First of all, my apologies if this has been covered in one of the TrainerRoad AI threads and I missed it. There’s so much traffic on those threads that it’s hard to keep up.
Here is my situation: Trainer Road has been consistently and significantly overestimating my FTP for years. While my actual FTP is in range of 240, TR estimates it in the 260 range. The new AI feature estimates 261 Watt.
If I go with this setting - as people have suggested on this forum - my prescribed workouts are set way too high: Sweet spot is now threshold, and threshold is now VO2Max.
I have been training with an inflated FTP for years, and it cost me a lot of progress because I was doing the wrong things at the wrong time. I learned my lesson, and I don’t want to go down that road again. I don’t know why algorithms overestimate my FTP, but they do.
Also, I am not particularly interested in raising my FTP as a goal in itself. For me, it’s a vanity metric. I race long time trials, what I care about is TTE. Given the focus of the new User Interface on “FTP increase prediction”, I am afraid that the TrainerRoad AI’s goals and my own goals are not very well aligned.
What am I supposed to do now? Just continue with my manually set FTP and continue testing as I used to do it?
What will then happen to my training plan? Will the AI continue to swap out workouts as it sees fit, and what will the basis of its decisions be? Will TR still track progression, or does a manually set FTP basically turn off any adaptive systems?
If this has already been covered somewhere I’d be thankful for links. But for me, this new feature is mostly confusing.
How and how often do you determine your FTP that you’re so certain of the value?
To answer one question - if you enter/override your FTP, the prescribed training should adapt to that. And will progress you based on that.
My experience - I got a near 13% FTP bump during the Beta when this new AI FTP detection was released. My thoughts and initial experience was exactly the same - prescribed Endurance rides were in upper Tempo and into SS. Thresholds were VO2 - with over-unders never “under” (not even near to old FTP!) - somebody coined the phrase “over-over’s!
BUT… if I pitched up well enough rested and I can actually do those, then the new FTP and the work it gave me is much more relevant. For the record, since this bump and leaving AI to prescribe the workouts, I had ONE semi failed workout - I skipped the very last 1 min section on threshold day, but it was after gastro the previous 2 days.
The biggest thing I got out of this was to respect my hard days. And I also do running, so it is difficult to juggle the spread of intensities. So I have learned to look at my intensity days across disciplines and often cut back on doing “just a quick 10+ k” or a loop in the mountain the morning before a threshold session in the evening. Or not to overload on fatigue on the weekend’s long ride, as much as that is possible.
Which FTP value you work from is up to you, TR will take it from there. Also play around with training intensities, etc on your plan as well, and see what changes that makes to the expected FTP impact. Even if you end up not using the value, it is good to have FTP predictor evaluate the impact of something you change - that is basically the way I use the FTP prediction.
Thank you for your perspective. I test my FTP after every meaningful block of training to catch increases or decreases. That’s roughly once a month.
Here’s my predicament:
I have a relatively strong 3 to 6 minute power output. I suspect that this is the reason algorithmically detected FTP is skewed towards the higher end.
When TR gives me a “threshold” workout which is set too high, but interval length is still within that range, I will be able to complete it. TR will be happy, and will continue to give me workouts that are too high, but completable. At least this is what the old TR did.
The result is that I will keep developing VO2Max, an area that is not useful to me (and I’m relatively good at already).
That’s the trap I fell into, and the one I really want to avoid. What I need to do is consistently work at threshold level to increase my TTE well beyond what TR gives as maximum interval lengths.
I’m in a similar-ish situation. What I did was compare the last Sweetspot workout I did pre-update (AIFTP ~15 watts below actual FTP) with the first one suggested by the new model (AIFTP ~15 watts above actual FTP)
I went from
3 × 30 at ~200w 0.85 IF 1265kj 90m TiZ Moderate
… to …
7 × 6 to 12 at ~225w 0.80 IF 1275kj 68m TiZ V Hard
The bottom line: The first was a bit easier than it should have been - more like high tempo, really. The second was a little harder than it should have been but I was still definitely below threshold. There were 7 intervals, and my HR on the 6th was only 3bpm higher than on the 2nd, which I think suggests I was managing the work well enough.
And I felt great afterwards.
I had a similar experience with my Threshold workouts as well. Before “Hard" - After “Hard".
I’m ‘only’ doing two hard workouts a week, so I’m quite comfortable with them being challenging.
So what I’d suggest is that, just out of curiosity, you accept the AI FTP Detection and have a look at the workouts it serves up. You may find that they’re eminently do-able. If they’re not, you can always set FTP manually.
This reads as if you’re conflating “improving your VO2 Max” with “doing workouts that TR labels as VO2 Max”.
These two things are not the same. All your aerobic work helps to improve your VO2 Max, all the way from Zone 2 to Zone 5. And if you increase your VO2 Max, FTP increases are pretty much inevitable.
This reads as if you’re conflating “improving your VO2 Max” with “doing workouts that TR labels as VO2 Max”.
