Forum menu
Looking at my power curve on Strava I notice something odd. I've always used the 20minute test to benchmark. Take the mean power number and multiply by 0.95. However, when I look at the power page on strava and check the estimated FTP box it has it at a value above my 20 minute power. About 12W higher so it is giving me a number over 30W higher than the one I would use from a test.
My 20min power is 363W. Not done an FTP test in a while but that was a TT so not a million miles away. FTP around 340. Strava estimates 375W which is crazy. I'd never maintain that for an hour, probably only 10-15min. Best hour output is 320W.
Anyone know how Strava calculates that FTP value? One of the training programs (TrainingPeak or TrainerRoad, I forget) estimates your FTP by using AI and comparing your performances against the dataset of others on the system. Apparently this works well and stops the need for regular FTP tests to structure your training around. Has strava started doing something similar.
One of the training programs (TrainingPeak or TrainerRoad, I forget) estimates your FTP
These are estimates. You're trying to extrapolate from one context to another. It will get you into the ballpark, but there are numerous sources of error so no point taking it too seriously.
I don't follow. I don't use trainer road or training peaks. They were just examples of ai FTP estimation without doing a test.
I do the odd FTP test but otherwise use my power curve 20 minute value *0.95 as it works well enough for an amature 40yo.
I was mostly curios about the Strava estimation as it seems wrong. Unbelievably so.
I found a few instances of this happening to others but no answer.
Should be able to ride at your FTP for about 40 minutes minimum, 20 minute tests get you in the ball park but a long form test (such as Empirical Cycling) will help you feel where your threshold is
https://www.trainingpeaks.com/blog/the-physiology-of-ftp-and-new-testing-protocols/
Google says that Strava just collates your recent efforts for various durations and adjusts for your weight, elevation change, etc.
So arguably I guess if you're sure it's reading high then maybe you've done a lot of stuff with a tailwind recently? Or downhill? 😉
I guess also we don't know what aero drag, drivetrain friction etc. Strava assumes, maybe your bike is just more efficient than Strava's assumptions?
I have manually entered my old FTP so don't get estimates, but I imagine (hope!) Strava would be estimating low as I doubt it takes in to account things like heavy winter tyres or riding in thick mud. It estimated my average power for a recent 55km of muddy singletrack as 88W 😆
At this time of year most intensity is done indoors. So normally a headwind 😉
It's all power meter data.
I'd be surprised if Strava are using AI to estimate FTP without making a big old fuss about it. Most apps seem to be very keen to say LOOK WE'RE USING AI and put some little magic sparkles next to a logo. My experience of looking at Strava's estimated power and calorie figures is that they can be reasonably self-consistent but rarely line up well with any measured values. That is to say they're not completely meaningless but if you have a power meter then the other tried and trusted methods are more likely to give you a useful value.
FWIW I've been using TrainerRoad's AI FTP detection for a few years now and it seems pretty good. It might be confirmation bias but if I do a threshold workout with their suggested FTP it feels like threshold to me. I think the key to their system is that you need to do at least 10 of their prescribed workouts indoors so they get some data in reasonably consistent conditions. Oh and they check RPE after every ride too.
I've not got a Strava sub these days to check, but I think its estimated FTP is not far off eFTP on the very good donate-ware site https://intervals.icu/
On Intervals. by default it does estimates based off 300secs iirc, but you can change that in settings.
Now has direct upload from Zwift, plus lots of other apps, soon to give https://crickles.casa/ data access and get the site working again after Strava's silly games recently.
It really doesn't matter what the number is as long as the measurement methods are consistent.
Being able to say 'my FTP is x' is useless, being able to say 'it's higher/lower than it was two months ago because I did this' is much more helpful.
Unless of course you’re setting training zones to an incorrect estimate?
That said, Intervals.icu and WKO5 get pretty close for me, but whatever model you’re using you need to feed it with good data on a regular basis.
I think the harsh reality is if you want to know your FTP the only way it to test it. Then you’ll most likely need to do a variety of tests over a wk or so. From those tests work out the average. Good explanation of testing here: https://uk.wahoofitness.com/blog/power-testing-what-when-and-why-do-them/
Why is everything 'AI' these days?
Strava has your weight/HR/Speed and terrain profile for pretty much every ride right? any directly measured power data is a bonus I suppose, they do some sums and it spits out a number, the fact that it's within ~12W of your tested figure is pretty impressive.
The real question has to be, is either number actually useful to you?
Personally I'd test it whichever way you feel, then try to ride at it for 30+ minutes to verify it.