Any downside to manual FTP adjustment with the new AI system?

[grwoolf]

The new AIFTP calculation bumped my FTP value up this week by about 8%. I’m not looking to get into a debate about whether the new AIFTP is better or worse or what FTP means for the general population, but the new value doesn’t work well for how I personally define and use FTP (particularly outside the TR system).

Listening to the podcast about the new features, I think @Nate_Pearson said the system doesn’t look at FTP anymore and is focused on actual work performed to determine future workouts/progression. And I’ve played around with changing my FTP back and forth and it seems the system is suggesting similar effort workouts going forward even with significantly different FTP values (which aligns with what Nate was saying). Sometimes the interval lengths and wattages vary a bit between the options, but they appear very similar in difficulty to my eye (and also align in the new difficulty predictor graph, which is a pretty awesome new feature).

So, assuming I’m not pushing up against duration/difficulty limits of available workouts for the effort TR wants to prescribe, is there any downside to setting my FTP to the value I feels aligns with my physiological FTP?\

Edit - I do understand that I loose the FTP prediction feature if I’m not using the TR calculated FTP, but it seems easy enough to just toggle that back on (ie - accept it) long enough to see where things are trending and then go back to the manually set FTP.

[kevistraining]

Only downside I have come across to using a lower FTP is that you reach “caps” on some types of workouts.

It doesn’t seem to prescribe sweetspot over level 8 for instance so you might find you either get stuck on a recurring level 8 workout or at least see less variety.

But otherwise no - I’ve stuck in some wildly different FTPs - 50w less and 50w more than my “actual” FTP - and the system seems to manage ok.

No way really for us to confirm which works the absolute best - but I don’t see a good reason to doubt that TR has modelled the best FTP to use.

[napoleon1981]

From what I understand, having high progression levels is relatively rare in the TR database. It begs the question wether this newAI approach is catering to the majority of TR users, and is more effective for the masses, but is giving up some of its efficacy for the edge cases like us.

So while it may be an upgrade overall, im still not convinced its an upgrade for everyone. Particularly the edge group that is highly trained. I would be interested in seeing some specific data around that.

[Savov7]

This January, I easily had my worst training block since I joined TR.

Week 1 : zones were off after my first AIFTP detection (10% overestimate), did threshold masked as SS and failed two workouts (one anaerobic and one supposedly sweetspot, but actually threshold watts, yet it was on a yellow day, so my bad).

Week 2: Manual FTP, good progression

Week 3: New AIFTP detection (the official release forced it). Just 4% overestimation. Still had a 20 min threshold interval in a SS workout and my OUs today felt like a poorly paced race. Completed my first 12-min set, but it blew me up and I failed the other 2 due to breathing limitations (cause the overs were very VO2-like).

I’m considering giving the AIFTP another chance, since it’s coming to its senses and isn’t planning another 3% increase anymore, but it also can’t find a good variety of workouts for me. Most of my alternates are not recommended. I think this will be better with a manual FTP.

@grwoolf we lose FTP prediction with the manual FTP, too.

[kevistraining]

@napoleon1981

Particularly the edge group that is highly trained. I would be interested in seeing some specific data around that.

Yeah I’m open minded but I’m sort of worried about a similar thing.

I like the idea of the “normalisation” and I’d love to be convinced that the normalisation value of threshold level 3-4 is based on hard facts and modelling - but I’m left wondering if they were scared of normalising in the 4-5+ range because too many people would get FTP downgrades?

So we who were on higher levels get unrealistically high FTPs

Likewise, I’m not certain that the FTP prediction is particularly useful for trained athletes because the expected increase in 28days (1 or 2w?) seems to be within the calculation error?

Personally I don’t think that the threshold level should be normalised down because it’s self limiting (you can’t complete 7+ workouts if your FTP is set correct) but i see the benefit of it being normalised up. Can’t believe so many people were doing level 1 and 2 threshold workouts and thinking they were ok.

I do think that sub-threshold workouts needed to be kept in check though - but I’m not sure that higher FTPs is the way to do it.

But, all that said, the new AIFTP isn’t too far off for me, my zones seem to be just about sensible. I’m rolling with it.

[Savov7]

@kevistraining

Likewise, I’m not certain that the FTP prediction is particularly useful for trained athletes because the expected increase in 28days (1 or 2w?) seems to be within the calculation error?

On one hand, I agree, but on the other hand, it’s a good guardrail to keep us from shooting ourselves in the feet when the weather gets better and we start mixing in unstructured outdoor rides.

[kevistraining]

@Savov7

it’s a good guardrail to keep us from shooting ourselves in the feet

That’s a good shout to be fair

[grwoolf]

@Savov7

@grwoolf we lose FTP prediction with the manual FTP, too.

