FTP without a test: how to estimate threshold power from data you already have

What FTP actually is (and what it isn't)

Functional Threshold Power was defined by Andrew Coggan as the highest power a rider can sustain in a quasi-steady state for approximately one hour [Allen et al. 2019]. The 20-minute test estimating it as 95% of average power is a protocol, not the definition itself. The distinction matters because the protocol fails for a meaningful share of riders.

The original concept dates to the early 2000s. Coggan and Hunter Allen formalized FTP in Training and Racing with a Power Meter [Allen et al. 2019] as the upper boundary between the heavy and severe intensity domains — the wattage you can hold for roughly an hour before fatigue forces you off. Physiologically, it sits near maximal lactate steady state (MLSS), the highest power at which blood lactate stabilizes rather than climbing [Karsten et al. 2018]. It is not the same number as MLSS, but it is in the same neighborhood for most trained cyclists.

The 20-minute test came later, as a workaround. A genuine 60-minute time trial is brutal and rare, so the protocol multiplies 20-minute power by 0.95 to estimate it. Group-level studies show this works well on average — Karsten and colleagues [Karsten et al. 2018] reported a bias of 1.4% versus MLSS with a near-perfect correlation across trained cyclists. But Borszcz and colleagues [Borszcz et al. 2018] showed that on an individual basis, the limits of agreement between the 20-minute estimate and a real 60-minute effort spanned 40-60 watts. For a 250-watt rider, that is the difference between a productive Tuesday session and a disaster.

What FTP isn't: a fixed physiological constant that only one test reveals. It is an estimated parameter of the power-duration curve, and like any estimate it has confidence intervals. Treating it as a single sacred number — the way most apps' onboarding flows do — bakes false precision into every workout that follows.

Why ramp tests and 20-minute tests are flawed (and when they are useful)

Ramp tests measure maximal aerobic power and approximate FTP at 75% of MAP — a relationship that varies between 72% and 77% across riders. The 20-minute test is highly pacing-dependent and warm-up-dependent. Both produce a number; neither produces ground truth. They are useful as periodic reality checks, dangerous as the only input to a training plan.

The ramp test's central assumption is that FTP equals 75% of the peak one-minute power achieved at the top of an incremental ramp [Allen et al. 2019]. The 75% figure is an average — published validation work places the actual ratio between 72% and 77% depending on the rider's anaerobic capacity. A pure sprinter with a high MAP and a modest aerobic ceiling gets a flattering ramp number that overstates their true FTP by 8-15 watts. A diesel-engine rider with a low MAP relative to their threshold gets the opposite. TrainerRoad's ramp test is the most-used FTP protocol in indoor cycling and it is honest about being a starting estimate — the criticism is the apps that consume the number as if it were the truth.

The 20-minute test is worse than its reputation. Borszcz and colleagues [Borszcz et al. 2022] showed that warm-up structure alone shifts the resulting FTP estimate by clinically meaningful amounts. Pacing matters — a rider who blows up at minute 14 produces a falsely low number, and a rider who paces conservatively the entire 20 minutes produces a falsely low number for the opposite reason. The test demands not just fitness but pacing experience, which most amateurs do not have.

When tests are useful: as a periodic reality check on whatever model you are using. A power-curve-derived FTP estimate that disagrees with a clean 20-minute effort by more than 5% means one of the two is wrong, and the field test is usually the tiebreaker. Once a season — typically at the end of base, before the build phase — running an actual 20-minute or 8-minute test [Allen et al. 2019] is good hygiene. Doing it every six weeks because an app demands it is mostly a recovery tax.

The Critical Power model: where modeled FTP estimates come from

The Critical Power model is the mathematical backbone of every modern FTP-without-a-test approach. Hugues Monod and Jacques Scherrer first formalized it in 1965 [Monod & Scherrer 1965]; David Poole, Andrew Jones and colleagues canonized it for endurance science in 2016 [Poole et al. 2016]. Two parameters describe almost all of your sustainable power output: Critical Power and W'.

