How we estimate power on climbs

The climb power figures on this site are our own calculation, not a measurement and not someone else's estimate. Every one is labelled Estimate · howtoturnpro. This page gives the formula, how the rider's weight is chosen, every effort we checked against a power meter, and what the method cannot see.

The formula

The same physics as Strava's own estimate: the power needed to move rider and bike up the climb at the measured average speed, against gravity, rolling resistance and air drag, with no wind and no drafting.

P = (m·g·v·sin a + m·g·v·Crr·cos a + ½·ρ·CdA·v³) / drivetrain

• m = rider + bike (6.8 kg, or the rider's 2026 road bike where his profile gives it) + 1 kg of kit
• Crr (rolling resistance) 0.0045; CdA (drag area) 0.32 m²; drivetrain loss 2.5 %
• ρ (air density) from the climb's mid altitude and a typical temperature for the month
• v from the segment length and the rider's time; a from the segment's elevation gain
• Climbs flatter than 6 % are left out: there, wind and drafting matter more than the formula.

The weight

W/kg divides the estimate by the rider's published weight. The weight comes from data/rider_profiles/, every value with its source. For each effort we take a weight given for that very race first, then by kind of source and closeness in years: the rider's own words or his team's page, then the press, then reference databases (such as a magazine's team guide), then an unsourced encyclopedia entry.

• A: the team's page from 2025–26, the rider's own words, or a weight given at a race in the last two years.
• B: an older or undated team page, or a dated press figure.
• C: a reference database or Wikipedia. Shown with an asterisk: weight from a reference database, not the team or the rider.

Where the sources for a rider differ by more than 3 kg and the best is tier C, we show the time only, not W/kg. A 3 kg error in weight moves W/kg by about 1 % but watts by about 4 %, which is why W/kg is the figure we put first and watts go in the detail table.

The check: 55 efforts with a power meter

For riders whose power file for that climb is public, we compare our estimate with what the meter recorded. Median error 2.0 %; median signed error -0.5 % (no systematic lean); 43 of 55 within 5 %, 52 of 55 within 10 %. Efforts from 2017 to 2023 are used here only, for the check, and never on a climb page.

What the method cannot see

• Wind and shelter. That is why climbs under 6 % are not included.
• The time is the rider's best time on the segment that day. On a climb raced twice in a stage it can be the first passage (Piancavallo, for example).
• Not every rider shares his rides. Riders who don't are missing, so a climb table is not a stage result.
• We tried correcting for the conditions of the day, using the other riders with a power meter on the same climb. It did not reduce the error (2.1 % without it, 2.6 % with it), so it is not applied. Most of the error is individual, and most of that is weight.

Climbs link only to the Strava segment, never to a rider's activity or profile.

All climbs →