I think the analogy is off.
You’re measuring two different things to predict the outcomes.
Imaging you’re trying to predict the price of a house: you used to guess based on square footage. It was ok. Now you’re, if we combine square footage with proximity to a school it gives better guesses.
The square footage measurement wasn’t wrong, it just wasn’t giving you the best prediction.
But then going ahead and saying ‘we have better square footage measurements’ is misleading.
Out of curiosity, how was the hyper parameter (?) of level 3 threshold chosen vs other ones?
To paraphrase my understanding here:
If you were under level 3, you’ll likely get a decrease in ftp but higher level workouts
If you were over level 3, you likely get an increase in ftp but a lower level workouts
The idea is to maximize the time in zone that the algorithm thinks is most effective for you?
I see, so a demanding plan was nerfing my ftp?
Does the balanced structure reduce my volume time or tss?
In this case, you probably shouldn’t be using ramp tests and the 20 minute test. Depending on the person tests have shown they can both be really, really off. For example, the 20 minute to FTP conversion can actually vary from 87-89% all the way to 97% for the conversion factor depending on the person.
Just get used to doing a steady state long form tests and being able to find that FTP tipping point and you’ll be a lot better off and you can pace off that.
I’m one of the lucky ones I guess, the new FTP and old AIFTP have been dead on for me and didn’t change, and both sanity check against what id call my “real” FTP for now.
But I’m wondering if when I start hammering Threshold workouts in a dedicated block for it, if it starts changing the number based on my Threshold progression…
@Nate_Pearson - if you have any insight why it’s been dead on accurate without a change for me between AIFTP (1w maybe), it’d be interesting as it seems like most people are seeing a change.
No, the previous system was AI FTP Detection + Threshold level 1-8. New system is AI FTP Detection + Threshold 3. So his power meter analogy is legit.
And in general, the new AI FTP detection is more precise and accurate.
@Nate_pearson, thanks for being so hands-on & visible with this!
I’ve gotta agree with Kev above though. Could this be possible for the athlete to set as an “advanced option”, or something for Support to set for them at their request? Would be especially useful for those who have no problems scheduling >1h workouts.
Same here. No change when I started the Beta, and while the initial predicted increase was large enough to make me skeptical, as I’ve done more workouts, the prediction has adjusted to be in line with what I’d expect from a base phase.
I don’t really agree here? It’s definitely measuring two different things.
Before it was measuring “old AI FTP Detection was predicting the result of a ramp test”
Now its measuring “Threshold 3”, which I assume is roughly correlating to ability to complete a L3 threshold workout.
Those are two different measures, which are trying to predict the best workouts for improvements going forward.
Is the new measure better and predicting the right workout? I’ll take your word at yes.
Is the the new measure more precise at predicting your ramp test outcome? no its not.
Pedantics aside, I’m looking forward to trying the new features.
Good points. No single metric for performance is going to be perfect and we can argue endlessly about testing protocol. However, I do believe “recalibrating to about a level 3 threshold workout” is an objectively worse definition for a platform-interoperable performance metric.
I see this change as the opposite. I’ve been riding exclusively on TR for the past 10 months. I have seen decent FTP increases in that time, but also Progression Level increases. My workouts have become all on the higher end of progression, level 7/8. This is the higher TSS @Nate_Pearson has been referring so since at these levels the only way to make workouts harder is to reduce rest. So my workouts mostly look like 3 x 18min SS with 30 sec rest. It’s a bit suffocating really. I got an FTP bump and a big reduction of levels so I feel like this has lifted the ceiling on what workouts I can be prescribed. I see a lot more variety of workout in my next block and I’m really excited for it.
So I feel like my old FTP was ‘undervalued’ but it didn’t matter I was still getting tough workouts and I was seeing fitness progression. With the adjustment I feel like it’s a more normalised value.
