Method
45% less error on your split times: how we measure it
The split times RAVIT gives you have a median error of 6.3%, versus 13.2% for a standard calculation. Here are the test, the numbers, and what they don't tell you.
The test
118 races already run, 98,479 finishers.
For each edition, RAVIT calculated every runner's split times from their UTMB Index and the results of previous editions only, just as for a runner preparing the next one.
Source
These times were compared with the actual times, checkpoint by checkpoint.
The same exercise was done with a standard calculation, for comparison.
The numbers
71% of our times land within 10% of the actual time, versus 39% for the standard calculation.
Basis: split times over the last two thirds of the race, 91 races, 79,095 finishers.
- RAVIT
- Generic model
- Standard calculation
What is the standard calculation?
A constant pace per effort-km: 1 km counts as 1, and every 100 m of elevation gain adds 1 effort-km.
Pace is derived from the UTMB Index: pace = exp(2.555 − 0.00123 × index) minutes per effort-km, a relationship fitted on all the other races in the test.
No fatigue, no race-specific terrain: a simple pace calculation.
What this number doesn't say
It compares RAVIT with a standard calculation, not with every app on the market.
It holds for races where we have results from previous editions. On the others, our generic model has 37% less error than the standard calculation (interval 27% to 45%).
It is a median error: for half of runners, the gap is larger.
Only runners who reached the finish are counted.
The UTMB Index used is the one published in each edition's results.
Measured on October 9, 2026, redone with every change to the engine. Confidence intervals: 1,000 random draws of the races.