These two things are not the same. All your aerobic work helps to improve your VO2 Max, all the way from Zone 2 to Zone 5. And if you increase your VO2 Max, FTP increases are pretty much inevitable.
Actually, it is “I am working above threshold in the VO2Max power zone when Trainer Road labels this workout as threshold”.
And if all you do is “go for 4 minutes as hard as you can”, I’m not sure how much this will do to your ability to ride hard for 60 minutes (assuming that we’re using this specific definition of FTP).
I know that it didn’t do much for me, and accepting a lower FTP did two things for me:
One, hard workouts that should just be hard didn’t completely wreck me, allowing me recover well and go hard again tomorrow.
And two, I was actually able to gradually increase the time I can hold that power up to the duration my races demanded of me. Which is the only thing I care about.
Of course, developing your VO2Max once you hit a threshold plateau (“raising the roof”) is important, but doing it “by default” by using an vastly inflated FTP for defining training zones did not work for me.
I’m curious about the AI adjusting my training plan to my performance, but I just want TrainerRoad to use a realistic FTP estimate, which for me is the result of one of the established test protocols (that correlates with my feelings and my race results) and not one it comes up with by looking at non-test performances.
But how are you testing it? Are you doing something in TR? Outside? Going to a lab? Other?
Also, my general takeaway from what I’ve read about the new TR AI is that you would be better off just ignoring the FTP number and doing the prescribed workouts (and, of course, rating them appropriately). Nate has said multiple times that the new logic looks at HR, wattage, and ratings, and ignores FTP. So, give it a try for a while to see if you can do the given workouts, or if it adjusts to something you think is more correct.
This is where I also similarly struggle with TR, despite the claims it doesn’t base VO2 work outs on 120% of FTP, it does do work outs based on percentages. My AI FTP currently I think is somewhat optimistic, but fine, I’ll go with it…..until my over/under work outs aren’t just a little bit “over” they are in fact now the wattage of my VO2 max.
It seems to be this bump of AI FTP for a few cycles and then when I just physically can’t do the work outs due my genetics, my AI FTP gets dropped. And cue the same up and down cycle. The AI doesn’t seem to learn that I couldn’t do particular work outs last time and perhaps the time before that, so perhaps it should offer me something different. Or a little less out of reach. Fitness cannot have an endless upward trajectory, but I could reach a plateau. TR seems bad at helping maintaining that plateau. And the plateau, as I say, could be down to genetics - that I have reached my ceiling.
Thanks everybody for your insights and perspectives! After reading up on a couple of things, I decided that I will continue with my tested FTP and not use the AI FTP (even though I’m not even sure anymore if AI FTP actually is supposed to be your FTP, so …). Let’s see if the AI will still be able to model future workouts and progressions for me.
How is keeping a more realistic physiological FTP limiting progression? We’ve been kicking this question around in another thread and the current “group think” is that it’s mostly a duration vs. intensity thing.
If you are limited to shorter workouts, I can see how someone can hit progression limits at their physiological FTP, but some of us feel that extending the interval duration is a more appropriate approach to continue progression. If you inflate FTP to get an appropriately hard “sweet spot” workout in a short period of time, I think you risk turning that workout into a threshold workout via the artificially high FTP. Which can absolutely be a good workout (and probably the best approach if time crunched). But maybe not the best approach for those with the time to extend the workout at sub-threshold wattages. Again, back to the intensity vs. duration balance.
Would love to hear your thoughts if you think that is off base and how/if we are limiting progression by keeping the FTP set closer to physiological limits while extending duration.
With all due respect, there are three things I will not do:
Continue hunting information that is spread across different media formats, in different locations and from different authors to try to understand the workings of the software I am using.
Deliberately fail workouts just so that an algorithm understands that it estimated my FTP too high, missing valuable training time.
Shift my training goals towards shorter, higher intensity workouts without my training plan explicitly saying so (the AI-scheduled workouts would be VO2, but still say threshold).
I understand that TR might down regulate the prescribed workouts once I rate them too high or fail them. But I fail to see how this is a productive approach to the problem.
Yes, I might be overly cautious. But after my experiences with an inflated FTP, I would much rather go the other way round: Start with my tested FTP that is too low in the eyes of the AI, and then increase it.
You certainty don’t have to (and shouldn’t) fail workouts to teach the AI. The good news is it considers all work you do, not just what’s prescribed.
You can:
Choose an easier alternate
Manually adjust your FTP
Find a workout in the catalog with similar watts to what you’ve been doing
Lower the workout intensity percentage
Lower the training approach to moderate or conservative.
Regardless of how you approach it, the most important part is that you fill out the survey for the work you actually did. The AI will adjust with this context.
I’m also curious how you are generally rating these rides in the surveys. Threshold rides for example, are usually supposed to be somewhere in the Hard or Very Hard range, but not Maximum Efforts.
That said, if you’re the type of rider that values playing it a bit more safe, in order to prevent burnout and preserve consistency (which is smart btw). You might want to consider a moderate or conservative training approach.
This all assumes everything is working correctly. I can’t see your calendar to get the full picture. But hopefully this helps.
