On the San Fermo della Battaglia, one rider attacks and the others either respond or get dropped. The critical power model puts a name to what happens. Critical power (CP) is the power you can hold for a long time, because below it muscle, lactate and oxygen uptake stay stable. W′ is the work that can be done above CP: a reserve that is used up faster the further the power is from CP.
In brief
- Critical power sits at roughly 70-90% of VO2max, depending on how well trained a rider is.
- The lactate threshold is about 76% of critical power: they are two different boundaries.
- W′ varies enormously from person to person: from 11.4 to 33.5 kJ in 14 competitive cyclists.
- Estimating CP and W′ takes several maximal efforts; the 3-minute all-out test gives CP and W′ from a single effort.
- W′ is not a fixed parameter: the simplest model works best on long efforts just above CP.
The model has only two numbers, which is why it is the simplest way to read an attack. One says where the boundary lies, the other how far beyond it a rider can go. An attack on the Civiglio, in these terms, becomes a sum: how much reserve the attacker has, how much the chaser burns.
Still to be clarified: how the two numbers are measured, how this differs from FTP, how the reserve refills when a rider eases off, and where the model gets it wrong.
Critical power is the boundary between a steady effort and one that gives way
You ride at a given power and the body settles: oxygen uptake rises and then levels off, lactate does not climb, the muscles do not keep changing. Beyond a certain point it no longer settles. Everything keeps drifting until exhaustion. That point is CP, the boundary between stable and unstable.
In physiologists’ terms, CP separates the heavy domain from the severe domain and is the power associated with the maximal aerobic steady state. Above CP the aerobic measure alone no longer describes the effort, and this is where W′ comes in, helping to individualise intensities and durations. For the relationship between this boundary and maximal oxygen uptake, see the guide to VO2max in cycling.
In theory, if the average power of a session stayed below CP, you could carry on indefinitely. For the most commonly used effort bands, see the guide to power zones.
W′ is the reserve spent above CP
The sum
- The group: 14 competitive cyclists, with a mean CP of 253 W and a mean W′ of 21.9 kJ.
- Take W′ in joules and divide it by the distance from CP.
- At CP + 100 W: 21.900 / 100 = 219 s, about 3′39″.
- At CP + 50 W: 21.900 / 50 = 438 s, about 7′18″.
Halving the distance from CP doubles how long the reserve lasts.
Above CP the reserve falls in a straight line
W′ is measured in kilojoules: it is the work available above CP. In the 3-minute all-out test it is 16.5 ± 4.0 kJ; other studies give 18.1 ± 5.7 and 20.6 ± 7.4 kJ, and about 22 kJ in a non-fatigued subject.
Above CP the reserve falls in a straight line: remaining W′ = W′ − (P − CP)·t, where P is power and t is time. Remaining W′ is the reserve left at any instant. The further from CP, the steeper the drop.
Two cyclists with the same W′ have different CPs
The same watts-to-kilos ratio can hide different boundaries. Two cyclists with the same W′ of 13 kJ may have a CP of 375 W at 78 kg and of 305 W at 64 kg. To understand what a ratio is worth, the guide to watts per kilo compares amateurs and professionals.
W′ is also a personal figure, like CP. A group average serves to do the sum, not to say what any single cyclist has. After 2, 6 and 15 minutes of recovery, W′ came out at 8, 14 and 19 kJ.
The power-duration curve: the longer you go, the less you push
Put CP and W′ together and you get a curve. The power you can hold for a time t is P(t) = W′/t + CP. The longer the time, the more the first term thins out and the curve flattens towards CP. On short efforts the reserve weighs heavily, on long ones almost nothing.
On the Selvino climb, at Il Lombardia 2026, every climbing time corresponds to a power per kilo. The amateur in the examples, 280 W and 75 kg, climbs it in 34′17″. The professional, 335 W and 65 kg, in 26′46″. The Climb Timer gives the watts per kilo for any time.
The watts per kilo for the climb come from the physical model of Martin and colleagues (1998), which accounts for gravity, air and rolling resistance: it is a different thing from the two-parameter curve. The curve provides the logic, the calculator the comparison. The same time can also be read as VAM.
Selvino
10.1 km 5.9% average 596 m elevation gain
Enter a rider’s climbing time: the watts per kilo, with their range, and the VAM will appear here.
