Intervals.icu interval detection and training tool

[Job]

I did a steady ride on zone 2 (180W) for 1 hour but still I get "bad" on model fit 

 

[shaper]

Tx a stack! I have made good progress with improving FTP estimation over time, expect to get some more of that live this weekend. As always suggestions for improvements welcome.

FTP estimation getting much better and close enough to what am getting using Golden Cheetah which have used for many years.

 

Look forward to seeing the improvements

[davidtinker]

I did a steady ride on zone 2 (180W) for 1 hour but still I get "bad" on model fit 

 

The "model fit" is the r^2 value which is a measure of the distance between the line and the data points. So bad isn't necessarily a problem. You can see visually if the line is fitting the points well or not. I don't think its worth a lot on a l hr ride.

Edited May 30, 2019 by davidtinker

[davidtinker]

How does your fitness graph compare or differ from Elevate, I'm looking at your fitness trend and it reads about 9% lower than what I see on Elevate.

 

I prefer yours BTW

 

Tx. I am having a look at Elevate now. If they are doing the same Coggan chart then the numbers should be pretty much the same. Does Elevate give the "weighted average power" / NP numbers for your rides and do the rides have the same FTP set? Those 2 determine training load / TSS which is what is used to calculate ATL (7 day) and CTL (42 day) exponentially weighted moving averages.

[RossW]

I did a steady ride on zone 2 (180W) for 1 hour but still I get "bad" on model fit 

Was the ride done on an indoor trainer? Erg mode? I've found that when doing an endurance effort indoor ride, the long duration without large change leads to the graph producing a bad fit. You'll see that when you give it a bigger spread to work with, the fit will improve.

I'm including a TR Pettit workout; bad result. And a TR Bluebell workout; good result. The spread of data in the good fit is larger because there were intervals and then recovery periods.

[davidtinker]

Tx. I am having a look at Elevate now. 

 

 

The Elevate PSS (TSS / Training load) scores are within 1 of mine so its normal TSS. The fatigue and fitness numbers are also very close so also the same algorithm. What Elevate does differently is "todays form" = "yesterdays fitness - yesterdays fatigue" whereas I do "todays form = todays fitness - todays fatigue". So with Intervals.icu you are seeing your form after whatever riding you did vs Elevate showing you what your form was before you did any riding.

 

Not sure which is better or more standard.

[shaper]

The Elevate PSS (TSS / Training load) scores are within 1 of mine so its normal TSS. The fatigue and fitness numbers are also very close so also the same algorithm. What Elevate does differently is "todays form" = "yesterdays fitness - yesterdays fatigue" whereas I do "todays form = todays fitness - todays fatigue". So with Intervals.icu you are seeing your form after whatever riding you did vs Elevate showing you what your form was before you did any riding.

 

Not sure which is better or more standard.

Trainingpeaks calculates TSB (form) the same as you do https://www.trainingpeaks.com/blog/applying-the-numbers-part-3-training-stress-balance/ which I think is considered the standard ?

[Job]

Indoor trainer and ERG mode yes. I am doing a zwift TOW stage 7 tonight which is on a hilly road. I will try and spread my power and HR at different zones, will see if I can get a good.

 

Was the ride done on an indoor trainer? Erg mode? I've found that when doing an endurance effort indoor ride, the long duration without large change leads to the graph producing a bad fit. You'll see that when you give it a bigger spread to work with, the fit will improve.

I'm including a TR Pettit workout; bad result. And a TR Bluebell workout; good result. The spread of data in the good fit is larger because there were intervals and then recovery periods.

[Job]

Finally I got a good

[Mkmilan]

Hi. I am new here. After one ride of 5h my calculated ftp and cp are way bigger than should be. How is this possible? Thank you

[davidtinker]

Hi. I am new here. After one ride of 5h my calculated ftp and cp are way bigger than should be. How is this possible? Thank you

 

Can you please email me (david.tinker@gmail.com) a link to the activity on Intervals.icu. Then I can have a look and see whats up.

[HdB]

Hi David

 

I noticed this evening that the power model to estimate FTP has changed, to a combination of Morton's 3P (the default previously?) and FastFitness.Tips - what is the difference?

 

The new default gives me an estimated FTP that is 34W higher than that of Morton's 3P.

We had a 30km TT event this morning, where I averaged pretty much exactly the M3P number, albeit for less than an hour, and I could MAYBE have kept around there for another few minutes.

So the M3P number seems plausible.... but I very much doubt I could currently do 34W higher for an hour.

 

I think I'll keep my FTP on Zwift as is for now - some of the workouts are difficult enough as is

[davidtinker]

I noticed this evening that the power model to estimate FTP has changed, to a combination of Morton's 3P (the default previously?) and FastFitness.Tips - what is the difference?

 

It uses any max effort of between 60s and 30m to place you on a predefined power curve and then uses Morton 3P parameters from that curve for your power model. This has the advantage of only needing one effort and not a particular set of efforts like Morton 3P to get good results. The Morton 3P stuff is way out if you don't actually have max 5s, 2-3m and 10m+ efforts.

 

Please mail me (david.tinker@gmail.com) your Strava ID (go to "My Profile" and get it from the browser address bar) and I will have a look.

 

I am using the new algorithm for "continuous" FTP estimation like Xert does, but its not helpful if the numbers don't seem right for lots of people.

Edited June 10, 2019 by davidtinker

[RossW]

It uses any max effort of between 60s and 30m to place you on a predefined power curve and then uses Morton 3P parameters from that curve for your power model. This has the advantage of only needing one effort and not a particular set of efforts like Morton 3P to get good results. The Morton 3P stuff is way out if you don't actually have max 5s, 2-3m and 10m+ efforts.

 

Please mail me (david.tinker@gmail.com) your Strava ID (go to "My Profile" and get it from the browser address bar) and I will have a look.

 

I am using the new algorithm for "continuous" FTP estimation like Xert does, but its not helpful if the numbers don't seem right for lots of people.

I'm finding the new estimation method pretty close to what I've achieved through the Trainerroad ramp test. Quite interesting to see estimated FTP line which changes based on your rides - red box in attached image. You'll be able to see my as tested FTP in the image and well and can see the closeness of the estimated value.

 

Perhaps the higher estimate is the result of a real out-and-out effort, HdB? Maybe you're holding back on your Zwift test 

[davidtinker]

You are not supposed to be able to see that estimated FTP line yet Thats a bug.

[RossW]

You are not supposed to be able to see that estimated FTP line yet Thats a bug.

Oops