Average power tells you the mechanical mean of a ride. Normalized Power® estimates what the same ride might have cost physiologically if the effort had been steady. Neither metric is universally better: the useful one depends on whether you are pacing an interval, evaluating a variable ride or estimating training stress.
When a ride is smooth, the two numbers should be close. When it contains repeated climbs, accelerations and coasting, Normalized Power is usually higher because harder efforts are weighted more heavily. Compare both numbers, then use Variability Index to quantify the gap.
Average power is an output metric; Normalized Power is an interpretation metric. Pace steady intervals with average power. Review variable rides with both. Do not treat a high Normalized Power as power you literally sustained.
If these metrics are new to you, begin with the Cycling Power Zones Guide. You can then use the Cycling Power Zone Calculator to turn a current FTP into practical targets.
Normalized Power vs average power at a glance
| Question | Average power | Normalized Power |
|---|---|---|
| What does it describe? | The arithmetic mean of recorded watts | An estimate of the steady power with a similar physiological cost |
| How are hard surges treated? | Every recorded watt has equal mathematical weight | Higher efforts receive disproportionate weight |
| Does coasting matter? | Yes, zero-power periods lower the mean when included | Yes, but NP does not simply delete zeros |
| Best use | Intervals, tests, mechanical work and steady pacing | Post-ride intensity, variable terrain, IF, TSS and VI |
| Main limitation | Can understate how costly repeated surges were | Is a model, not a directly measured physiological response |
| Short efforts | Useful when the averaging window matches the effort | Less informative because the calculation uses a 30-second rolling average |
What is average power?
Average power is the sum of the recorded power values divided by the number of samples. If a one-hour ride records 200 watts on average, the rider produced a mean external power of 200 watts across that file.
This simplicity is useful. Average power is directly connected to mechanical work: at a constant 200 watts for one hour, the work performed at the cranks is approximately 720 kilojoules. It is also the appropriate headline number for a continuous maximal test, a steady interval or a time trial in which you want to know what power was actually sustained.
Some head units can display an average that excludes zero-power samples, while most training analysis relies on the complete recorded file. “Non-zero average power” will be higher whenever you coast. Compare like with like before drawing conclusions.
What is Normalized Power?
Normalized Power is a proprietary cycling metric developed to account for the nonlinear cost of variable effort. A short surge well above threshold can create much more strain than the same amount of time spent slightly below the ride average. NP weights those harder periods so a stochastic ride is not reduced to one deceptively moderate mean.
The result is expressed in watts, but it is not another direct measurement from the power meter. It is an estimate generated from the power file. A ride with 190 watts average and 225 watts NP does not mean the rider continuously produced 225 watts. It means the variable ride is estimated to have imposed a cost resembling a steadier effort near that level.
Normalized Power, Intensity Factor and Training Stress Score are trademarks of Peaksware. Other platforms may show names such as weighted or adjusted power and may use a related—but not identical—calculation. Small differences between apps do not necessarily indicate bad data.
How is Normalized Power calculated?
The calculation is easier to understand as a sequence than as one formula:
-
Create a 30-second rolling average
Second-by-second power is smoothed over 30 seconds to approximate the delayed response to a change in effort.
-
Raise every rolling value to the fourth power
This is the key weighting step. Higher values grow dramatically faster than moderate values.
-
Average the transformed values
The calculation combines the entire selected ride or segment after the intensity weighting has been applied.
-
Take the fourth root
This converts the result back into a watt value that can be compared with average power and FTP.
The algorithm does not merely remove coasting. It smooths the file and disproportionately weights the harder rolling averages. That is why repeated one- or two-minute surges can raise NP substantially even when easy descending or drafting keeps average power modest.
How two rides can have the same average power but feel different
Consider two 60-minute rides. Both finish at 200 watts average:
Steady indoor effort
The rider stays close to 200 watts with only small fluctuations. Average power is 200 watts, NP might be 203 watts and VI is 1.02.
Hilly, surging effort
The rider repeatedly climbs at 280–330 watts and coasts downhill. Average power is still 200 watts, but NP might be 235 watts and VI is 1.18.
The second ride is likely to create greater fatigue despite identical duration and average power. The exact NP depends on the complete power trace, not just a few summary numbers, so these values are illustrative rather than a way to reconstruct NP manually.