@Nate_Pearson Also as I look through the workout database, the workout that I did today 3x20 at what the new AI FTP Detection computed would equate to Pinnacle +3, Threshold 7.2, with a little bit shorter of a rest period (3-4 minutes vice the 5 of Pinnacle).
And I did the same workout last week at the same power on Jan 14th.
Thoughts?
Do adaptations still exist in the new version? I’ve only done 2 workouts since the update, but you used to either get a notification of adaptations or a notification that there were none.
It’s more automatic than that now. The workouts will briefly show the little magic icon, then resolve into what they’re planned to be. You could view the 28-day window as a forecast of what your plan should look like, but changes will occur automatically.
Also worth noting that you can pin workouts if there is something you didn’t want to adapt, for whatever reason. And I believe if you override a fatigue-related suggestion it automatically pins that choice.
My FTP increased slightly, but my WO levels took a giant leap, I can’t remember the levels before, but they weren’t this high.
And if I understand correctly it should balance around a level 3 threshold.
A caveat to that is that i prefer to do harder workouts, like the 3 x 20 SS rather than going back and doing level 3 SS work. And that I haven’t done a lot of threshold lately (only did Beacon recently and that was only moderate, bedore that I tried an over/under but had a bad day when I tried, and failed).
As you can see it also still recommends high level WOs, and not even a yellow day after a max effort VO2 (for instance that was a 6.5 that I failed), but Garmin had a load of 283 and recommends recovery, which is closer to reality. 100 TSS max effort VO2, is a lot harder than two hours of 70 % IF zone 2, but same TSS, which it seems is what triggers yellow days.
The question is if this looks right?
I get all the points TrainerRoad and @Nate_Pearson make about FTP and I agree.
From experience I can say that TrainerRoad manages to select the workouts which are “spot on”. I said that already long before the update: it does not matter how u feel about the number of FTP in TrainerRoad or how it is compared to other tests outside TrainerRoad - as long as the system is consistent in itself (which it is)
BUT: please @Nate_Pearson would u expand on the question which has never answered in any post I saw the last two years:
I train to race outdoors. How should I use any number from TrainerRoad to pace in a race (say: I want to make sure not to go harder than x% of my FTP on climbs in order to be able to complete a Gran Fondo well)?
If your GF has, say, three 10 minute climbs and you’ve been doing Threshold workouts like 3 × 10 at 300 watts at an appropriate RPE, I’m going to guess that aiming for 300 watts for your climbs isn’t going to lead you too far astray, all other things being equal.
These are both inaccurate analogies. In both cases, there’s an underlying true value, and one of the devices is closer to the truth than the other, so one is better and the other is worse. You can mitigate that through some kind of calibration (surely people around here aren’t new to PM calibration pre ride), but at the end of the day, you always look for whatever’s closer to the truth, call it a higher quality product, and pay more for it. I’m not OK substituting my very accurate AIFTP detector (or any device) with one that reads at least 3% off (would be considered subpar if it were a PM). I’m fine with two values that have two different names and meanings though. That’s fair game. Just don’t sell me a scale that doesn’t allow me to answer my doctor when they ask me how much I weigh, even though it shows me KGs each time I step on it.
This is the same question I have. I completely understand the metrics etc. in making these changes and I can’t wait to jump onboard. However I train to race, specifically TT. I can’t see myself in my current condition completing a 16km TT at my new FTP which jumped 6.7%. This puts my FTP at point that I would have struggled to hold for 20ish minutes in my best days which I know I am nowhere near currently!
I think the plan ultimately is that the AI will give you an FTP value but will also spit out predictions for 1 hour, 20 minute and 5 minute power.
Ahhh now I understand the intention. And in my case it is spot on level 3 (I struggled at 3.4).
Question: why level 3? That gives me approx 5x9min over unders. Would t threshold be better to train longer durations? Say 3x15 or 2x25, to really target towards hour power workouts. So maybe target level 5 and decrease FTP?
