Trust AI/ML ? All workouts are too easy!

Hey guys I’ve been using tr for about four years now, maybe longer. I’ve been using and trusting the AI coaching element since it was introduced and quite enjoy it. Makes the whole process so much easier and takes a lot of the guesswork out of the equation.
Or does it?? It will always suggest a new training ftp at the start of indoor season and it always sets it low enough that any ss workouts feel alarmingly easy! Using breath rate as an indicator I don’t feel I’m even remotely close to zone 2 (partial but not strained sentences etc) let alone sweet spot.
I rate the workout as easy (because it was) and I’m feeling I’m not getting what I should from tr as far as training response goes.
What say you? Any insight that will make me feel I’m on the right training path? I’m reluctant to manually set ftp cause I’ll never get it right lol!
Is a ramp test what I should do?
Great work your all doing btw!!

1 Like

When in doubt, test.

  • You can use any test of your preference… be that the Ramp Test, 1x 20-min, 2x 8-min, Kolie Moore, 60-min, etc. There are pro’s con’s to any of those test options, but you’ll find someone who swears by at least one of them (and possibly swears AT the others :wink: ? ).

If the core of your issue comes from worry about AI FTP Detection accuracy for you, an actual FTP test of your choosing may be the best next step.


Does AI boost your FTP after base phase? How’s RPE for harder workouts (threshold, VO2 etc.)? I think each approach to figuring out your FTP has its challenges and they probably dependent on the individual.

I’ve found that AI seems to underestimate my FTP relative to a ramp test; however, I’m not sure this is a bad thing, as I think I tended to overachieve on ramp tests for whatever reason. Resulting FTP would be a nice high number, but I found myself at risk of failing workouts. I’ve found the AI FTP to make many workouts feel easier, but this also makes it much easier to stick to plans and avoid burnout. Over time my AI FTP has crept up to my previous ramp test FTP, but I’m objectively fitter now than I was in the ramp test days. (Basing this on outdoor rides and Strava PRs.)

Adaptive training also adds a further wrinkle here; I currently have a high adaptive training VO2 progression number, and am able to do those workouts at my current FTP. If my FTP increases much though, I may start failing these workouts; is higher FTP but lower adaptive training progression metric preferable to the converse?

@andrewe172 a couple questions come to my mind as I read your post…

Regarding this particular statement… I could be wrong here, but this implies you ride outdoor all summer (or whenever weather permits)? Do you have power data that is feeding into TR so when you come inside, it sees what you’ve done outside (power wise) OR are you going off RPE when outdoor? I’m speculating, but if TR’s isn’t getting power data for an extended period of time, it might be underestimating a “starting” FTP when you do get on the trainer…

Of course, if you are entering “EASY” on all the surveys, I would expect Progression Levels to ramp up at a faster rate AND AI FTP bumping up your FTP much faster as well once it starts getting all that data from your indoor rides and surveys…

My second question… if all the workouts feel easy (even once the PL’s and FTP start getting up higher), maybe your trainer needs to be calibrated (assuming you are using a smart trainer)?

All that said, as @mcneese.chad stated, do a test and see what number it gets you and go from there!

1 Like

There’s something to be said for beginner gains, where anything works.

There’s something to be said for getting a bump every time you FTP is estimated.

Neither of the above will continue, but if they do I’d keep quiet personally! I’d love to be in that situation.

Of course, if you feel you can perform better than stated FTP, feel free to do a test but I’d recommend a longer one…having done the exact opposite and over-performed on shorter tests, then gone on to fail a large number of workouts afterwards it definitely doesn’t help with motivation!

So nothing beginner going on here. Garmin pedals with both look and spd bodies so road mountain or gravel, I always have power on the bikes and tr can see that. When I remember I fill out the surveys after each ride. Saris h3 is what I’m currently on indoors. Carbs are high when needed but generally about 30g for each workout.
Guess I’ll have to carb up and do a proper test here soon

What is reporting power to TR for your indoor rides? Pedals or the Saris?

It tends to err on the easier side if you don’t have a lot of recent indoor history. If the last sweet spot workout you did was that easy - next SS workout you have just use the alternates feature to swap in one with a much higher PL - probably a stretch or a breakthrough. If you complete it without struggling then it will adjust all of your future workouts up to and above that level.

I’m using the saris for power on erg mode. I looked for a way to calibrate (albeit not that hard) the trainer but haven’t found it yet.

That could be what’s causing the difference. Power recorded by my trainer (Kickr Core) is off by about 20 watts vs. my Assioma pedals. I’d move your pedals over and run TR with Power Match. See how that works.


Hey there!

If your workouts feel easier at the moment, they may become more challenging as you work through your plan and your Progression Levels increase, which will trigger Adaptive Training to adapt your plan for you.

If things consistently seem a bit too easy, though, you could try manually bumping your FTP up by 2-3% (so not too much), or you could do a Ramp Test (or any fitness assessment you’d prefer :wink: ).

I also like the recommendation @maletero gave you – try using PowerMatch to see if that makes things feel how they should in terms of power. You can also calibrate your trainer in the TR app itself if you haven’t given that a shot yet.

Feel free to let us know if you have any additional questions as you continue along your training plan!

1 Like

TRs implementation machine learning is not determining whether the workout is the right workout for you (as a person comprehensive of all your training history) it’s determining if you are effectively progressing through their pre-described plan. It’s a subtle but important difference. If you feel like the plan is too easy for you then your adherence to it is irrelevant. The plan can still be too easy even if the machine learning shows perfect compliance