The model is simple in form. For any duration above roughly two minutes, the highest power you can sustain follows a hyperbolic curve with two parameters: Critical Power (CP), the asymptote, and W' (W-prime), a fixed amount of work in kilojoules you can do above CP before you fail [Poole et al. 2016, Burnley & Jones 2018]. Below CP, physiological responses stabilize and you can ride for hours. Above CP, lactate, phosphocreatine and pH spiral until fatigue ends the effort. The threshold between those two domains is what FTP approximates.

What makes CP useful for FTP-without-a-test is that the model can be fit from any three or more maximal efforts of different durations — a 5-minute climb, a 12-minute time trial, a 20-minute KOM attempt — without forcing you to do all of them on the same day. Vanhatalo, Doust and Burnley [Vanhatalo et al. 2007] validated a single 3-minute all-out test that estimates CP within 6 watts of conventionally-measured values. Software with months of your ride history can do better than that, because it has dozens of near-maximal efforts of varying duration to fit against rather than relying on one acute test.

The relationship between CP and FTP is close but not identical. Critical Power tends to run a few percent higher than 60-minute power for highly-trained athletes, and the offset is athlete-specific [MacInnis et al. 2021]. Practical apps account for this by reporting an FTP that is CP scaled down by a small individual factor, calibrated against the rider's own time-trial efforts when available. The core insight is that the data needed to estimate FTP already exists in any rider's power file — it just has to be parsed correctly.

How AdaptCycling estimates FTP from your Strava history

We read your last 90-180 days of Strava activities at sign-in, build your mean-maximal-power curve from every ride, and fit a Critical Power model to it. The output is an FTP estimate accurate to within roughly 3-5% for most riders — usable on day one with no forced test. The estimate updates as new rides come in.

The pipeline is straightforward. On connection, we pull every activity with power data from your Strava history. From each ride, we extract the mean-maximal power for durations from 5 seconds to 60 minutes — the same MMP curve Allen and Coggan describe [Allen et al. 2019] and that GoldenCheetah, Intervals.icu and TrainingPeaks have all surfaced for years. With dozens to hundreds of rides feeding the curve, the longer-duration points are statistically reliable rather than dependent on whether you happened to feel good last Tuesday.

We then fit the curve to a Critical Power model and derive FTP from CP, applying a small individualizing offset based on your longest sustained efforts. The result skips the test entirely. Riders who have been training with power for at least three months get a usable FTP within seconds of connecting Strava. Riders with thinner histories get a more conservative estimate that updates as they accumulate quality efforts.

Two things matter for accuracy. First, you need at least one near-maximal effort in the longer-duration range — a 12-minute climb, a 20-minute time trial, a hard 30-minute solo segment. Without that, the curve is anchored only by short bursts and over-estimates FTP. Second, the data has to be clean — power-meter spikes, indoor sessions with broken calibration, and ERG-mode workouts at suppressed power all distort the curve. We filter the obvious outliers; the rest is on the rider's own data hygiene.

How Xert, Intervals.icu, and TrainingPeaks each derive FTP without a test

Three competitors have credible no-test FTP approaches. Xert builds a three-parameter Fitness Signature from breakthrough efforts. Intervals.icu fits a Critical Power model to your MMP curve and reports an eFTP. TrainingPeaks Auto-FTP detects a recent best 20-minute power and computes 95% of it. Each method has tradeoffs worth understanding before you choose a primary tool.

Xert's model is the most ambitious. Their Fitness Signature combines Threshold Power, High Intensity Energy (their W' analogue) and Peak Power into a three-parameter description of your power-duration curve [Xert MPA]. The system updates from breakthrough efforts — rides where you push your limits — and predicts your real-time Maximum Power Available second by second. The strength is the granularity. The weakness is that breakthrough rides are required for the model to update; passive zone-2 weeks leave the signature drifting.

Intervals.icu is the free standard. The platform fits a CP model against your MMP curve, reports an eFTP based on the best fit, and lets you set the model parameters manually if you disagree [Intervals.icu]. The eFTP value updates whenever a new ride extends or lifts the curve. For self-coached riders who already use Intervals.icu for analytics, eFTP is essentially free FTP estimation that runs as a background process.

TrainingPeaks Auto-FTP is more conservative. The system watches for any 20-minute effort that exceeds your previous best, computes 95% of it, and proposes that as a new FTP. It will not estimate from short or long efforts, only from a clean 20-minute peak. That makes it the most predictable but also the least proactive — riders who do not produce 20-minute peak efforts in normal training will see Auto-FTP go stale for months.

