Why your ramp test FTP is too high (and how to correct it)

Why the ramp test overshoots your real threshold

The ramp test never measures your hour power directly. It measures the peak one-minute power you reach as the load climbs, then multiplies by 0.75 [Allen et al. 2019]. Because that final minute is partly fueled by a finite anaerobic reserve, riders with a big reserve buy a high peak — and inherit an inflated FTP.

The power-duration model explains the leak. Any effort above your critical power draws down W', a fixed quantity of work — measured in kilojoules — that you can spend above threshold before you fail [Burnley & Jones 2018, Vanhatalo et al. 2007]. The last 60 seconds of a ramp sit far above critical power, so they are paid for almost entirely out of W'. A rider with a 25 kJ anaerobic reserve reaches a higher final minute than a rider with 15 kJ at the identical aerobic ceiling, even though both hold the same sustainable power. The ramp cannot tell those two riders apart — it reads only the peak, and the peak is inflated by anaerobic capacity.

This is why peak power output and sustainable power are different physiological measures, not two views of one number. Bentley and colleagues [Bentley et al. 2001] found that the peak workload from an incremental test and the power a cyclist actually held over a 90-minute time trial were correlated but distinct, tracking separate determinants across their nine well-trained subjects. The ramp collapses that distinction into a single 75% multiplier. For most of the field the collapse is harmless, because most riders cluster near the average ratio — but the further your anaerobic capacity sits from the mean, the more the multiplier lies.

Worse, the peak minute is not even a stable number. Michalik and colleagues [Michalik et al. 2019] ran the same men through two ramp rates and found peak power was 51.69 watts higher — roughly 15% — on the faster ramp, because a steeper climb reaches exhaustion before the aerobic system is fully taxed. TrainerRoad's and Zwift's ramps use fixed step sizes, so your result is partly an artifact of the protocol you happened to ride. The ramp is honest as a starting estimate; the error enters when an app treats that estimate as ground truth and prescribes threshold work against it.

The rider types the ramp flatters (and the ones it shortchanges)

The 75% ratio is a population average. Validation and field data place the real ratio near 72% to 77% depending on anaerobic capacity [Allen et al. 2019]. Punchy sprinters and criterium riders land at the low end and get inflated FTPs; aerobic diesels land at the high end and get numbers that undersell them.

If you sprint well, close crits from a small group, or win the last-100-meters drag race, you are the rider the ramp flatters most. Your high maximal aerobic power and large anaerobic reserve inflate the final minute, and 75% of an inflated peak overstates your true threshold — often by 8 to 15 watts on a 250-watt rider, roughly a 3 to 6% error. That is exactly enough to turn a prescribed 4x10 at 'sweet spot' into an over-threshold session you cannot finish.

The mirror image is the aerobic diesel — the rider who time-trials well but gets shelled the instant the pace turns explosive. Your anaerobic reserve is modest, so your ramp peak is comparatively low, and 75% of a low peak understates a threshold you can genuinely hold. Karsten and colleagues [Borszcz et al. 2019] showed FTP sits close to maximal lactate steady state, the aerobic, sustainable boundary — precisely the quality a diesel is built around and the ramp fails to reward. These riders are frequently stronger at threshold than their ramp FTP claims, and correcting upward is as real a fix as correcting down.

The scatter is not small. Borszcz and colleagues [Borszcz et al. 2018] reported that on an individual basis the agreement between a short FTP estimate and a genuine sustained effort spanned 40 to 60 watts across their cyclists, even though the group averages matched almost perfectly. A number that is accurate for the population mean and 40 watts off for you is worthless for prescribing your Tuesday intervals. The ramp's convenience is real; its per-rider precision is the part the marketing quietly omits.

The tell: workouts that keep blowing up at your ramp FTP

You do not need a lab to confirm an inflated FTP — your training log already holds the evidence. If threshold and sweet-spot sessions fail repeatedly at the prescribed watts while your easy rides feel normal, the FTP is too high, not your discipline. A pattern beats any single bad day.

The clean signal is a repeated failure at one specific intensity. Two or three sweet-spot workouts — say 3x15 at 88 to 94% of FTP — that you abandon in the final interval, week after week, point at the anchor, not the legs. One failed session is noise: poor sleep, heat, under-fueling. Three failures at the same relative intensity with adequate recovery between them is signal. The giveaway is where the failure lands — inflated-FTP riders die in the last third of threshold work, when the aerobic system is finally asked to hold a wattage it never actually owned.

