How KomPredict Works: From Strava Data to a Time Prediction
KomPredict answers one question: is today a good day to attack this segment? Here is the full path from data to answer.
Step 1: your data
After connecting Strava, your starred segments are imported — with length, gradient, bearing and your existing best time. In your profile you store weight, power values and bike type, which sets the aerodynamic baseline.
Step 2: hourly weather
For the segment location a rolling 24-hour forecast is loaded: wind speed, wind direction, gusts and temperature. Wind direction plus segment bearing produce effective tailwind and the crosswind share.
Step 3: best time as anchor
If a best time exists, it is the foundation. The model does not compute what you could theoretically do, but how today's conditions compare with the conditions of your best time. Only without history does the pure physics model take over from power, CdA, rolling resistance and gradient.
Step 4: simulation and probabilities
Over a thousand runs with scattered inputs produce a distribution of possible times. From it come PR probability, top-10 and top-50 chances, KOM chance and a confidence band.
Step 5: decision and explanation
Everything condenses into an attack score with a clear verdict and a best time window. Every metric carries an explanation with formula and an interpretation of your specific value — so you do not have to believe the numbers, you can follow them.
In short
Best time as anchor, hourly weather, simulation for probabilities, one score for the decision.
Frequently asked questions
Do I need a power meter?
No. For segments where you already have a best time, time and weather are enough. Power values improve predictions on new segments.
Does it work outside Germany?
Prediction quality depends on weather data coverage; German-speaking Europe is currently covered best.
What happens to my Strava data?
Only the data needed for predictions is stored. Details are in the privacy policy.