# How to estimate FTP w/out a trainer?

#

**1**Full Member

Thread Starter

Join Date: Mar 2010

Posts: 222

Mentioned: 0 Post(s)

Tagged: 0 Thread(s)

Quoted: 188 Post(s)

Likes: 0

Liked 11 Times
in
7 Posts

**How to estimate FTP w/out a trainer?**

So I don’t have an indoor trainer (other than a Peloton) and certainly don’t have a road where I can ride all out for 20 minutes on. How else could an FTP test be conducted, or is really the best method to get on a trainer to do it?

**Likes For InvertedMP:**

#

**4**• —

Join Date: Jun 2015

Location: Land of Pleasant Living

Posts: 10,690

Bikes:

*Shmikes*Mentioned: 56 Post(s)

Tagged: 0 Thread(s)

Quoted: 8826 Post(s)

Liked 4,778 Times
in
2,546 Posts

intervals.icu will generate a power curve based on any all-out effort > 3 min using Morton's 3 parameter critical power model, FWIW. Mine actually has a pretty good fit for efforts from 1 min to 1 hr.

#

**5**Full Member

Join Date: Jul 2020

Location: Ellay

Posts: 216

Bikes:

*2002 Eddy Merckx Team SC Resto-Mod; 2019 Ibis Hakka MX; 2017 Spot Brand Ajax Belt Drive*Mentioned: 1 Post(s)

Tagged: 0 Thread(s)

Quoted: 105 Post(s)

Liked 141 Times
in
83 Posts

Yes, but unless the Peloton is properly calibrated (which many - or even most - aren't), the FTP result won't be useful for road riding.

Drive your bike somewhere you can find a suitable road. There must be somewhere within an hour of you.

Drive your bike somewhere you can find a suitable road. There must be somewhere within an hour of you.

**Likes For mattcalifornia:**

#

**6**Perceptual Dullard

Join Date: Sep 2009

Posts: 2,020

Mentioned: 33 Post(s)

Tagged: 0 Thread(s)

Quoted: 647 Post(s)

Liked 703 Times
in
292 Posts

Not any. For the Morton 3-param model you'll need at least 3 all-out efforts. You'll get slightly better (more robust) results if the all-out efforts are reasonably steady, though a nice thing is that they don't have to be slavishly steady. Also, the results will be slightly more robust if the shortest and longest efforts aren't too close together in duration, so maybe 3-4 minutes, 8-9 minutes, and maybe 16 or 17 minutes. The advantages of the CP/ W' model are that you get twice as many parameters to evaluate your condition than FTP, and you also get a way to assess the consistency of the tests.

#

**7**Advocatus Diaboli

I never understood how, using examples above, measuring output for a 3 minute or so effort, can tell you what you could manage for a full hour.

#

**8**• —

Join Date: Jun 2015

Location: Land of Pleasant Living

Posts: 10,690

Bikes:

*Shmikes*Mentioned: 56 Post(s)

Tagged: 0 Thread(s)

Quoted: 8826 Post(s)

Liked 4,778 Times
in
2,546 Posts

#

**9**• —

Join Date: Jun 2015

Location: Land of Pleasant Living

Posts: 10,690

Bikes:

*Shmikes*Mentioned: 56 Post(s)

Tagged: 0 Thread(s)

Quoted: 8826 Post(s)

Liked 4,778 Times
in
2,546 Posts

#

**10**Perceptual Dullard

Join Date: Sep 2009

Posts: 2,020

Mentioned: 33 Post(s)

Tagged: 0 Thread(s)

Quoted: 647 Post(s)

Liked 703 Times
in
292 Posts

Here's an example, taken from my data. I haven't ridden outside since March, so these data points are taken from my indoor rides, but you can get an idea of what's happening. The top panel shows the "familiar" power-duration curve (for an old fat slow guy who hasn't been outside since March). The exact same data, but transformed from watts of power to Joules of work, are shown in the bottom panel. The thing to notice is that the data, when translated into Joules, is

**almost**linear. (The correlation coefficient is 0.998). Since the data fit the regression line really really well, you can summarize these data pretty well with only two parameters: a slope, and an intercept. That's the basis for CP and W'. The slope of the regression line is in Joules/second, but a Joule per second is a watt -- so the slope is in watts. I haven't been riding much so the slope is really, really, low so please don't calculate it. In addition, the y-intercept is in kJ, and the intercept is called W'.

