LIGHT BOX RESEARCH

Tag: biking

  • How many rides does your power curve actually need?

    Your power-duration curve, the best watts you can hold at every
    duration from a sprint to your longest ride, is treated as a measurement
    of what you can do. But it is really a best-of. And you can’t produce
    your best across the full power-duration span in just a ride or two. So
    how many rides of a decent duration would you need to get a solid
    estimate of your ability?

    We were building a durability study on GoldenCheetah field data and
    wanted to know the effect of different observation windows on the MMP
    (mean maximal power), that same best-watts-at-every-duration curve.
    Working with our cleaned dataset, GCclean, we took 200,050 rides of 1.25
    to 2.25 hours, across 1,854 riders and 3,966 rider-seasons. These are
    only the rides that fall in that duration band, not the whole season,
    about 51 per rider-season on average. For each season we recomputed the
    power curve while shrinking the pool of rides we searched: all of them,
    then 75%, 50%, on down to a single ride.

    No surprise, the curve sags

    Compared to searching all of those rides, a single ride will have an
    MMP 39% lower at 15 seconds, and 22% at 20 to 30 minutes. The sprint end
    of the curve will on average be further off the all-rides MMP than the
    endurance end. In this dataset, the MMP estimate got much better at
    about 15 rides and was solid by 25 or so. These are randomly sampled
    rides so that number could likely be lower with a few intentional
    efforts spread out over several rides.

    Power-duration curves as the search space shrinks from all of a rider’s in-band rides (top) down to one ride (bottom). The sprint end collapses; the aerobic end holds.

    Thinner data is also noisier

    An MMP built from a couple of rides is not only lower, but jumpier as
    well. The variability is higher at the short end, with a one-ride sprint
    peak (Pmax) varying about 27%, while the endurance end varies by about
    14% from ride to ride.

    How reliable an MMP estimate is, plotted against how many rides went into it. Fewer rides means more scatter, worst at the sprint end.

    What to take from it

    I think the main take home is the importance of throwing in some
    dedicated efforts sprinkled across the duration of the curve on a
    regular basis. Then when thinking about monitoring trends, 2 weeks is
    going to be a bare minimum for looking for changes in trends. For
    reliable data, month to month comparisons would be the way to go.

    One caveat: these numbers are averages across about 4,000
    rider-seasons, so treat the ride counts as a rule of thumb for the
    group, not a guarantee for any one rider.

    Trzymaj się