comparable index and scores teams using standard five-runner cross-country scoring. It is a projection from available evidence, not a prediction that replaces racing. Form, weather, tactics, health, and course fit still matter on meet day.
The published scale
The working rating line uses three seconds per point after race correction. Higher is faster. Altitude is an independent model inspired by the general corrected-time concept documented by TullyRunners, but its automated corrections, evidence weighting, confidence labels, and published outputs are its own.
Read the historical TullyRunners explanation ↗