FTP without a test

How often should you test your FTP? Twice a year, not every six weeks

Most training apps nag you to retest FTP every four to six weeks. For a formal field test, that cadence is roughly three times too frequent. The honest answer: run an actual 20-minute or ramp test about twice a year — once at the end of base before the build phase, and once as a sanity check two weeks out from a goal event [Allen et al. 2019]. Test off that schedule only when something specific invalidates your data. Here is the decision framework, with numbers.

By Jim Camut · Former pro & ex-Bruyneel Academy racer

Updated Jul 17, 20264 chapters7 citations

01 / 04

The honest cadence: about twice a year for a formal test

For a dedicated field test, twice a season is enough: end of base before the build phase, and a pre-goal check about two weeks out [Allen et al. 2019]. Threshold fitness moves over weeks to months, not days, and a test's own error can exceed the change you are trying to detect in six weeks.

The two-test logic is about anchoring, not measuring for its own sake. The end-of-base test resets your training zones before the hardest block of the year, when riding at the wrong intensity is most expensive; a 10-watt error on a 250-watt rider drags every threshold interval off target. The pre-goal test, run 10 to 14 days out [Allen et al. 2019], confirms the number your taper is built on. Two well-placed tests do more for a season than eight scattered ones.

Testing more often mostly measures noise. Borszcz and colleagues [Borszcz et al. 2018] found that the individual limits of agreement between a 20-minute FTP estimate and a true 60-minute effort spanned 40 to 60 watts. A realistic six-week block might raise threshold by 2 to 5 percent — 5 to 12 watts for that same 250-watt rider. When the test's own error band is wider than the fitness change you are chasing, a monthly retest is measuring your warm-up and pacing, not your engine.

There is also a recovery cost. A maximal 20-minute or ramp test is a genuinely hard session that needs a day of freshness going in and 24 to 48 hours to clear afterward. On a six-week cadence that is roughly eight tests a year, each one displacing a quality workout and each one vulnerable to a false low — Borszcz and colleagues [Borszcz et al. 2022] showed that warm-up structure alone shifts the resulting FTP by clinically meaningful amounts. Test under-recovered and you stamp a deflated number onto the next block.

02 / 04

Why apps push a four-to-six-week retest

The four-to-six-week cadence is an artifact of software design, not physiology. Static plan generators consume FTP as one fixed input, so the only way to keep your zones honest is to re-measure on a timer that matches a training block. The cadence serves the tool's need for a fresh number, and pushes the cost onto you.

Consider what a conventional plan generator actually needs. It multiplies a single FTP value by fixed percentages to set every zone and every workout target. If that number drifts as you get fitter, the whole plan quietly goes stale, so the software schedules a retest every mesocycle — typically four to six weeks — to refresh the input. The logic is sound for the tool. It just assumes your only source of truth is a dedicated test.

The cost lands on the athlete. At a six-week cadence you run roughly eight tests a year; each is a hard day you did not spend training, plus the taper and recovery around it. Worse, any test caught on an under-recovered day reads low, and a number that is falsely low by 15 watts then poisons every zone for the following block — your sweet-spot work drifts toward tempo, and the plan silently under-doses you for weeks.

03 / 04

When to test off-cadence: the three real triggers

Ignore the calendar and test when a specific event invalidates your data. Three triggers justify an off-schedule effort: a layoff long enough to start detraining, a suspected real change larger than about 5 percent, or new equipment. Each one breaks the assumption that your recent power history still reflects who you are today.

A layoff is the clearest trigger, because detraining is fast at first. Coyle and colleagues [Coyle et al. 1984] tracked highly trained subjects after they stopped: VO2max fell 7 percent in the first 21 days and settled 16 percent below trained values by 56 days, with muscle oxidative enzyme activity falling on a roughly 12-day half-life. Mujika and Padilla [Mujika & Padilla 2000 Part I] attribute the earliest losses — inside the first two to four weeks — largely to a drop in blood volume. Practically: after about two weeks fully off, your zones have started to drift, and that is the point to consider a re-anchor.

Longer breaks reset the curve outright. Reviewing stoppage beyond four weeks [Mujika & Padilla 2000 Part II], Mujika and Padilla report that recently acquired VO2max gains are lost completely and lactate threshold falls, even though a long-term athlete stays above untrained values. After six weeks or more off — illness, injury, a life crisis — your old FTP is fiction, and any model built on those stale rides is anchored to a rider who no longer exists. Retest before the first hard block back, not three weeks into it.

The second trigger is a suspected real jump. A stable FTP should not move more than 1 to 2 percent on the strength of any single ride, so if a benchmark climb suddenly comes in 5 percent higher, or a continuously modeled estimate steps up by more than 5 percent, it is worth a clean confirmation effort. A change that large is either genuine adaptation or a data artifact — a power spike, a mis-set weight, a tailwind PR — and a controlled test is the tiebreaker.