Notice where the problem does not appear. Zone 2 and endurance rides feel entirely normal, because they sit well below the inflated number and the error never bites. Short VO2max intervals may also feel fine — 30-second and 2-minute efforts are paid for by the same anaerobic reserve that inflated your ramp in the first place. The collapse is specific to the 88 to 105% band, the sustained region where FTP is supposed to live [Borszcz et al. 2019]. If your log shows strength up top, comfort down low, and failure in the middle, that hourglass shape is the fingerprint of an FTP set too high.

How to correct an inflated ramp FTP

Four fixes, cheapest first: shave the multiplier to fit your rider type, cross-check with one sustained effort, re-anchor on a curve that pools many efforts instead of one, or let software fit your whole power history. Each moves you from a population guess toward your own sustainable power.

The fastest correction costs nothing. If you are a proven sprinter or crit rider, drop the ramp multiplier from 75% toward 72% and re-run your workouts against the lower number — on a 250-watt ramp FTP that is a roughly 8-watt haircut, usually enough to make sweet-spot work completable again. Then cross-check with one clean sustained effort: a well-paced 20-minute test, or better, a genuine 30-to-40-minute solo time trial gives you real sustained data the ramp never collected [Allen et al. 2019]. Where the sustained effort and the ramp disagree by more than 5%, trust the longer effort.

The deeper fix is to stop relying on any single test. The power-duration or critical-power model fits your threshold from several maximal efforts of different lengths — a 5-minute climb, a 12-minute TT, a 20-minute segment — so no one anaerobically-loaded minute can dominate the estimate [Vanhatalo et al. 2007, Burnley & Jones 2018]. This is the frame of our pillar on estimating FTP without a forced test: your threshold already lives in the data you have ridden, and the ramp is just one flawed test the curve-based approach sidesteps. A curve fit to dozens of real efforts is far harder to fool than 75% of a single peak minute.

This is what AdaptCycling does with your Strava history. We build your mean-maximal-power curve from every ride, fit a critical-power model, and derive FTP from your sustained efforts rather than one ramp peak — so an anaerobically gifted rider is not punished with threshold targets built on a single minute of W' spend [Bentley et al. 2001]. Because the estimate is anchored in the power you actually hold for 10, 20, and 40 minutes, it corrects the individual bias the ramp bakes in, and it re-derives every time a new hard effort lands. The plan then adapts to that corrected number instead of asking you to survive an inflated one.

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
   Bentley et al. 2001. Peak power output, the lactate threshold, and time trial performance in cyclists. Medicine & Science in Sports & Exercise.
3. 03
   Michalik et al. 2019. Influence of Intensity RAMP Incremental Test on Peak Power, Post-Exercise Blood Lactate, and Heart Rate Recovery in Males: Cross-Over Study. International Journal of Environmental Research and Public Health.
4. 04
   Borszcz et al. 2019. Is the Functional Threshold Power Interchangeable With the Maximal Lactate Steady State in Trained Cyclists?. International Journal of Sports Physiology and Performance.
5. 05
   Borszcz et al. 2018. Functional Threshold Power in Cyclists: Validity of the Concept and Physiological Responses. International Journal of Sports Medicine.
6. 06
   Vanhatalo et al. 2007. Determination of Critical Power Using a 3-min All-out Cycling Test. 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.

See the direct comparisons

• AdaptCycling vs TrainerRoad →

More inside FTP without a test

Other articles in this series

1. 01

   Indoor vs outdoor FTP: why the numbers differ

   Why your indoor FTP reads lower than outdoor — heat, cooling, motivation, and power-source differences — and whether to keep two numbers.

   →
2. 02

   How to estimate FTP without a power meter

   Estimating FTP from heart rate, RPE, and Strava when you don't own a power meter — how close you can get and where the method breaks down.

   →
3. 03

   20-minute vs 8-minute FTP test: which to use

   How the 20-minute and 8-minute FTP tests differ, the multipliers each uses, and which one fits your riding — plus why both are only protocols.

   →
4. 04

   Why cycling apps show you different FTP numbers

   Strava, Xert, Intervals.icu, and TrainingPeaks can each report a different FTP. Why the estimates diverge and which one to trust.

   →
5. 05

   How often should you test your FTP?

   How often to re-test FTP as a self-coached cyclist — twice a year, not every six weeks — and why modeled estimates change the cadence.

   →