I’m returning to TR after a period of minimal riding (so new to AI FTP & PLs), in preparation for a week in the alps come September. So far I have about 4 weeks of outdoor riding (7-10hrs/week), and both Garmin & intervals.icu calculate my current FTP to be in the 235-240 range, which also matches my previous experience with returning to consistent riding.
However, running the AI FTP for the first time estimates my FTP at 262! Whilst I’m already feeling stronger than I did a month ago, this number seems significantly higher than it should be, given both Garmin/intervals estimates & my previous experience.
Can I trust the AI FTP enough to base my training plan on, or should I start with the lower FTP?
I’ve never used intervals.icu but I think Garmin will look for your best long effort (?maybe 20min plus) to do the FTP calculation. If you haven’t been doing those type of all out efforts then probably the FTP will read low depending on your other riding.
Personally I’ve found AI FTP to work really well and would trust that. Then try and do some of the recommended threshold workouts and see how you fare. If you get through them then you are probably good. If you can’t finish and get destroyed early on, then just lower it.
Start with a ramp test and see how it compares to Garmin and Intervals. I’ve got a Garmin 530 and it’s pretty accurate if I do a hard 20+ minute TT, otherwise it’s low by 3-5%. In other words I can beat the Garmin estimate if I haven’t done any long/hard efforts. Better to be too low than too high.
In my case where I joined TR with lots of indoor structured training experience, AI ftp does a good job when it has 10 plus recent structured indoor (controlled environment) rides to work with. If you are just returning and don’t have structured rides in your recent history I would initially err on the side of caution. My ramp test, AI ftp, and intervals.icu are all within 2 watts. But that is with a lot of indoor structured rides. Either way you will know very quickly which number is accurate.
If in doubt, verify: I like @WindWarrior’s suggestion to do a ramp test as a way to see where things are. Then use a few threshold workouts to gauge whether the suggested FTP works for you or not.
When doing a Ramp Test, their eFTP is pretty similar with the result but will stay the same until your next ramp or a similar activity. It’s not a good a day-to-day indicator.
In my case, it’s 20W below my current FTP.
In doubt, decrease intensity by 10% and increase it little by little.
It will drop over time. eFTP requires maximal efforts of some minimal duration, if you never do them it won’t update. (AI FTP detection doesn’t really have this problem)
Garmin FTP would calculate on my chaingangs (pacelines) last year when I was going hard but not flat out. I would go with AI FTP the system will adjust quickly enough if you answer your post workout surveys honestly.
Thanks for the reply’s everyone. Some interesting points to consider. Like Dylan mentioned, I’m wondering if the difference is due to AI FTP currently only having unstructured outdoor rides as data,
As my workouts will also be outside, I think I’ll start with the lower FTP numbers and see how I handle them. If I’m finishing everything with ease, I can increase it, if the PLs and adaptations don’t do it automatically.
Intervals icu and Garmin can only use the data it gets. If you “artificially” cap the length and intensity of your efforts by following a plan, this will - by design - be a lower estimate.
If it goes well with the workouts so far, I’d continue on your current track. I also like the idea of a ramp test, but why change what is working…
I’m saying this from my experience long before AI FTP was around. Ramp test FTP was consistently much higher than the intervals icu estimate, and I was training well without problems. Mostly long sustained work (triathlon/sustained plans).
once I came outside and do rides with longer or higher efforts the icu estimate came up while my ramp test ftp didn’t change.
Go with AI FTP, answer your post ride surveys honestly and let adaptive training do its thing. My intervals.icu estimate currently is 235W whilst AI FTP has me at 270 W it’s just that it has been a long time since I have done a single very hard/max effort
When I came to TR I had 3 recent FTP numbers from different tests. I averaged them for a starting point, then I focused on getting my 10 ‘indoor’ workouts in so AI FTP could do it’s thing. Also, if you do the workouts as indoor mode from a device you can adjust the intensity up/down based on how you feel that day.
2-Standard 1hr FTP test with that dreaded 20min interval had me at 284 & 287
Wahoo Systm 299
TR ramp test 275
After 5 weeks AI FTP now has me at 284, I’m totally rekt and in dire need of the recovery week that’s finally here. wheew lol
Use aiFTP and do 3 mintute Threshold intervals and 5 - 6 minute sweetspot intervals. Dont worry adaptive training will adjust (well that is what happens if it is not ballpark correct)
Intervals.icu will typically give a low eFTP unless you make all-out efforts. The eFTP for a ride is calculated based on the power vs duration curve for the ride. So you will get the same eFTP for a 1x threshold as a 4x threshold workout unless the rest time is very short.
AI FTP Detection, however, seems to take this into account. And also multiple rides.
Garmin uses 95% of your best 20 minute average power within 90 days to estimate your FTP. It doesn’t account for the 5 min VO2 blow out test that comes before the 20 min TT which is what Hunter Allen calls for in his FTP test protocol. In most cases, it will over estimate. Doesn’t intervals.icu use one’s PDC to estimate the FTP? A variation of Monod model over the last 90 days. It’s as good as your dataset and whether you match the “average user” for the model.
Normally I’ll just bite the bullet and do one of the available test protocols. I prefer doing a 30 min TT or the Empirical Cycling’s baseline FTP test, see The Physiology of FTP and New FTP Test Protocols. I use the MAP test as a quick validation between FTP tests.
Not what’s specify on their website assuming your not referring to an actual test through the device or estimate one enter by hand. Care to provide actually info?