3.7 W/kg approx.
between 3.5 and 4.1 W/kg
- VAM
- 1,043 m/h
- Average speed
- 17.7 km/h
- Power
- about 280 W
| Power | Time | Speed |
|---|---|---|
| 5.0 W/kg | 26'50" | 22.6 km/h |
| 5.5 W/kg | 24'53" | 24.4 km/h |
| 6.0 W/kg | 23'15" | 26.1 km/h |
| 6.5 W/kg | 21'53" | 27.7 km/h |
| 7.0 W/kg | 20'43" | 29.2 km/h |
How we calculate it
Physical model by Martin and colleagues (1998): gravity, rolling resistance, air (on the rider and on the wheels) and bearing friction, at constant speed and gradient. Central assumptions: bike and equipment 8.0 kg, rolling resistance 0.0040, aerodynamic drag area 0.32 m², drivetrain efficiency 97.6%, air density 1.09 kg/m³ (average altitude 1,000 m), no wind.
The range is the result with all the favorable assumptions (drafting, a 1.5 m/s tailwind, a light bike) and with all the unfavorable ones (riding alone, a headwind, a heavier bike). These estimates are typically off by 6–10% (Millet, Tronche and Grappe, 2014): that’s why we never give a single number. A rider who surges and eases off spends a little more for the same time. VAM, the meters climbed in an hour, is a measurement instead: it depends only on elevation gain and time.
How CP and W′ are measured
You need several maximal efforts of different durations. The curve is fitted to the points obtained, and the two numbers come out of the fit.
Three to five trials between 2 and 15 minutes
The classic protocol calls for three to five trials at maximal effort between 2 and 15 minutes: the shortest of 2-5 minutes, the longest of 12-15. Below 2 minutes there is no guarantee of reaching VO2max; beyond 15, glycogen and motivation weigh in.
With at least three trials the standard error is low: 2-5% for CP and under 10% for W′. Just two trials give a perfect fit by construction (R² = 1.0) and the error cannot be calculated.
In the laboratory, constant-load trials are 3 to 7 efforts to exhaustion, over 2-7 visits. On the road, the proposal is a sprint of about 10-15 seconds and at least three efforts, with at least 30 minutes of active recovery between trials, even on two consecutive days.
The 3-minute all-out test gives everything in one trial
The all-out ramp test and the 3-minute all-out test derive CP and W′ from a single trial. In the latter, W′ is exhausted in the first 2.5 minutes or so, and the power over the last 30 seconds reflects CP.
Cadence matters: if the final cadence is about 10 rpm above the preferred one, the measured CP drops significantly. The test can overestimate CP, and with it capacity in the severe domain.
CP and FTP are not the same thing
FTP is defined as the power sustainable for one hour, but it is almost always estimated from a short test. How that is done is explained in the guide to what FTP is.
Its limitation is that it represents a single point on the curve of power against duration and has no solid physiological basis. CP describes the whole curve and estimates the boundary between a steady and a non-steady metabolic state.
The limits of agreement between FTP and CP are too wide for the two measures to be used interchangeably. In several studies FTP differs significantly from CP, maximal lactate steady state and lactate threshold.
The lactate threshold is yet another boundary, lower than CP. It is not CP. For the effort bands, see the guide to power zones.
| Measure | What it describes | How it is estimated |
|---|---|---|
| Critical power (CP) | The whole power-duration curve: the boundary between steady and non-steady state | Several maximal trials between 2 and 15 minutes, or one all-out trial |
| FTP | One point on the curve: a proxy for average power over 60 minutes | 95% of the 20-minute average, or 90% of the 8-minute average |
| Lactate threshold | A boundary different from CP, and lower; in the Sanders study: FTP 341 ± 33 W against 279 ± 20 W at lactate threshold | Dmax method, in the Sanders study |
How W′ refills when you ease off
W′ recovery is a curve
The reserve does not return at a constant rate: recovery is a curve, quick at first and then ever slower, and it runs faster the further recovery power sits below CP. In Skiba’s model the time constant is τ = 546 × e^(−0.01·ΔCP) + 316 seconds, where ΔCP is the difference between CP and recovery power. Genuinely easing off on a climb puts the reserve back in the bank quickly; staying just under CP recovers less of it.
The constant is about 377 s with recovery at 20 W. In the moderate zone it rises to about 452 s, in the heavy zone to about 580 s. If recovery power stays above CP, the constant reaches about 7056 s: a non-physiological value, meaning almost no recovery.
