Critical Power vs FTP: which number should you train with?

If you have spent any time in cycling forums you will have seen the argument: FTP is a coaching fudge, Critical Power is the real thing, and anyone still using FTP is behind the times. Or the reverse — CP is academic overreach and FTP works fine, thank you. Both positions contain something true and quite a lot of noise.

Here is the short version. Critical Power is a physiological boundary with a defensible meaning. FTP is a practical estimate of roughly the same region of intensity, arrived at differently. For most riders the two numbers land close together — and where they diverge, the divergence itself is informative. This post is about what each one actually is, and which of them should anchor your zones.

What Critical Power is a boundary between

The power–duration relationship has a shape everyone recognises: you can hold enormous power for a few seconds, much less for a few minutes, and the curve flattens as duration extends. What the Critical Power model says is that this curve does not flatten towards zero. It flattens towards a specific, non-zero asymptote — and that asymptote marks a real change in what your body is doing.

Below that asymptote, your physiology can find a steady state: oxygen uptake levels off, muscle phosphocreatine and blood lactate stabilise at elevated but constant values, and you can keep going for a long time. Above it, nothing settles. Oxygen uptake drifts upward towards maximum, lactate rises continuously, and exhaustion arrives predictably once the finite work capacity above CP is spent. That is a real, measurable change of state, and it is what gives CP a physiological basis that a percentage of a time trial does not have.

The intensity domains

The model divides exercise into domains with genuinely different internal responses. CP is the wall between the two that matter most for endurance training.

Where FTP came from, and why it persists

FTP arrived from the opposite direction — not from a model, but from coaches needing something they could measure on a road with a power meter. The definition is pragmatic: roughly the highest power you could sustain in a quasi-steady state for about an hour, usually estimated by riding 20 minutes flat out and taking 95% of the average.

It has held on for good reasons. It requires one effort, the arithmetic is trivial, every training platform speaks it, and its validation record against laboratory markers is respectable. In trained cyclists, FTP tracked the lactate threshold closely — though the same study found it systematically under-estimated the lactate threshold in recreational riders, with the size of the error related to fitness. That is a useful warning: FTP behaves best in the population it was developed on.

Side by side

How far apart are they in practice?

When both were measured in the same trained cyclists and triathletes, CP came out at 256 W against an FTP of 249 W — CP higher by about 7 W on average. The correlation between them was very strong. But the limits of agreement ran from roughly −19 to +33 W, which for an individual rider is a meaningful spread.

The honest reading is that they agree well as group averages and less well as individual measurements. If your CP and your FTP differ by 5 W, that is noise. If they differ by 25 W, that is telling you something about the shape of your curve — usually that you have a large W′ and your short efforts are dragging the fit, or that your 20-minute test was not fully committed.

There is also a subtlety that trips people up: CP is often sustainable for rather less than an hour — commonly cited in the 20–40 minute range — despite being defined as the boundary of steady state. This is not a contradiction. The boundary is about whether physiological variables stabilise, not about how long willpower and fuel last at that intensity.

Which should anchor your zones?

• For everything up to threshold, CP is the better anchor if you can measure it well. It marks a real domain boundary, so zones built beneath it correspond to actual physiological states rather than arbitrary percentages.
• For everything above threshold, neither number is enough on its own. Prescribing severe-domain work sensibly needs the second parameter, W′ — which is why the CP model gives you more to work with there. We cover that in W′ and anaerobic capacity.
• If your CP estimate is shaky, FTP is the safer choice. A well-executed 20-minute test beats a badly fitted curve every time. Garbage in the model is harder to spot than garbage in a time trial.
• Do not switch back and forth mid-block. Changing your anchor changes every zone boundary at once, and you will spend a month unable to tell whether the training or the arithmetic changed.

Moveee's power profile fits both from your own best efforts, so you can see the gap for yourself rather than trusting a single protocol. If the two numbers sit close, either will serve. If they sit far apart, that is worth understanding before you build a training block on top of it.