Coaches face hard enough challenges without guessing how to compensate for drift in their riders’ power meters. Accurate, repeatable power measurement makes top-level training more focused and meaningful.
For coaches, ‘drift’ is a real and important issue. Broadly, drift is the tendency of a measuring instrument to wander off from its settings.
When a power meter is used to measure the outcome of a carefully designed training regime, drift is more than a nuisance. It undermines a strategy and compels the trainer to guess at the true values recorded by the rider. And the higher the rider’s level, the more significant the effect of this guesswork is.
All will know the well-rehearsed maxim of the digital era, ‘garbage in, garbage out’.
Dirty data caused by drift compromises even the best thought-through training strategy.
In a bid to combat this effect, some of the better power meters provide a so-called ‘auto-zeroing’ function to limit drift. This may or may not work. But why not prefer InfoCrank—a power meter that’s accurate by design, and not subject to drift? A power meter that calculates power from first principles, that is, tangential torque x cadence.
Here, our aim is to give you one practical example of how drift, however subtle, undermines the value of the tool used to measure the rider’s power output and compromises a training plan.
One of our sponsored riders (a national champion in his sport) reported this story to us. He was using a power meter that’s claimed to be ‘tried, trusted and true’. For this three-hour ride on an undulating course, he performed three 20-minute intervals at race pace. Intervals of this kind, when prescribed by a coach as part of a training regime, are meant to be carried out at an exact cadence and power level.
The rider was preparing using the key metric of weight/power. On the day, this translated to 263 Watts as measured by InfoCrank.
The other power meter read 245 W for the first interval. This upset the rider because he thought the PM was accurate. (A clever algorithm supposedly compensates for the inaccuracies introduced by measuring elsewhere than at the crank, in the direct load path.)
There was another problem with the measurement: the actual average cadence for the first 20-minute interval, measured by InfoCrank, was 89 rpm. The other PM recorded a cadence of 95 rpm. Anyone in serious training knows this is quite a different style of riding!
When speaking to coaches, we find they often suggest results like these are not really significant. They say, and their clients echo it (reviewers too): ‘As long as the numbers are consistent, it’s OK!’
We say that’s not OK. Let’s see how the rest of the training session pans out.
The data from the rider’s second interval was similar to the first (relatively speaking). Power was still lower than actual, varying 0.5% from its first reading. This meant that the wattage had drifted 0.5% during the rest period—not much to worry about.
More important, though, was that the cadence reading on the other PM had also drifted. Now it was almost accurate. In the first interval, it was out by 5.6% compared to InfoCrank, this time it was out by only 1.1%. The rider was actually riding at the same cadence (1 rpm different over 20 minutes) in interval two compared to interval one, but the other PM suggested he was riding 7 rpm slower!
In interval three, results were problematic. The other PM now had a variance of >3% compared to its previous reading. This nullifies the vague notion that the coach can infer true output because this PM measures consistently low. The third interval had varied by nearly 3.5% from the first interval. Just for clarification, the rider’s actual power had decreased slightly in the final interval by just over 2.5%, but the PM measured his power as increasing by 0.4%. That is not consistency—it’s drift!
Just as in virtually all areas of life that involve measurement (speed in the car, weight on your bathroom scales, telling the time), the measurement of human power output on a bicycle evolves—from an approximation using indirect methods, to highly accurate (and technically correct) direct measurement.
Few in the coaching community will disagree that, to improve human performance, we must prefer accurate power measurement. With InfoCrank, that is now available, and at a price comparable to that of first-generation indirect systems.
We say the argument for ‘consistency’ is contrived and specious. Close examination of just this one outdoor, real-world training ride shows it’s nonsense.
Coaches—dump those worn-out ideas! It’s time to engage with and embrace accuracy!
*The raw data from the comparison ride over three hours with 3 x 20 minute intervals is below. Only the data for the intervals was reviewed for this article.
For complete transparency: This article in its draft form was sent to the athlete for comment. He agreed with the summary of the data. However, he decided to do another test, this time indoors—it is the middle of winter here. We have not analysed the drift of this recent 1hr 45min ride, but the average deviation of the other PM is minus 3.5% compared to the same InfoCrank. As with the initial test, the athlete wrote that he believed the PM was also more erratic than the InfoCrank—that is, greater deviation.