Notice also that the data being so linear make it easy to extend or project out to 3600 seconds ( = 1 hour). The nice thing is that you can see that at longer durations my average power decreases, so the data points are below the regression line. In fact, we have all the uusal regression diagnostics that tell us how well the data fit, and let us do estimates at various other durations I didn't test at. And, although there are around 10 points on that chart, you can sorta see that if I had only tested at around 3 minutes, 8 minutes, and 15 minutes, the points would still have been pretty close to the full line and pretty linear. So that's why we have some confidence that we can use data points at 3, 8, and 15 minutes to predict how much power I could put out for close to an hour. (Once I get much beyond a hour or so, my power drops off a lot -- but you can see that).

If your power were wildly inconsistent at different durations, you'd know that predicting out to an hour would be less reliable. You don't get that kind of insight from a single FTP test.

**Likes For RChung:**

#

**11**Advocatus Diaboli

Here's an example, taken from my data. I haven't ridden outside since March, so these data points are taken from my indoor rides, but you can get an idea of what's happening. The top panel shows the "familiar" power-duration curve (for an old fat slow guy who hasn't been outside since March). The exact same data, but transformed from watts of power to Joules of work, are shown in the bottom panel. The thing to notice is that the data, when translated into Joules, is

**almost**linear. (The correlation coefficient is 0.998). Since the data fit the regression line really really well, you can summarize these data pretty well with only two parameters: a slope, and an intercept. That's the basis for CP and W'. The slope of the regression line is in Joules/second, but a Joule per second is a watt -- so the slope is in watts. I haven't been riding much so the slope is really, really, low so please don't calculate it. In addition, the y-intercept is in kJ, and the intercept is called W'.

Notice also that the data being so linear make it easy to extend or project out to 3600 seconds ( = 1 hour). The nice thing is that you can see that at longer durations my average power decreases, so the data points are below the regression line. In fact, we have all the uusal regression diagnostics that tell us how well the data fit, and let us do estimates at various other durations I didn't test at. And, although there are around 10 points on that chart, you can sorta see that if I had only tested at around 3 minutes, 8 minutes, and 15 minutes, the points would still have been pretty close to the full line and pretty linear. So that's why we have some confidence that we can use data points at 3, 8, and 15 minutes to predict how much power I could put out for close to an hour. (Once I get much beyond a hour or so, my power drops off a lot -- but you can see that).

If your power were wildly inconsistent at different durations, you'd know that predicting out to an hour would be less reliable. You don't get that kind of insight from a single FTP test.

#

**12**Senior Member

If you do a search on Monod 2 parameter model or some such, you can find much better information. The basic idea is that if we ignore very short term sprinting, the power to drive cycling (or any exercise) can conceptually be broken down into two parts: an aerobic component which can be maintained for a very long time and an anaerobic part, which like a battery, has a fixed amount of power that can be used to supplement the aerobic part. This anaerobic part can be delivered quickly or slowly. In the model, the rate isn't limited, just the total work done. So then if you do an all out effort of any duration, you'll be working the aerobic part to its max (this power is referred to as critical power or Cp) and completely drain the anaerobic part, call W'. So then total work performed for any duration will be Wt=Cp*t+W' and plotting Wt against t will give a straight line. Of course, the assumptions in the model break down at some point due to fatigue, which is why we look at and model a persons full power-duration curve to account for very short and very long efforts.

#

**13**Perceptual Dullard

Join Date: Sep 2009

Posts: 2,020

Mentioned: 33 Post(s)

Tagged: 0 Thread(s)

Quoted: 647 Post(s)

Liked 703 Times
in
292 Posts

**predict**how well you'd do over long durations, but they do influence what you'd do over long durations. Suppose you had a great sprint compared to one of your friends, and your left hand (short duration points) went up, but the two of you had essentially the same right hand (long duration) points. Then your slope would be slightly lower than your friend's, and it would mean that your FTP or CP would be slightly lower. OTOH, if your sprint was weak ass and sucked worse than his, but your endurance engine was good, the right hand points would pretty much stay where they are but the left hand points would be lower. So in order to produce the same number of Joules at longer durations, your slope would have to be slightly higher.

[Edited to add:] Looking at the data in work-duration terms is really a lot clearer. You can see whether your data are close to linear or not; and if not, where it starts to bend.

*Last edited by RChung; 09-05-20 at 04:55 PM.*