Hi all, Looking for advice on using the AI or ramp test to determine FTP and other profiles after a long time away from TR
My question: how much does the TR AI engine need to get an accurate read of our profiles?
My Situation: . this fall, I’ve been faffing about with non-TR workouts. Since September, total ride time ~4 hours of non-TR workouts a week, once of which was sometimes a zwift race. There have been two 2-week periods when I was barely in training (bike or otherwise). One of those periods was two weeks ago due to work travel, and for the past two weeks, when I did get on a trainer, I focused only on SS and some long z2 sessions, which was good for the soul. But that means in the past 4 weeks, 2 have been nearly free of cycling-training, and the last two are SS or Z2 only.
So, should I trust the AI engine to put me in the right ballpark for my profile, or should I conduct a ramp test? I haven’t settled on any goals for 2026 yet, but i do want to make sure that my profile is lined up well so that my first training block is a good one when I re-sub.
I would do the ramp test. In a few weeks TR will gather enough data from your workouts to make the workouts fit your fitness. It’s not only FTP that determines what the best course of action is. Be truthful when rating your workouts. TR will test you with some really hard workouts.
This is something new people have been experiencing.
If I had to start with TR, I would probably pick 4 workouts (Z2, Z4, Z5, Z7). Choose something from achievable and something from stretch difficulty range i.e. 8 workouts in total. All those contribute toward your initial 10 workouts plus give good power profile for yourself and dial in your Progression Levels for further workouts.
I have taken some time off any real training from September to mid December. Just odd workouts here and there will no consistency whatsoever.
I decided to start a plan and started with a ramp test. In my situation the ramp test result was 3 watts from what my AIFTP was suggested to be.
Maybe this was because of the ad-hoc workouts I had done within TR feeding it at least some data? I’m not sure. What I do know is that I don’t need to look forward to another ramp test any time soon
Based on the riding you’ve been doing, what do you think your FTP is?
For example, I recently did a ride and threw in 3 efforts of btwn 10 and 15 minutes. The first one felt like sweetspot. The second felt like threshold. The third started like threshold and then as I increased the power I could feel that I’d pushed beyond a sustainable effort and into VO2 Max towards the end.
If you did a similar ride, you could probably ballpark your FTP to within 5 or 10 watts, and get a good workout as a bonus.
In my case, I am coming back to TR after a 3 year (maybe 4 year) absence. If you are fairly confident with your present FTP then I would just let AI do its thing. If you are not confident, then do an FTP test. After importing my history, AI thought that my FTP was within 3-4 watts of what I am in confident therein. I will just let AI do its thing.
Do a ramp test if you want to know the absolute truth with no question. I always thought the 2x8 min test was the best test but found the ramp test was pretty accurate.
TR moved on from the ramp/8min test for good reasons. Given appropriate data (ride history/workouts done in tr) their ai ftp gives a seemingly much more accurate result.
Yeah. Any tests other than the FTP defining 1hr test are approximations, and different riders will approximate differently. Bests to stick with one type to gauge progress. For starting TR, it’s best to to input or determine by test a reasonable starting FTP so that the first workouts are productive and provide good initial inputs to the AI FTP algorithm. Last I checked, AI FTP only uses actual TR workout data, so any non-workout rides or history don’t factor in.
Your last statement isn’t correct, I believe AI FTP takes into account all rides if you have power data.
From the TR support page for AI FTP:
‘Tired of stressing about FTP testing? So were we! That’s why we created AI FTP Detection. It looks at all your training—Indoors, outdoors, structured, and unstructured— to understand your fitness and detect your FTP, with no all-out efforts required! Now, you can skip testing entirely and stay focused on getting faster.’
Yeah. Now that you mention it, that’s correct, but it was that you had to do a number (5?) TR workouts before AI FTP would kick in. Is that still the same?
Disclaimer: I actually like ramp tests (and I reckon any other FTP test).
Over the years, AI FTP has improved and I have done less and less actual FTP tests. The ramp test has always worked very well for me. When I tried out AI FTP initially, it wasn’t very good at the beginning of a season or after a longer break. I’d severely underestimate my FTP. Now it works so well that I rarely do ramp tests these days.
Here is what I’d do if I were you:
To get a baseline, do a ramp test (or any other FTP test for that matter) at the beginning of your training plan.
Verify your FTP with threshold intervals. In my experience, both, steady 4 x 10 minutes at 100 % FTP work as do over/unders. I can typically pinpoint my FTP +/- 2 W. Adjust your FTP if necessary.
Then follow your training plan and let AI FTP do its job.
TR moved on from the ramp/8min test for good reasons. Given appropriate data (ride history/workouts done in tr) their ai ftp gives a seemingly much more accurate result.
