Thanks for your thoughtful response, and for all the effort you put into answering questions on the forum. TBH I’m not really on here much any more, in part because it seems totally swamped with posts about AI FTP.
It seems like if you could have some kind of explanation for predicted FTP changes, that might help a lot. Since this is AI-based I know it’s not always clear how input and output are related. But if there were a little (?) button that said something like what you posted above (“Your predicted FTP decreased because, based on workout X, workouts A-B-C were down-graded to easier versions”) I think that would help a lot.
I also think you could consider just de-emphasizing the predicted FTP number on the main page. A minor thing, but right now it’s by far the largest and most eye-catching thing you see if you visit trainerroad.com. Just having that number be “current AI FTP” with a smaller box below with days to next prediction and predicted FTP, with no “+/- watts” notation, might help. I feel like some users may confuse current FTP/fitness with the predicted FTP number, leading to questions like “why did my FTP drop 5 watts from 1 skipped workout?” (It didn’t, the prediction for 3 weeks from now did.)
I’d say that if your Sweet Spot workouts are feeling “Hard,” then they’re probably in a good spot.
Keep rating your workouts how they feel. If you find that you aren’t challenged enough with those VO2 workouts let me know and I can take a look at your plan again.
Or you could provide a predicted FTP range. Like with a 95% confidence interval your predicted FTP is X to Y watts.
this would provide a goal for those who want to keep predicted FTP showing but also the uncertainty.
I’m just guessing but predicted FTP isn’t based on some mathematical equation (e.g. FTPpredicted is a function of TSS, NP, HR, HRV, difficulty rating, W’ or other metrics we can see)? If it were, TR could provide feedback after every change in predicted FTP as to the reason for the change.
Anecdotally I can say that workout difficulty rating obviously affects the prediction. That’s not earth shattering news. You did 2h of 5’ VO2 efforts with 5’ rests and rated it “easy?” Newsflash: Those weren’t VO2 power so your FTP is set too low.
It’s not correct that it only happens with outdoor workouts.
Yesterday I had an indoor recovery type ride in my plan. I completed it & rated it as easy. My AI FTP prediction dropped from the previous prediction that was made after previous workout one day earlier.
I don’t understand why if I complete a scheduled workout, with RPE as expected, my predicted FTP should change?
The question I haven’t seen an answer to is why does the AI think that an experienced cyclist with many years of structured training can increase their FTP by 30 watts in 4 weeks? Why can’t the initial projection be more realistic?
Does TR treat workouts done on Zwift as “Outside”? I ask because I’ve certainly seen my predicted FTP drop after completing a workout, many times, and this winter I did all of mine in Zwift.
Would you be ok if that range dropped? Originally I was thinking providing a range would be helpful, but we’d be having the same discussions why is my range dropping. Or would you want a really wide range that incorporates non-compliance in it?
Maybe TR can post stats on the percentage of users who followed the plan exactly and the prediction was spot on (within a few watts) and x, y,z percentages off. They have all the data.
They may need to dig into individual accounts to get it, but they could I imagine pull it quickly. It would be nice to just see the scatter plot and I wonder if they could use that to estimate the variability without needing to run more simulations for the individual user.
I have yet to see a scenario where the Outdoor Workouts bug has affected indoor workouts in the same way.
I could take a look at that workout and your predictions if you’d like, and let you know what I find. Let me know if you’re cool with discussing that stuff here.
I think that It might have happened again today - after I completed the prescribed over under workout, rated it hard (as expected) and I think the expected FTP dropped (could be I’m misremembering the previous and didn’t drop today)
The more I use Trainer Road, the more I am getting convinced that the FTP prediction is just a gamification of the platform. It’s main function is to keep you hooked in to training with in the system. I am not sayng this to discount the value of the training, I actually think Trainer Road does a good job at prescribing workouts, it’s just that at the end of the day, they need us to become dependent on the system so that they can keep the recurring revenue.
Is this bug only for outdoor workout or also unstructured outdoor rides? After an unstructured outdoor ride my prediction went up and showing a 7 watt increase. Prior it was showing a 2 watt increase which is much more realistic.
So, mine dropped with perfect (or so I thought) compliance. Posted elsewhere about it. But I did rate some VO2 and anaerobic intervals as hard and did many of them outside. But did not miss or fail workouts did no added workouts.
So I would accept a range in this scenario: I follow the plan perfectly and end up with the bottom of the range for the final FTP at the end of the block. the message I take away: “you did everything right, but due to things you can’t control (HR? morning vs afternoon workout? Phase of the moon? Outdoor vs indoor? Bugs in the AI?) you only managed to hit the lower number of the range. You can’t do anything about it. Here is your new range for the next block.”
But suppose I did the plan perfectly and the range itself drops so the final FTP is below the initial bottom of range. In that case I agree. Having a range is useless.
