Average power vs Normalized Power
How two rides can have the same average
Ride steadily at 200 W for twenty minutes and you complete 240 kJ of mechanical work. Alternate between 100 W and 300 W for equal lengths of time and the arithmetic average, duration and work can be the same. The trace tells a different story. This matters when comparing a steady endurance ride with a stop-start race: the same average can hide very different patterns of effort.
Move the variation control. Our example alternates two-minute blocks around a fixed 200 W centre. Every watt added to the high block is removed from the low block. The average stays put by construction.
Solid: raw power. Dashed: 30-second mean, beginning after the first full window. Two-minute alternating blocks; 20-minute example.
How two rides can have the same average
Ride steadily at 200 W for twenty minutes and you complete 240 kJ of mechanical work. Alternate between 100 W and 300 W for equal lengths of time and the arithmetic average, duration and work can be the same. The trace tells a different story. This matters when comparing a steady endurance ride with a stop-start race: the same average can hide very different patterns of effort.
Move the variation control. Our example alternates two-minute blocks around a fixed 200 W centre. Every watt added to the high block is removed from the low block. The average stays put by construction.
How Normalized Power works
Normalized Power applies a 30-second rolling average, raises those averages to the fourth power, averages them, then takes the fourth root. Large values therefore have more influence than they do in an ordinary average.
It was designed to estimate an equivalent steady power for the physiological demand of a variable effort. It remains a model-derived summary, not a direct measurement of your metabolism.
A closer look at the calculation
At zero variation, both lines read 200 W. Increase variation and compare the displayed raw trace with the smoothed trace. The smoothing softens each transition, while the fourth-power step gives the higher plateaus extra weight.
Our calculation uses complete 30-second windows only. It never fills the beginning with fictional zeroes. Selecting a short extract in another tool can change edge handling; compare the same full recording and conventions.
When to use each number
Average power answers how quickly you did mechanical work over the selected time. Normalized Power adds information about variability. Neither tells you exactly how long you can recover in a race or how you should feel tomorrow.
Avoid using this metric to judge a single short sprint. TrainingPeaks documents that short-segment values can behave unexpectedly. Inspect the trace and the effort duration before explaining a surprising result.
Why surges can feel harder than the average
Power is an external measurement: it describes mechanical work at the meter. Your body must supply energy to produce it, and the response to a hard surge does not instantly disappear when power drops. Repeated accelerations can therefore create a different pattern of strain from riding steadily at the same average watts.
In a study of 15 elite cyclists, researchers compared 28-minute steady bouts with bouts varying around the same average. The variable bouts produced higher oxygen uptake, heart rate and blood lactate under the tested conditions. That experiment shows why a simple average can miss a difference; it does not prove that one formula predicts the response of every rider.
What the calculation rewards
The thirty-second smoothing step reduces the influence of an isolated instantaneous spike. Raising the smoothed values to the fourth power then gives harder periods more weight. A ride with repeated efforts can therefore have a higher Normalized Power than a steady ride with the same average.
The result is still one number. Two rides can have similar Normalized Power but different interval lengths, recovery opportunities and order of efforts. A five-minute climb followed by rest is not the same experience as many short accelerations just because a summary happens to match.
Use the two numbers together
Use average watts when you want to understand total mechanical work: average power multiplied by elapsed recording time gives work if the samples and duration are complete. Use Normalized Power as a model-based description of variability over a suitable interval.
When a value surprises you, open the trace. Check that you selected the intended segment, that zeros and pauses are handled consistently, and that spikes or missing samples are not distorting the result. Neither number by itself tells you recovery time or the quality of a workout.