The honest comparison: Xert is the most sophisticated; Intervals.icu is the best free option; TrainingPeaks Auto-FTP is the safest if you distrust modeled estimates and just want a small bump when you actually go faster. AdaptCycling's approach is closest to Intervals.icu — CP-derived, continuously updated, no test required — with the difference being we use the FTP estimate to drive an adaptive plan rather than just displaying it on a dashboard.

When you actually do need a test (and which test to do)

You need an actual FTP test in three situations: at the start of a structured block when you have less than 90 days of power data, after a long layoff that invalidates your MMP curve, and as a sanity check before a goal event when the modeled FTP feels off. Outside those three, modeled FTP is sufficient for nearly all amateur self-coached training.

The first situation is the most common. A new rider, or a rider who has just bought their first power meter, has nothing for the model to fit against. Two weeks of zone 2 and one ramp test produces a usable starting FTP within a 5% margin of the truth, and the modeled estimates take over once the rider has accumulated 60-90 days of varied riding. Skipping the early test in favor of pure modeling means accepting wider error bars for the first month.

The second is the long-layoff case. After 6+ weeks off — illness, injury, life crisis — the MMP curve no longer reflects current fitness. Mujika and colleagues [Mujika & Padilla 2000] showed that VO2max and threshold-related performance markers begin meaningful decline after roughly 10 days of complete cessation, and the rate accelerates after three weeks. Coming back from a long break, a fresh test re-anchors the model and prevents the first six weeks of the new block from being run at the wrong intensity.

The third is the pre-goal sanity check. Two weeks before a target event is a good time to do a clean 20-minute or 8-minute effort [Allen et al. 2019] and compare it to the modeled FTP. If the two agree within 5%, you taper with confidence. If they disagree by more, you have time to investigate before the event rather than discovering on race day that the plan was built against a number 15 watts off the truth.

The protocol that holds up best in research is the 3-minute all-out test [Vanhatalo et al. 2007] for direct CP estimation, paired with a separate 20-minute effort for FTP cross-check. It is unpleasant but short. For most amateurs the friction-free alternative is the standard 20-minute test — well-warmed-up, well-paced, ideally outdoors on a known climb. Whatever you do, do it twice a year, not every six weeks.

Quick answers

Sources cited in this guide

1. 01
   Allen et al. 2019. Training and Racing with a Power Meter (3rd ed.). VeloPress.
2. 02
   Borszcz et al. 2018. Functional Threshold Power in Cyclists: Validity of the Concept and Physiological Responses. International Journal of Sports Medicine.
3. 03
   Borszcz et al. 2022. Functional Threshold Power Estimated from a 20-minute Time-trial Test is Warm-up-dependent. International Journal of Sports Physiology and Performance.
4. 04
   Karsten et al. 2018. Is the Functional Threshold Power Interchangeable With the Maximal Lactate Steady State in Trained Cyclists?. International Journal of Sports Physiology and Performance.
5. 05
   Monod & Scherrer 1965. The Work Capacity of a Synergic Muscular Group. Ergonomics.
6. 06
   Poole et al. 2016. Critical Power: An Important Fatigue Threshold in Exercise Physiology. Medicine & Science in Sports & Exercise.
7. 07
   Burnley & Jones 2018. Power-duration relationship: Physiology, fatigue, and the limits of human performance. European Journal of Sport Science.
8. 08
   Vanhatalo et al. 2007. Determination of Critical Power Using a 3-min All-out Cycling Test. Medicine & Science in Sports & Exercise.
9. 09
   MacInnis et al. 2021. Do Critical and Functional Threshold Powers Equate in Highly-Trained Athletes?. International Journal of Sports Physiology and Performance.
10. 10
    Mujika & Padilla 2000. Detraining: Loss of Training-Induced Physiological and Performance Adaptations. Part I: Short Term Insufficient Training Stimulus. Sports Medicine.
11. 11
    Xert MPA. Maximal Power Available (MPA) — Methodology. Xert / Baron Biosystems.
12. 12
    Intervals.icu. Power and eFTP — Intervals.icu Documentation. Intervals.icu.