Your FTP Is Probably Wrong. Here's Why That's Okay.

You did the test. You suffered for 20 minutes, divided by 0.95, and came out the other side with a number.

That number — your Functional Threshold Power — now runs your training life. It sets your zones, governs your interval targets, determines your pacing strategy. It's the North Star of power-based cycling, the one figure coaches and apps and forum arguments orbit around.

And it's probably wrong.

Not "off by a few watts" wrong. Conceptually fuzzy wrong. A number that means genuinely different things depending on who's measuring it, what protocol they used, and what physiological landmark they think it represents.

The good news — and I mean this — is that it doesn't really matter.

A Useful Hack That Became Global Doctrine

To understand why FTP is probably wrong, you need to know where it actually came from.

In the early 2000s, exercise physiologist and competitive cyclist Andrew Coggan needed a practical way to set training zones for the power meters that were starting to appear on bikes. The physiological gold standard for "threshold" was Maximum Lactate Steady State (MLSS): the highest power you can sustain without blood lactate continuously accumulating. Pinpointing MLSS required multiple lab visits, blood draws every few minutes, and a physiologist to read the data. Completely useless for a Tuesday evening training session.

Coggan's workaround was pragmatic and clever: a well-paced, all-out 60-minute effort produces an average power very close to MLSS. The problem is that most athletes won't voluntarily suffer through a 60-minute maximal effort often enough for it to be useful. So the protocol simplified: ride flat-out for 20 minutes, take 95% of the average, call it FTP.

That 95% correction factor? Not derived from a lab. It was an observation — most riders' 20-minute maximal power runs about 5% higher than their 60-minute power. Close enough.

A field expedient became a global standard. Apps built entire architectures around it. Training plans assumed it. And somewhere along the way, we collectively forgot that FTP was never supposed to be precise. It was supposed to be useful.

What FTP Actually Measures — And What It Doesn't

This is where the science gets a bit uncomfortable.

When exercise physiologists want to find your true threshold, they don't run a 20-minute bike test. They measure one of several physiological landmarks:

Maximum Lactate Steady State (MLSS): The highest power you can sustain without lactate continuously accumulating. Multiple lab visits, serial blood draws. The actual gold standard.

Respiratory Compensation Point (RCP): During a ramp test, the point where breathing rate accelerates sharply relative to oxygen consumption — your body buffering acidosis by blowing off CO₂. No blood required, just a metabolic cart.

Critical Power (CP): Derived from multiple all-out efforts at different durations (typically 3, 5, and 12 minutes). The asymptote of the power-duration curve — the highest power you can theoretically sustain "indefinitely," which in practice means roughly 30–60 minutes before other fatigue mechanisms catch up.

Coggan's FTP is supposed to approximate MLSS. Whether it actually does depends on who you ask — and what the research says.

Strong Correlation. Wide Agreement Limits.

A 2024 Spanish study compared FTP test results from 49 amateur triathletes against their Respiratory Compensation Point measured in a lab. FTP and RCP correlated at r = 0.916 — genuinely strong. If you're a researcher studying groups of athletes, FTP is a reasonable proxy for RCP and MLSS. The population-level relationship holds.

But the researchers also ran a Bland-Altman analysis — the statistical test that actually matters for individuals rather than populations. The limits of agreement between FTP and RCP were wide enough that for any given athlete, your FTP result could diverge meaningfully from your actual physiological threshold. Sometimes high. Sometimes low. The error bars are large enough that building precise training zones on a single FTP number is, statistically speaking, building on a well-educated guess.

Source: "Equivalence between FTP and RCP: Assessment of association, agreement and sex effect." PMID: 42060964, 2024.

Critical Power and FTP Aren't the Same Thing

A 2021 study asked this directly: do Critical Power and Functional Threshold Power actually match up in highly-trained athletes?

They don't.

Researchers tested highly-trained cyclists and triathletes using three different Critical Power models — hyperbolic, linear, and 3-minute all-out — and compared the results to each athlete's FTP from a standard 20-minute test. The different CP models didn't even agree with each other, let alone with FTP.

This matters for more than pedantic reasons. Critical Power has a stronger theoretical foundation than FTP. CP marks a genuine physiological boundary — the line between the "heavy" and "severe" exercise intensity domains. Below CP, you can theoretically sustain effort for a very long time, limited by fuel and cumulative fatigue rather than an inexorable metabolic spiral. Above CP, you're on a countdown. Your W' — your anaerobic work capacity above critical power — is being depleted, and when it hits zero, you're done.

FTP sits near this boundary. It correlates with it. But it doesn't map onto it cleanly, and treating them as interchangeable leads to miscalibrated zones.

Source: "Do Critical and Functional Threshold Powers Equate in Highly-Trained Athletes?" PMID: 34055164, 2021.

Even 55% vs. 85% of FTP Isn't Consistent

A 2025 UK study had 12 healthy males cycle for 30 minutes at 55% and 85% of their FTP, measuring oxygen consumption, heart rate, blood lactate, and RPE at baseline, 15, and 30 minutes.

Physiological and subjective responses differed significantly both between and within exercise intensities. Even when you perfectly control for FTP percentage, the same relative intensity produces meaningfully different internal loads in different people — and the divergence grows over time within a single session as cardiovascular drift and fatigue accumulate.

Your training zones are approximate guidelines. They were never precision instruments.

Source: "Physiological and subjective measures of exercise intensity cycling at 55% and 85% of functional threshold power." PMID: 42148565, 2025.

The Metric FTP Can't See: Durability

The most important thing FTP won't tell you is whether you can still produce that power after three hours of racing.