The third is new equipment. Power meters routinely disagree by 3 to 5 percent, so a new meter, a new trainer, or a swap between crank- and pedal-based measurement makes your entire historical power curve non-comparable overnight. A wattage that was FTP last week may read 10 watts high or low on the new device. The first week or two on new hardware needs a benchmark effort to rescale everything that follows; without it, every zone inherits the offset.

04 / 04

The modern reframe: cross-check, not re-measure

Continuous modeling changes the question entirely. When FTP is estimated from your power curve on every ride, the number updates as a background process, so cadence stops being how often you re-measure and becomes how often you cross-check the model against a real effort. For most riders that is one or two efforts a year, not eight.

This is the practical payoff of the idea at the center of the parent guide — that FTP can be estimated from the ride data you already have, with no forced test. Once the estimate refreshes continuously, a scheduled monthly retest is redundant by construction: the model has already moved. The only thing a dedicated effort still buys you is verification, so the whole question collapses from 'how often should I re-measure' to 'how often should I cross-check.' The honest cross-check cadence is the same twice-a-year rhythm from the top of this article — end of base and pre-goal — except now you are confirming a number rather than generating one.

AdaptCycling works this way. We read your Strava history, fit a Critical Power model to your mean-maximal-power curve, and re-derive FTP after every ride, applying a small per-rider offset because Critical Power runs measurably above 60-minute power — McGrath and colleagues [McGrath et al. 2021] measured a 16-watt gap (282 versus 266 watts) in highly trained cyclists. Because the number stays current, we ask for a real test only at the two moments it earns its keep: re-anchoring after a long layoff, when the curve is built on stale rides, and a pre-goal sanity check. Cross-check, don't re-measure.

Common questions

Quick answers

How often should I actually test my FTP?

For a dedicated field test, about twice a year: once at the end of your base phase before the build block, and once as a sanity check 10 to 14 days before a goal event [Allen et al. 2019]. Test outside that rhythm only when a layoff, a suspected change over 5 percent, or new equipment invalidates your recent data. The every-four-to-six-weeks cadence most apps push is roughly three times more often than the physiology requires.

Why does my app tell me to retest every four to six weeks?

Because a static plan generator uses FTP as a single fixed input and has no other way to notice you have gotten fitter — it must re-measure on a timer to keep your zones current, and four to six weeks matches a typical training block. The cadence serves the software, not you. Each test costs a hard day plus recovery, roughly eight a year, and any test caught under-recovered stamps a low number onto the next block [Borszcz et al. 2022].

How long off the bike before I need to retest?

About two weeks fully off is when zones start to drift — VO2max drops around 7 percent in the first three weeks of no training [Coyle et al. 1984], with the earliest losses driven by falling blood volume [Mujika & Padilla 2000 Part I]. After six weeks or more off, recently gained fitness is largely gone [Mujika & Padilla 2000 Part II] and your old FTP is fiction; retest before the first hard block back rather than training weeks at the wrong intensity.

My FTP jumped 15 watts after one hard ride — is that real?

Probably not from one ride. A stable FTP should not move more than 1 to 2 percent on the strength of a single effort, so a sudden 15-watt jump — roughly 6 percent on a 250-watt rider — is more often a data artifact than an overnight adaptation: a power spike, a tailwind, a mis-entered weight. Treat it as a flag to cross-check with a clean effort, not as a new number to train against immediately.

If my FTP updates automatically, do I still need to test at all?

Only to cross-check, and only once or twice a year. A continuously modeled FTP already tracks real change within days, so a dedicated test stops being a measurement and becomes verification. The two moments still worth a real effort are re-anchoring after a long layoff, when the model is fit to stale rides, and a pre-goal check so you taper against a number you have confirmed [Allen et al. 2019].

References

Sources cited in this guide

01
Allen et al. 2019. Training and Racing with a Power Meter (3rd ed.). VeloPress.
02
Borszcz et al. 2018. Functional Threshold Power in Cyclists: Validity of the Concept and Physiological Responses. International Journal of Sports Medicine.
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.
04
Mujika & Padilla 2000 Part I. Detraining: Loss of Training-Induced Physiological and Performance Adaptations. Part I. Sports Medicine.
05
Mujika & Padilla 2000 Part II. Detraining: Loss of Training-Induced Physiological and Performance Adaptations. Part II. Sports Medicine.
06
Coyle et al. 1984. Time Course of Loss of Adaptations After Stopping Prolonged Intense Endurance Training. Journal of Applied Physiology.
07
McGrath et al. 2021. Do Critical and Functional Threshold Powers Equate in Highly-Trained Athletes?. International Journal of Exercise Science.

Compare head-to-head

See the direct comparisons

In this series

More inside FTP without a test

Start here · Foundational guide

FTP without a test: estimating threshold from real rides

How to find FTP without a 20-minute or ramp test — using your power curve, critical-power modeling, and the rides you've already done.

Read the full guide

Try the adaptive coach yourself.

Connect Strava and the coach reads your last eight weeks — intensity balance, ramp rate, what's working, what's holding you back — then drafts your training plan around it.

Analyze my training→

Free 14-day trial, no card.