Yeah, I get that. You can always accept the AIFTP at any time and then see the prediction, but it’s going to change you plan also when you accept the FTP (so the prediction is based on an updated plan). You could manually adjust all the workouts back to match the old FTP workouts to see what the prediction would be based on that, but that’s probably more of a pain than it’s worth.

I do think the FTP prediction is a cool feature though. At the end of the day, it’s just a guess, but likely a well informed guess with all the data they have. For me, the system is showing a predicted FTP that is basically unchanged or a couple watts up depending on my workout choices. Interesting to play with, but not enough to make up for all the other issue around using the inflated AIFTP value. If it turns out to be the ultimate “what if” tool for laying out a plan, I might change my approach to better leverage it, but I think I’ll just live without it for now.

I’ve changed my FTP back to my physiological FTP and I like the changes to my plan so far. Longer intervals with lower intensity resulting in similar levels of difficulty compared to the workouts prescribed on the inflated FTP. It is going to require me to extend my sweet spot workouts longer, but it was time to do that anyway. Much more appropriate than inflating my FTP and turning sweet spot workouts into a threshold workouts with shorter intervals IMO. And the system was smart enough to tell me I needed longer SS workouts to work with my lower FTP, so that’s a cool new feature as well.

[Savov7]

@grwoolf

And the system was smart enough to tell me I needed longer SS workouts to work with my lower FTP, so that’s a cool new feature as well.

I like that as well! If you can go back to the detected FTP to see the prediction, I’ll be going manual as soon as my endurance week ends, too.

[Caro.Gomez-Villafane]

@grwoolf I think the answer to your question is here: Metrics on new AI FTP Detection + TrainerRoad AI

But in short the new FTP allows the system to develop your fitness further into each training zone.

[BCM]

While I appreciate the work everyone’s doing, I don’t think that does answer the question.

While maybe a minority in the grand scheme of things, there are a number of people where the system is materially overstating their FTP, effectively biasing them towards a lot more intensity (e.g. making Sweet Spot Threshold, Threshold into VO2, etc.)

What I’d like to hear specifically - for this group of people, what’s the downside if we manually override our FTP and lower it? (And maybe there isn’t any other than losing prediction. Is it just longer workouts at a lower workout level for the lower set FTP?)

[Rizzi]

The following post reflects exactly my thoughts. Especially:

@kevistraining

I like the idea of the “normalisation” and I’d love to be convinced that the normalisation value of threshold level 3-4 is based on hard facts and modelling - but I’m left wondering if they were scared of normalising in the 4-5+ range because too many people would get FTP downgrades?

@kevistraining

Personally I don’t think that the threshold level should be normalised down because it’s self limiting (you can’t complete 7+ workouts if your FTP is set correct) but i see the benefit of it being normalised up. Can’t believe so many people were doing level 1 and 2 threshold workouts and thinking they were ok.

I do think that sub-threshold workouts needed to be kept in check though - but I’m not sure that higher FTPs is the way to do it.

[Bones]

@BCM

What I’d like to hear specifically - for this group of people, what’s the downside if we manually override our FTP and lower it? (And maybe there isn’t any other than losing prediction. Is it just longer workouts at a lower workout level for the lower set FTP?)

As I mentioned in another thread, I’m one of those folks who has too high of an AIFTP (IMO I should state). Once I complete this weeks prescribed intervals, I’m going to reassess in my recover week if I should just manually lower it.

I have been experiencing SS workouts at threshold and over/unders at FTP/vo2. I can do the workouts, but I really don’t think I’m working in the correct zones for my base period.

This also affects my ‘recovery week’ z2 rides. I’m essentially doing these in tempo watts!

I have been in the beta since the beginning, and I do recall Nate mentioning folks like me were in the very small minority who got big FTP bumps. I also recall him mentioning that the AIFTP may not work as well for 1 or 2% of folks. Are these people possibility the same population?

It seems like there are handful of vocal people chiming in who are experiencing, what they think, is too high of an AIFTP. My question to you folks, what is your profile? I’m wondering if we can see some similarities among us. I can go first for what I think my be relevant, but feel free to add more data points.

Male

50

Years cycling: 30 plus

Power profile: high 4-8 min. power relative to other zones

Previous highest w/kg: 5.1 w/kg (10 years) ago

Current w/kg (real vs predicted): 4.4w/kg / 4.55 w/kg

Predicted w/kg in 10 days: 4.68 w/kg

avg annual hours over the last 5 years: 600 or so

[lifeintheslow]

@Bones

My question to you folks, what is your profile?

Quite similar to yours to some degree.
Male, 50

Years Cycling: 21 years

Power Profile: high 4-8 min. power relative to other zones

Current w/kg (real vs predicted): 3.38w/kg / 3.68 w/kg

Around 360 h/year.

Every algorithmic estimation overestimated my FTP so far, and I think it’s because of that high 4-8 minute output. But that’s just an hypothesis.

[Bones]

@lifeintheslow

I think it’s because of that high 4-8 minute output. But that’s just an hypothesis.

