Your bike computer says your critical power is 251 watts. Run a recent 5K through a running-power calculator and it spits out 302 watts of critical speed converted to power. Treat those as the same number and your bike zones just jumped 20% for no real reason.
That’s not a rounding error. A 2023 study tested 18 trained triathletes on both bike and run. Running-derived critical power overestimated cycling critical power by 20.2%. Cycling power explained just 26.7% of the variance in running power (R² = 0.267). In short: your two thresholds live in different data sets. One doesn’t predict the other.
This is the question triathletes actually ask and rarely get a straight answer to. Does critical speed relate to FTP at all, or are they two separate numbers wearing the same “threshold” label? The research says: related, yes. Interchangeable, no.
Critical Speed vs FTP: Two Rulers for the Same Ceiling
FTP (Functional Threshold Power) is the highest power you can hold on a bike for about an hour before you blow up. Critical power (CP) is a close cousin. It’s calculated from your best efforts across several durations, usually 2 to 15 minutes. Critical speed (CS) is the running version of CP: the pace you could theoretically sustain forever if fatigue built up in a straight line.
All three describe the same idea. They mark the line between an effort you can hold and one that eats you alive. None of them measure it exactly the same way.
The model behind CP and CS looks like this:
Power (or Speed) = W' / time + CP (or CS)
Translation: below CP, you can theoretically go forever. Above it, you’re spending a fixed energy reserve called W’ on the bike or D’ on the run. Once that reserve is empty, you slow down. Hard.
The Cycling Numbers: How Close Is FTP to Critical Power?
Even within cycling alone, CP and FTP don’t match exactly. A 2021 study of 17 trained cyclists and triathletes found CP averaged 256 watts against an FTP of 249 watts. That’s a 7-watt gap, with a 91.7% probability that CP ran higher, and the two numbers still tracked tightly together at r = 0.969.
The same kind of small offset shows up in running too. A 2024 study of 15 highly trained runners found running-specific critical power exceeded running FTP by about 14 watts.
Here’s the practical read: within one sport, CP and FTP are close cousins. Test either one and you’re in the right neighborhood. Cross sports, and the neighborhood changes entirely.
That’s a 50-watt difference in the same athletes, tested in the same block. If you only have FTP, that’s close enough to set cycling zones. It tells you nothing about your run.
Running Alone: Critical Speed and Power Agree Almost Perfectly
Here’s the good news for runners. A 2023 study using Stryd running-power data tested 10 trained runners. Critical speed and critical power told the same story. Running power at critical speed measured 271 watts. Critical power from a separate protocol measured 270 watts. Statistically, no difference (p = .940), with a correlation of 0.980 between the two.
Translation: inside one sport, critical speed and running power are the same threshold wearing two units. The running-only conversion problem is basically solved. It’s the cross-sport jump that breaks.
Do Cycling and Running Thresholds Transfer? The Study That Falsifies the Shortcut
This is the study behind the chart above, and it’s worth seeing in full. Researchers tested 18 trained triathletes for critical power on the bike, using a power meter and a 3-minute all-out test. They tested the same athletes on the run, using Stryd and a 9- and 3-minute protocol.
Running CP measured 301.8 watts. Cycling CP measured 251.1 watts. Running CP overestimated cycling CP by 20.2%, and cycling power explained only 26.7% of the variation in running power.
| Comparison | Cohort | Metric A | Metric B | Gap | Agreement |
|---|---|---|---|---|---|
| Cycling CP vs FTP (Karsten 2021) | n=17 cyclists/triathletes | FTP 249W | CP 256W | +7W | r = 0.969 |
| Running CP vs FTP (Ñancupil-Andrade 2024) | n=15 trained runners | Running FTP | Running CP | +14W | not reported |
| CS vs running power (2023) | n=10 trained runners | CS-power 271W | CP 270W | ~1W, ns | ICC = 0.980 |
| Cycling CP vs running CP (MDPI 2023) | n=18 triathletes | Cycling 251.1W | Running 301.8W | +20.2% | R² = 0.267 |
| CP vs MLSS (2024 meta-analysis) | Pooled cohorts | CP | MLSS | +8-18% | High individual variance |
Same athlete. Same week of testing. Two very different threshold numbers, depending on which limb did the work.
One threshold does not fit both sports.
Two Batteries, Two Recharge Rates: W’ and D’
Above CP or CS, you’re burning through W’ on the bike and D’ on the run, your two fixed-size anaerobic reserves. They don’t behave the same, even in one athlete.
Picture two phone batteries with the same rated capacity but different chemistry. One drains fast under load and recharges slowly. The other drains steadily and refills fast. Same number on the label, different behavior in your hand.
W’ itself isn’t even tied to threshold power. Two cyclists can share an identical W’ of 13 kilojoules despite CPs of 375 watts and 305 watts, a 70-watt gap. Translation: knowing someone’s threshold tells you nothing about the size of their sprint reserve.
Recharge speed matters just as much. Research on W’ reconstitution below CP found:
| Recovery Time Below CP | W’ Reconstituted |
|---|---|
| 2 minutes | 37% |
| 6 minutes | 65% |
| 15 minutes | 86% |
For a triathlete, that’s the real cost of a hard bike leg. Drain 80% of your bike W’, and your run legs start with a run-specific D’ that’s already dented, damage a standalone run test would never catch.
Where Critical Power Actually Sits vs. What You Can Really Sustain
One more wrinkle before building a training model. CP and CS both run hot compared to maximal lactate steady state (MLSS), the lab-measured line between a truly sustainable effort and one that slowly drifts toward failure. Pooled research puts CP 8 to 18% above MLSS, with a mean gap of 12.42 watts.
Translation: “critical” doesn’t mean “forever.” A 2022 critique of the model makes the same point directly. Athletes reliably fatigue at their calculated CP well before the “indefinite” duration the name implies.
Treat CP and CS as a strong ceiling estimate. Not a promise.
A Triathlete’s Race Week: Borrowing the Wrong Number
Take a triathlete we’ll call Priya, 41, racing her second 70.3. She tested running critical speed on the track, a clean protocol that converted to 288 watts of running power. She never tested bike CP on its own. She just took the running number, subtracted a rough 5%, and set her bike interval target at 274 watts.
Her real cycling CP, tested properly two weeks later, came back at 231 watts. She’d been building bike intervals nearly 20% too hot, almost exactly the gap the cross-modality research predicts. Her bike sessions felt unsustainable, and she arrived at the run leg already drained. Once she rebuilt her zones from an actual bike test, her long-ride power held steady. Her next race, her run split off the bike improved by several minutes.
One test per sport. No shortcuts between them.
Building One Threshold Framework Across Both Sports
The fix isn’t a better conversion factor. It’s not converting at all. Cycling and running each need their own critical-power curve, built from that sport’s own data, then read on a shared scale of percent-of-threshold so you can compare intensity across disciplines honestly.
AthleteOS builds exactly that. It models critical power and critical speed independently from your actual Garmin, Strava, and Stryd history for each sport, then displays both on a normalized threshold scale instead of assuming a fixed cross-sport conversion the data doesn’t support. It also tracks W’ and D’ balance separately, so if a hard bike leg has drained 80% of your bike reserve, your run pace target adjusts for what’s actually left, not a number pulled from a stale test done weeks ago.
If your zones currently come from one test converted across two sports, that’s worth fixing before your next build. Pair this with tracking your fitness, fatigue, and form scores, watching aerobic decoupling on long efforts, and comparing TSS against TRIMP across disciplines. That combination treats swim, bike, and run as separate signals, not one shared number. Sign up for AthleteOS and connect your training history to see both curves built from your own data.