The half-life of W′ recovery is 234 seconds. That of VO2 kinetics is 74 seconds, that of lactate 1366: W′ sits between the two.
The recovery model has clear limits
The model has clear limits. It was built on small samples, often from 4 to 11 subjects and sometimes untrained. The constant varies widely from one person to another. And it does not account for the slowing of recovery after repeated efforts.
Attacks and surges: the W′ that is spent and refilled
Real cycling is intermittent, and the model applies there too with some assumptions. Work sits above CP, recovery below, and the session’s average power above CP. W′ expenditure is linear, recovery curvilinear. At exhaustion W′ is assumed to be spent, while CP and the time constant may change with fatigue.
Four factors decide performance: intensity and duration of the work periods, intensity and duration of the recovery periods. In one experiment, with recovery at lower power the test lasted 47% longer with heavy recovery, 100% longer with moderate recovery and 219% longer with light recovery, compared with about 384 seconds at constant power.
After exhaustion, those who went back to a power above CP kept going for about 39 seconds. Dropping below CP allowed them to exceed 10 minutes.
The model is useful for building a breakaway, for pacing a time trial, in the team pursuit.
The professionals’ numbers on the climbs of Il Lombardia 2026
At Il Lombardia 2026 the two hardest efforts are the first climb of San Fermo della Battaglia, with Pellizzari, of about 6 minutes, and the Civiglio, of about 12. They sit above threshold, so in those minutes W′ was in use. Selvino and Valpiana, efforts of about 26 minutes, have lower powers: they sit above one-hour threshold.
The Ghisallo is left out of the comparison: it was climbed fast and air resistance accounted for more than 15% of the effort. The rest is the power-duration curve seen live: a short time is worth higher powers.
For those who ride, the curve is the same, with lower values. The Climb Timer turns your own time into watts per kilo, and the guide to watts per kilo for amateurs and professionals says where you stand.
Where the model falls short
Below 2 minutes and above 40, CP gets it wrong
Outside the severe domain, CP overestimates the sustainable power, both over short times, under 2 minutes, and long ones, beyond 40.
How long CP can be held is debated: observed times range from 20-40 minutes to over an hour.
Duration also shifts the parameters. After more than 80 minutes of intense cycling CP falls, while with carbohydrates it is maintained beyond 120 minutes. W′ begins to fall after 40 minutes, with or without carbohydrates.
W′ is not a fixed parameter
In a study of 14 competitive cyclists, 9 men and 5 women, W′ was consumed by 70% and then power was dropped to CP + 20 W. They did more work than predicted: +5.4 ± 0.7 kJ and +4.3 ± 1.3 kJ in the two trials, with longer endurance of 256 ± 38 s and 240 ± 55 s. With W′ theoretically at zero, it was possible to continue by reducing the power: +2.7 ± 0.4 kJ and +109 ± 12 s.
The model seems to hold when W′ is consumed over long times at low intensity: 70% in 10 minutes at about 112% of CP produced no extra work, in 3 minutes at about 140% it did. W′ used to be seen as a fixed anaerobic capacity; today it is thought of as a cascade of fatigue from metabolic processes above CP.
When an attack on a climb comes to an end, the W′ sum helps to read it. It remains a useful sum, not a verdict.
Article written with the help of Mnemosyne.
Sources
- Jones A.M., Vanhatalo A., “The ‘Critical Power’ Concept: Applications to Sports Performance with a Focus on Intermittent High-Intensity Exercise”, Sports Medicine, 2017, PMID 28332113, doi:10.1007/s40279-017-0688-0.
- Chorley A., Lamb K.L., “The Application of Critical Power, the Work Capacity above Critical Power (W’), and its Reconstitution: A Narrative Review of Current Evidence and Implications for Cycling Training Prescription”, Sports (Basel), 2020, PMID 32899777, doi:10.3390/sports8090123.
- Leo P., Spragg J., Podlogar T., Lawley J.S., Mujika I., “Power profiling and the power-duration relationship in cycling: a narrative review”, European Journal of Applied Physiology, 2022, PMID 34708276, doi:10.1007/s00421-021-04833-y.
- Welburn A.J., Bailey S.J., Ferguson R.A., “The effect of reducing work rate on total work done and time to task failure during severe intensity exercise”, European Journal of Applied Physiology, 2026, PMID 41222680, doi:10.1007/s00421-025-06046-z.
- “Il Lombardia 2026”, the race page: the climb times and our watts per kilo.