Two rides, same average watts, yet one left you flat. Normalized power is a weighted average that counts hard, variable efforts more heavily, so it reflects the cost of a ride better.
It is one of the core numbers in training with power.
| Item | Value |
|---|---|
| Rolling average window | 30 seconds |
| Exponent applied to each window value | 4 (fourth power) |
| Final step | Fourth root of the average |
| First window starts at | 30 seconds into the ride |
| Coaches taught to dismiss NP for durations under | 10 minutes |
These figures come from TrainingPeaks’ coaching blog. Figures and definitions can change, so check the source before you rely on them.
Why average power can mislead you
Average power treats every second equally. A steady ride and a ride of surges, coasting and sprints can share the same average, yet the second is far harder on your body.
The TrainingPeaks coaching blog says: “It’s something that, as experienced cyclists, we know from experience; two rides with the same average power can be vastly different in terms of their difficulty.” It adds that a ride with high variation, such as intervals, is often more difficult than a ride of moderate steady-state riding.
Coasting is the big distortion. As the same source notes, “zeros in a math equation can dramatically affect the average.” That is why a hilly group ride with plenty of freewheeling can show a modest average while still being a hard day.
Normalized power, often shortened to NP, is an attempt to put a number on that difference. In the words of the source, “The aim of NP is to reflect the fact that these variable efforts are physiologically more costly.”
How normalized power is calculated
The method has four steps, as described by TrainingPeaks:
- Calculate the 30-second rolling average of power, starting at 30 seconds and repeating for every second.
- Raise each of those values to the fourth power.
- Average the results of step two.
- Take the fourth root of that average. The result is your normalized power.
Raising to the fourth power makes big numbers count disproportionately, and the root brings the result back to watts. In practice your head unit, app or analysis software does this for you, second by second, from the recorded file.
It helps to think of the result as a weighted average. The source describes it as giving “a moderately higher value on larger power numbers by reducing the impact of low power numbers.”
A worked example
TrainingPeaks offers a scenario: a cyclist rides at 400W for 70 seconds, does a 30s surge at 600W, and then settles back to 400W for 20s. A plain average would blur the surge into the steady riding.
Normalized power does not let the surge disappear. Because the surge is raised to the fourth power, it pulls the result above what the average alone would suggest.
The source also gives a real-world style case: an example file has an average power of 450W and a normalized power of 478W, as calculated by TrainingPeaks. That gap is the model’s way of saying the ride was harder than its average implies, because the harder stretches are weighted more heavily than the easy ones.
Why a 30-second window and a fourth power?
TrainingPeaks explains that these values reflect that physiological responses are not instantaneous and do not scale in a straight line with intensity. A short window smooths out momentary spikes, and the exponent rewards harder work.
Your body does not jump to a new state the instant you stand on the pedals. It takes time to respond and to recover, so a rolling window is a closer match than a raw second-by-second reading.
The same source notes that if the exponent and root were removed, the calculation reduces to essentially average power. The weighting is what makes the metric useful.
It also explains what happens when the window grows. As the window approaches the length of the whole ride, the value trends toward average power, because the smoothing swallows the variation. The window is deliberately short for that reason.
Reading normalized power next to average power
Compare the two numbers for the same ride. The relationship tells you something about the shape of the effort.
- NP close to average power: the effort was steady, such as a flat solo ride at a constant pace.
- NP well above average power: the ride was variable, with surges, climbs, intervals or a lot of coasting.
- NP never meaningfully below average: because it weights hard work more, it should not come out lower than the plain average.
The gap between the two is a rough guide to how stop-start the ride was. Looking at the raw trace in your bike power data shows where the gap comes from, whether that is a few big sprints or constant small accelerations.
This comparison is most useful across similar rides. If your NP and average were nearly identical on a flat tempo ride last month and are far apart on a rolling route today, the difference likely reflects terrain and pacing; it does not by itself establish a change in fitness.
A practical habit is to look at both numbers after every ride. Over time you learn what your typical gap looks like on flat days, on hilly days and in group rides, and an unusual gap becomes a prompt to check how the ride was paced.
If the gap is large and you felt the ride was easy, check the file for dropouts or spikes in the power trace before trusting the figure.
Limits and cautions
Normalized power is a model, not a measurement. It is less meaningful on short efforts: at TrainingPeaks, coaches are taught to dismiss NP values for durations under 10 minutes.
The reason is simple. With a 30-second window, those first 30 seconds make up too large a share of a short effort, so the number says little about the ride as a whole.
The name is also tied to one vendor’s method, so other platforms may compute a similar figure slightly differently. Compare numbers from the same software when you track trends, and avoid mixing exports from different apps.
NP also feeds other metrics. Your training stress score builds on it, so any distortion in NP carries into that figure.
Finally, remember that NP knows nothing about heat, sleep, fuelling or the wind. It describes the power you produced, not how you felt, so treat it as one input alongside your own judgement.
In short, use normalized power to compare the cost of rides that look alike on average power but felt different. Keep it in context, check the file quality, and let the number support your training decisions rather than make them for you.
Frequently asked questions
Is normalized power the same as average power?
Show answer
No. Average power weights every second equally, while normalized power weights harder efforts more heavily through the fourth-power step.
Can normalized power be lower than average power?
Show answer
Not in normal use. The weighting favours higher values, so over longer efforts NP usually comes out equal to or above the average, though it can be lower over short intervals.
Is normalized power useful for short rides?
Show answer
Less so. TrainingPeaks coaches are taught to dismiss NP values for durations under 10 minutes.
Do I need to calculate it myself?
Show answer
No. Most cycling software computes it automatically, but knowing the four steps helps you read the number sensibly.