Why is Normalized Power higher than average power?
NP rises above average power when the ride contains enough variation for the weighting of hard efforts to outweigh easy sections. Common causes include:
- Repeated accelerations out of corners
- Short climbs ridden well above the ride average
- Hard pulls followed by drafting in a group
- Intervals separated by easy recovery
- Technical terrain with bursts and coasting
- Starting too hard and fading later
A large gap is not automatically good or bad. It can be appropriate in a criterium, mountain-bike race or interval workout. It is usually less desirable during a long-course triathlon bike leg, where unnecessary spikes can consume energy needed for the run.
Variability Index: quantify the gap
Variability Index compares the two metrics:
A ride with 220 watts NP and 200 watts average has a VI of 1.10. The closer VI is to 1.00, the steadier the power delivery was.
| VI | Typical interpretation | Possible context |
|---|---|---|
| 1.00–1.05 | Very steady | Indoor endurance, flat time trial, controlled long-course triathlon |
| 1.06–1.10 | Moderately variable | Rolling endurance ride or hilly race paced with control |
| 1.11–1.20 | Highly variable | Hard group ride, interval session or technical road race |
| Above 1.20 | Very stochastic | Criterium, mountain biking, repeated attacks or extensive coasting |
These bands are descriptive examples, not pass–fail standards. Terrain, race format, tactics, stops and the selected file segment all affect VI. Compare similar rides and ask whether the variability served the purpose of the session.
Which metric should you use?
| Situation | Primary metric | Why |
|---|---|---|
| 20-minute FTP test | Average power | The protocol is based on power actually sustained across the test interval |
| Steady threshold interval | Lap average power | Shows whether the target was held; NP can hide poor pacing within the interval |
| Hilly endurance ride | Both | Average describes output; NP and VI reveal the cost and variability |
| Interval workout summary | Both | Interval averages assess execution; NP helps describe the complete session load |
| Long-course triathlon | Real-time power plus NP and VI | Short averaging controls immediate effort; NP and VI help cap cumulative cost |
| Mechanical work | Average power and kilojoules | NP is not used to calculate actual external work |
| Training stress | NP with FTP and duration | NP feeds Intensity Factor and Training Stress Score |
How to use both metrics after training
-
Confirm the file is trustworthy
Look for power dropouts, calibration problems, an incorrect recording device or long pauses. A sophisticated metric cannot rescue bad input data.
-
Judge the key work directly
For intervals, inspect each work period using average power, duration, cadence and perceived effort. Do not use whole-ride NP to decide whether a specific interval was successful.
-
Compare NP with average power
Calculate or read VI and decide whether the variability fits the terrain and objective. A high VI on an easy endurance ride may reveal avoidable surging.
-
Add duration and FTP context
NP alone cannot tell you total load or individual difficulty. Use the correct current FTP, duration, heart rate, RPE and recovery response.
-
Compare like with like over time
Track similar routes, workout formats and file selections. A two-hour hilly group ride should not be judged against a 45-minute indoor session from NP alone.
Normalized Power for triathlon pacing
Triathletes need to optimise the complete race rather than produce the highest possible bike number. On a flat or rolling course, a relatively small gap between NP and average power generally indicates controlled pacing. On steep or technical courses, some variation is unavoidable, but large spikes can still carry a disproportionate cost.
To turn these metrics into an executable 90 km plan, use the 70.3 bike pacing strategy, including duration-based IF ranges, watt tables, TSS estimates and hill caps.
Use three layers of feedback:
- 3- or 10-second power: prevents immediate spikes and guides short changes in terrain.
- Lap average power: checks whether the current section is matching the race plan.
- Lap or ride NP: estimates the accumulating intensity of the variable effort.
Do not chase a low VI by riding an unsafe or mechanically inefficient line, and do not chase NP uphill because it is lagging. Set conservative caps, ride the terrain smoothly and validate the plan through race rehearsals. Our Ironman Pacing Strategy explains how bike restraint protects the marathon.
How NP connects to IF and TSS
For a full worked explanation of both metrics, see our guide to cycling TSS and Intensity Factor, including the formulas, practical ranges and the limitations of comparing unlike sessions.
Normalized Power becomes more useful when it is related to the rider's threshold and the duration of the session.
