Normalized Power, TSS and the fitness–fatigue chart
Average power hides how hard a ride really was. Normalized Power, Intensity Factor and TSS fix that, and the fitness–fatigue chart turns them into a training load you can plan with.
Why average power is not enough
Two rides can have the same average power and feel completely different. A steady 200 W for an hour is easy to digest. Five minutes at 300 W followed by five minutes at 100 W, repeated for the same hour, also averages 200 W but leaves you flat. Physiology does not respond linearly to power: the cost of an effort rises faster than the watts do. Normalized Power (NP) is a way of reflecting that on a single number: the steady power that would cost roughly the same as your variable ride.
How Normalized Power is calculated
- Take the power at 1-second resolution and compute a 30-second rolling average at every point.
- Raise each of those averages to the fourth power.
- Take the mean of all the values.
- Take the fourth root.
In plain text: NP = (mean of P30⁴)^¼, where P30 is the 30-second rolling average. The 30-second window smooths out single pedal strokes and brief spikes; the fourth power then gives disproportionate weight to the hard parts. For the interval example above (5 minutes at 300 W, 5 at 100 W), the average is 200 W but NP is about 253 W, because 300⁴ is eighty-one times larger than 100⁴. For a constant 200 W ride, NP equals 200 W.
GearWatts calculates NP from the moving time of the ride at 1 Hz, including the zeros when you coast and excluding stopped time. A ride needs at least 30 seconds of data for NP to exist, and the number is only meaningful for rides of at least 20 minutes or so; on a 12-minute ride it is just a heavily smoothed average.
Intensity Factor and TSS
Intensity Factor (IF) compares NP with your threshold, your FTP (see FTP and power zones):
IF = NP ÷ FTP
Training Stress Score (TSS) multiplies that by duration:
TSS = hours × IF² × 100
The scale is anchored so that one hour exactly at FTP is 100 TSS. A few examples with an FTP of 250 W:
| Ride | Moving time | NP | IF | TSS |
|---|---|---|---|---|
| Easy spin | 1 h | 150 W | 0.60 | 36 |
| Steady endurance | 3 h | 180 W | 0.72 | 156 |
| Hilly club ride | 2 h | 215 W | 0.86 | 148 |
| Hard race | 1.5 h | 240 W | 0.96 | 138 |
Note that the 3-hour endurance ride costs more TSS than the hard 90-minute race. That is not an error: the model treats stress as quadratic in intensity but linear in time, so a long easy day can exceed a short hard one. Whether that matches how tired you feel is a personal question, which is why the numbers are a guide, not a verdict. GearWatts computes the TSS of each ride from NP, FTP and moving time, and stores it with the ride.
Fitness, fatigue and form
Summing TSS over days gives a stress history. The performance chart smooths it with two exponentially weighted averages:
- CTL (chronic training load, "fitness"): average of daily TSS over about 42 days.
- ATL (acute training load, "fatigue"): the same over about 7 days.
- TSB (training stress balance, "form") = CTL − ATL.
A common way to update them each day is CTL_today = CTL_yesterday + (TSS_today − CTL_yesterday) ÷ 42, and the same with 7 for ATL. Because the 7-day average reacts quickly and the 42-day one slowly, a hard block raises ATL much faster than CTL and pushes TSB negative.
A numerical example. Start with CTL = ATL = 50 and ride 70 TSS every day for 4 weeks. CTL rises to about 60, ATL reaches about 70, so TSB is around −10. That slightly negative form is typical of a productive build: you are carrying fatigue, but fitness is climbing. Now cut to 30 TSS a day for a week. CTL only slips to about 55 while ATL drops to about 44, and TSB becomes roughly +12. That is the idea of a taper: lose a little fitness, lose much more fatigue.
How to read it for a build or a taper
- Building: aim for CTL to rise gradually, often a few points per week, rather than jumping. TSB between roughly −10 and −30 is what many riders see in a hard block; the deeper you go, the more recovery you will need.
- Recovery week: a clear drop in ATL while CTL barely moves; TSB crosses zero.
- Tapering for an event: reduce volume while keeping some intensity, and look for a slightly positive TSB on the day, commonly in the range of +5 to +20. Too positive means you rested too long and lost sharpness.
- Returning after a break: CTL has decayed; ramp it up gently, because the fatigue (ATL) responds in a week while your tissues take longer to adapt.
These ranges are rules of thumb from coaching practice, not laws; individuals differ widely.
Limits of the model
- It is only as good as your FTP. If your FTP is too low, IF and TSS inflate; too high, they shrink. GearWatts uses the FTP in your profile first, then the one stored in the file, and only if neither exists it estimates 95 % of that ride's best 20 minutes. The last case makes an easy ride look easy and a hard one look harder, so set an FTP you trust.
- No power, no TSS. Rides without a power meter contribute nothing, so CTL underestimates your real load if part of your riding is unmeasured.
- Constants of 42 and 7 days are conventions. They fit the average rider; yours may recover faster or slower.
- Blind to everything outside the bike: sleep, work, illness, heat, altitude, and strength sessions do not enter the number.
- Same TSS, different stress. A 100-TSS ride made of long threshold intervals is not the same as a 100-TSS ride of easy hours.
Use the chart to see trends and to plan, and use how you feel to veto it. A rising CTL with a calm TSB is a good sign; a falling TSB you cannot explain deserves a rest day, whatever the numbers say.