A buddy of mine and I compete quite a bit and we both use TrainerRoad. I’ve always based my training loosely off of the training plans and would do a ramp test every so often.
My buddy has always had a much lower FTP than me. He recently started riding very consistently on TR and putting in work. He selects his workouts based on what will move his ‘progression level’ up the most. So he’ll see that his V02 is lower than the other levels and will select a workout that brings V02 up.
His FTP has gone up shockingly fast! He went from 190w FTP to his current 264w FTP in just a few months. This could be due to the volume he’s putting in, but it makes me wonder if he’s on to something with his workout selection. Is this how TR is meant to be used? Did he find a hack for AI FTP Detection? Clearly what he’s doing is working, but this is the first time he’s had real consistency so maybe he’s just reaching his real potential for the first time.
I’m wondering if AI FTP Detection looks at your PL’s and gives you an FTP based on that. For example: If my FTP is set at 300 and I am a 7.0 in all metrics then AT FTP says my FTP should be set higher.
To be fair, I think his FTP is accurate. He did a ramp test and scored slightly higher than his AI FTP gave him a couple weeks prior. I’m wondering if his method of training is a good way to go, or if sticking to a specific plan is best.
No, not typically. The normally recommended use is starting with Plan Builder to make a plan on you calendar to start. Then perform workouts, answer surveys and let Adaptive Training make tweaks to workouts pending on the plan if/when needed.
Maybe, maybe not. That is a can of works to a degree and I can just about guarantee there is not concrete or singular answer.
People can and do use TR differently than the default ideal mentioned above. Some pick and choose manually by themselves, as guided by a coach or other methods.
Important note that leads into a broad comment… What works for him may or may not work for you, period.
Despite the shared use of TR between you, there are nearly countless variables that likely combine to give the differences in your results. Sure, if you mirrored his training it might be better for you. But it could equally be a lesser option compared to pure TR use or some other solutions entirely.
We have no real info on the deeper details of how AIFTPD works. The best thing to do for starters is to read what is provided in the support site.
Here’s a fun thing to check.
look at his, and your tss changes in the timeframe you are talking about.
Then compare any differences to the changes in your ftp. This would answer the ‘volume’ query.
Th equation is shockingly simple. Consistency + volume (2 hard workouts per week and fill the rest with as much as volume as you can handle). The rest is mostly smoke and mirrors. Some of it probably a positive distraction such as PL to help gamify your training and motivation. Some is probably negative distraction, worrying about the exact level of Z2 to make up your volume.
I would even submit that for most, just randomly picking your hard workouts is fine, as long as it is hard enough, which is similar to what your friend is doing. Follow this formula, ignore all the smoke and mirrors in regard to the lastest bestest way to train and I would bet you would be in the top 25% of all TR users (for a given volume).
I’ve started using hrv4training, it does FTP estimation too, it’sestimate seems pretty reasonable (I don’tuse TR anymore so can’t see what TR’s estimate currentlyis). The article below explains how hrv4training does it, I’m pretty sure trainerroad uses similar methods but with a much larger dataset
Someone please correct if I’m wrong, but TR does not use HR in any respect. The HRV4Training method indicates that it does us HR. As much as I’d like to see TR use HR, I’m also glad they don’t since my HR monitor(s) give faulty date as often as they give good data. I can be doing threshold and my HR says I’m at resting HR. It’s frustrating.
You are wrong. In fact the CEO encouraged users to keep their HR devices on until the end of their workout. TR didn’t mention exactly how they use it and what for, but I’d imagine it’s for some of the newer ML datasets such as AIFTP, PL2, unstructured outside rides, Red/Green light, etc.
It seems to me the differences are consistency and a structured training plan.
TR uses heart rate in several ways, be it to estimate TSS when you did not have power data for a ride (HR TSS). I’d be surprised if their other ML-based algorithms for e. g. AI FTP, AT and upcoming features didn’t use heart rate.
There are other ways to predict FTP/CP too eg model fitting like golden cheetah and wko do. Maybe TR uses several different methods to get an accurate figure