IF = NP ÷ FTP
A 210-watt NP with a correctly set 250-watt FTP produces an IF of 0.84. The same NP represents a different relative intensity for a rider with a different FTP.
Duration plus relative intensity
TSS combines time and intensity derived from NP and FTP. It is useful for tracking workload, but remains an estimate rather than a direct measurement of recovery required.
If your threshold value is stale, IF and TSS become misleading. Use a consistent testing method, described in our guide to ramp and 20-minute FTP tests, and judge the number against real workouts.
Common mistakes
Using NP to set FTP
A high NP from a stochastic ride is not proof that you can sustain that wattage near threshold. Estimate FTP with an appropriate test or performance effort, then validate it through training.
Chasing the bigger number
NP is often higher than average power, but higher is not inherently better. Efficient racing sometimes means conserving energy until the moments that decide the result.
Ignoring the selected time range
Whole-ride, moving-time and segment NP can differ. Warm-ups, cool-downs, café stops and paused recording change the comparison. State the time range when comparing files.
Using whole-ride NP to grade intervals
A workout can have an impressive NP while the actual intervals were erratic. Assess each repetition with lap average power and execution quality.
Comparing different platforms as if the metric were identical
Weighted or adjusted power may resemble NP without reproducing it exactly. Use one platform consistently for long-term trends.
Forgetting that training should solve a problem
Metrics describe the session; they do not prescribe the next one. If the goal is to raise sustainable power, use the progression principles in our How to Improve FTP Guide instead of simply trying to maximise NP.
Normalized Power vs average power: frequently asked questions
Is Normalized Power more accurate than average power?
It answers a different question. Average power accurately describes the arithmetic mean of recorded watts. NP is intended to estimate the physiological cost of variable output. NP can be more informative for a surging ride, while average power is more direct for steady intervals and mechanical work.
Should Normalized Power always be higher than average power?
For a sufficiently long, valid cycling file, NP is normally equal to or higher than average power. Very short selections, recording behaviour or differences between software implementations can produce confusing results, so inspect the file and selected time range.
Can Normalized Power be higher than FTP for an hour?
It can appear that way after a highly variable effort because NP is a model rather than power continuously sustained. It can also reflect an underestimated FTP, calibration error or an unsuitable file selection. Do not automatically raise FTP from one high NP value.
Does average power include zeros?
Most complete-file analysis includes zero-power periods, but some cycling computers can display non-zero average power. Check the device or platform setting before comparing values.
What is a good difference between NP and average power?
It depends on the ride. A steady workout or long-course triathlon may be near 1.00–1.05 VI, while intervals, group rides and technical races can be much higher. The useful question is whether the variability matched the purpose and terrain.
Should I show NP on my bike computer?
It can be useful for long rides and races, but it should complement—not replace—short-duration power and lap average power. NP responds too slowly to control every surge and can tempt riders to chase a lagging number.
Is Strava weighted average power the same as Normalized Power?
No. It serves a similar purpose but is not guaranteed to use the identical proprietary calculation. Compare trends within one platform rather than expecting exact agreement across platforms.
Which power should I use for a 20-minute FTP test?
Use the average power sustained across the prescribed 20-minute test interval. Do not substitute NP. Then apply only the calculation required by the specific test protocol.
Final takeaway
Average power and Normalized Power are complementary. Average power tells you what the power meter recorded on average and remains the clearest metric for steady intervals, tests and mechanical work. NP helps interpret why two rides with the same average can create different levels of strain. Read the numbers together, quantify the gap with VI, add duration and FTP context, and always judge the result against the purpose of the ride.
This guide treats Normalized Power as a practical model rather than a direct physiological measurement. Definitions, the four-step calculation and related metrics were checked against the official TrainingPeaks explanations of Normalized Power, Variability Index and NP, IF and TSS. Practical limitations were cross-checked against the 2026 CTS coaching review of how NP should and should not be used. The physiological context—that intensity and variability matter beyond total mechanical work—is supported by a systematic review of cycling durability. Race-specific context was informed by research on power demands across sprint-triathlon courses and a 2025 open-access study using normalized-power principles in cycling time-trial pacing optimisation. Example wattages and VI bands are explanatory scenarios, not universal physiological thresholds or individual coaching prescriptions.