A 2025 study examining Tour de France GC contenders found something that should probably reshape how most cyclists think about training. When comparing Top 5 finishers to Top 15 finishers across multiple Grand Tours, there were almost no meaningful differences in peak power output, time in zones, or TSS accumulated. The one statistically significant difference? Top 5 riders spent more time in Zone 1 — at or below 55% FTP.

The riders who finished on the podium didn't produce more power. They didn't spend more time at threshold. They went easier when it was easy — preserving their legs for km 180 when the race actually broke apart.

This is durability: the capacity to resist fatigue and maintain power output late in a long effort. It's not measured by an FTP test. It's barely measured by any consumer cycling metric currently on the market.

Source: "What Does It Take to Reach the Podium? Power Output and Heart Rate-Derived Racing Demands of Top Cyclists During Grand Tours." PMID: 40375449, 2025.

A separate 2024 paper made the case that durability should be treated as an independent parameter of endurance performance — distinct from FTP, VO2max, or efficiency. Two athletes with identical FTP and identical W/kg can have dramatically different race results if one's power degrades 3% after four hours and the other's degrades 15%. A fresh 20-minute test captures none of that.

Source: "Durability as an independent parameter of endurance performance in cycling." PMID: 40375449, 2025.

What Machine Learning Is Starting to Catch

A 2024 European paper applied machine learning to cardiovascular drift data — the gradual heart rate rise during sustained exercise at fixed power output. The goal was to detect when an athlete's actual fitness level had changed, without requiring a formal test.

The model could identify genuine fitness shifts from patterns in how heart rate drifted during ordinary training sessions. An athlete whose FTP had genuinely increased showed a different CV drift signature than someone having a good day or testing under favorable conditions.

The implication: FTP doesn't have to be something you test for every 4–6 weeks. It can be something software infers from your daily training data — a continuously updated estimate rather than a number you defend until next month's sufferfest.

Source: "Quantifying training response in cycling based on cardiovascular drift using machine learning." PMID: 40687435, 2024.

A separate 2024 proof-of-concept showed that Functional Data Analysis (FDA) — evaluating the entire shape of a physiological response curve rather than a single summary point — outperforms single-metric approaches for characterizing what an athlete can actually do. The future of training metrics isn't a more accurate threshold number. It's abandoning the obsession with single numbers altogether.

Source: "Beyond Static Assessment: A Proof-of-Concept Evaluation of Functional Data Analysis for Assessing Physiological Responses." PMID: 42029519, 2024.

So What Do You Actually Do With This?

If FTP is imprecise, if CP and RCP don't agree with it, if durability matters more than a fresh 20-minute number, and if the whole concept began as a pragmatic workaround — what should you actually do with your FTP?

Treat it as a range, not a point. Your real threshold isn't exactly 312 watts. It's somewhere in the neighborhood of 300–325. Training zones built on 312W will get you close enough, and the 5-watt precision your head unit displays is mostly an illusion. A workout targeting 105% of FTP will be hard whether your actual FTP is 300 or 315. A recovery ride at 50% FTP will be easy either way. The zones are broad enough that the errors mostly wash out.

Take RPE seriously — it's not a backup metric. The 2025 FTP intensity study found that Rating of Perceived Exertion tracked physiological strain as well as any power metric. If a workout feels harder than the numbers suggest, it is harder. Full stop. The cyclists who improve most consistently tend to be the ones who calibrate effort by feel first and use power as confirmation, not the other way around.

Retest less often, but watch your trends daily. The standard "retest every 4–6 weeks" advice burns real physical and mental energy chasing a number that fluctuates with sleep, nutrition, motivation, and test conditions. A more useful approach: retest when you have genuine reason to believe your fitness has shifted. Every 8–12 weeks works for most riders. Between tests, watch what's actually happening — are your intervals getting easier at the same power? Is your heart rate lower for the same effort? Are you recovering faster? These signals are more useful than a monthly 20-minute time trial.

Track durability separately. If you race or do long events, your ability to hold power after 3–4 hours probably matters more than your fresh FTP. Start monitoring your power output in the final hour of long rides and compare it to the first hour. That degradation percentage tells you something no 20-minute test can.

Remember what FTP was designed to do. It was invented to set training zones. It does that job well enough. It was never meant to predict race performance, compare your fitness to someone else's, or serve as a number for your Instagram bio. Use it to anchor your workouts. Don't use it to anchor your identity.

How Cadence Thinks About This

We built Cadence partly because most training apps treat FTP like a fixed truth you serve, rather than a rough anchor that serves you.

Every workout in Cadence's 600+ library is structured around your FTP — but the zones are ranges, not tight prescriptions. Your 85% FTP interval works whether your real threshold is a few watts higher or lower than the test said.

The dashboard proposes workouts based on your current freshness (TSB), recent training load, and goals — but you override it whenever you want. The number suggests. You decide. Cadence doesn't claim to know your body better than you do; it just tries to make the data less annoying to interpret.

Your training history is yours to export, back up, and analyze however you like. We don't lock it behind a subscription wall. And FTP sits alongside CTL, ATL, TSB, ramp rate, W/kg, and recent rides on one screen — because fitness isn't a single data point. It's a pattern that develops over months.

The Bottom Line

Your FTP is probably wrong in the way a weather forecast is wrong — close enough to pack the right jacket, not precise enough to leave the umbrella at home. That's fine. That's always been fine.

The cyclists who improve consistently aren't the ones with the most accurate FTP. They're the ones who train regularly, pay attention to how they feel, and use the number as an anchor rather than a verdict.

Trust the trend. Trust the feel. The number is just a starting point.

Cadence is the indoor cycling app you buy once and own for life. Structured workouts for your smart trainer, built to raise your FTP. One month free, then €49.99 once — no subscription, no AI babysitter, just watts.