Race time predictions

VDOT source

The engine uses the current VDOT, defined as the median of the top-3 VDOTs in your last 28 quality activities with clean data.

Each individual VDOT comes from Daniels' formula: VO2 = 0.000104 × v² + 0.182258 × v - 4.60, with v in m/min. We solve the quadratic for vVO2max and derive VDOT.

VDOT → race time mapping

Each distance has a coefficient that links VDOT to time. 5K ~98% vVO2max; 10K ~92%; HM ~88%; M ~84%; 50K ~78%; 100K ~70%.

Tables from Daniels' Running Formula (2014), complemented with elite data for long distances (Magness, 2018).

Altitude correction

If the race is > 5000 ft (1500 m), the time is penalized. Model: for every 3000 ft (1000 m) above sea level, aerobic performance drops 6-10% (higher = worse).

For somebody training in Veladero (13800 ft / 4200 m), this correction is critical. Without it, projected times would be 8-15 min faster than reality.

Where prediction loses precision

Trail with lots of vert: prediction is less reliable when D+ > 3300 ft (1000 m). We use GAP when trail data is available, but for UTMB/MIUT-style events, prediction can be off ±10%.

No recent PR (< 21 days): VDOT may be stale. Vetta flags it as "low confidence" in the dashboard.

Very advanced athletes (VDOT > 70): Daniels' formula was built for amateurs. For sub-elite the predictions tend to be conservative.

Public API

GET /api/v1/predict?vdot=50&distancia_km=42&desnivel_pos_m=200&altitud_media_m=0&tipo=calle

Returns seconds, formatted, pace_s_per_km, pace_formatted, breakdown.

Public calculator

/herramientas/vdot computes your VDOT from a race result and projects times for 5K-50K with altitude correction.