I want to hear from others, but my hypothesis is the partly the power profile as you mention, but my mind goes to experienced ‘older’ cyclists who likely have seen close to their peak potential a few years ago, but still train a lot. I don’t see this population making big gains in their FTP, sadly this includes me! What this population needs is more of IMO, is to work on their TTE, therefore higher PL zones and likely highish volume. Again, this is just my thoughts…

[grwoolf]

@Bones

My question to you folks, what is your profile? I’m wondering if we can see some similarities among us. I can go first for what I think my be relevant, but feel free to add more data points.

Male

57

cycling most of my life, but proper training/racing ~17 years

Very flat power curve, not much punch but can hold high percentage of FTP for a long time

w/kg swings between ~3.8w/kg in the off season to ~4.2/w/kg at peak fitness/weight. The swing is a combo of weight and watts, so my watts don’t drop too much in the off season (~15-20w).

New AIFTP was 318 from 295. I manually downgraded to 302, which is about as high as I can convince myself based on recent efforts. I just started my season, so I expect to get to ~310 once I get my easy early season gains. AIFTP was predicting 321 in a month.

500-650 hours per year

[grwoolf]

@Bones

my mind goes to experienced ‘older’ cyclists who likely have seen close to their peak potential a few years ago, but still train a lot. I don’t see this population making big gains in their FTP, sadly this includes me! What this population needs is more of IMO, is to work on their TTE, therefore higher PL zones and likely highish volume. Again, this is just my thoughts…

That all aligns with my thoughts and training philosophy. My season peak for FTP hasn’t budged much in years, I consider it a big win just to maintain as I get into my late 50’s. But I’ve been able to keep nudging my TTE out. I set all kinds of power PR’s last year for efforts that exceeded 3 hours. After dropping my FTP back down to a reasonable level, next week’s sweet spot workout became sweet spot again and the duration had to be extended to get the required work in. I can appreciate that this is not ideal for someone who has to stay within time constraints, but I want to push that sweet spot work out longer rather than jumping into shorter threshold work this early in the season.

[DLanzone]

I tend to fall into this bucket as well, but my case is slightly different. I’ve been seriously cycling only for only 1 year, my AIFTP is ~3.2W/kg whereas my practical FTP is right around 3W/kg. Mid-30s. Not much endurance sport experience. I have a long history of weight lifting and have great shorter term power (10 minutes or less, and high sprint power), but my longer power durations are not great.

I originally tried the platform last year and experienced a similar issue as this thread mentions, where sweet spot workouts were threshold and the threshold workouts were basically vo2. Lots of suffering involved even though I was able to complete most of them. I was pretty intimidated of all of my workouts, haha.

I signed back up after this most recent release, and based on the planned workouts, my sweetspot work will be in the range of what I usually consider my threshold. However, I’m thinking it may be possible after getting several of the initial workouts in my legs, but I know it won’t feel like sweet spot, and the over/unders will be at the upper end of very hard. At least initially.

I am curious if it is intended to feel like this in the beginning as you ramp up and it gets to know you. Recently, I’ve been doing extended sweet spot workouts at 2.8W/kg and I’m pretty convinced that’s my true sweet spot, but maybe I just need to force myself to suffer a bit more.

[noahhudson]

This is the thread I’ve been looking for. I was starting to feel like that SNL alien abduction skit – everyone was talking about how perfect their new workouts are and I’m thinking “yeah, that’s not quite what I went through”.

My new FTP went up 31W from 253W to 284W, and it said it would go up another 30W in the next 4 weeks. I knew it was way too high but wanted to give it a shot because there are so many posts about it working great. I got through the first SS workout and finished it fine, but it was definitely a Threshold level workout. I’d been doing longer intervals before so I think I was set up for this okay – it was work, but I got through it. But then I had an O/U threshold workout where the unders were 1W higher than my overs from the week before. Ol’ Cloudripper -3 was light on the clouds and very heavy on the ripping. I made it through the first 9’ set, but the cadence fell off at the end. Got most of the way through the second before the wheels fell off – dropped the power 12% and finished it. Tried that for the 3rd but had to drop it 25% just to crawl to the finish. I don’t know when the last time I failed a TR workout was, but I’m not looking forward to doing that again.

I have no desire to get into a “what is FTP debate” – there are plenty of threads on that. But my new FTP is set to the max power I’ve held for 7’ in the last year. 3x9’ centered around my 7’ max power was never going to work. And worst of all, the text in the workout keeps telling me that near the end of the “under” I should be feeling great and ready for more. Let me assure you – I was not feeling great.

I get that the system will get to work making things easier, but like others here have noted I’m not sure that this is the right approach. It’s not lowering the intensity, it’s shortening the intervals. If it drops down to a PL of 1.0, I’m quite sure I can finish the set – but the set itself isn’t going to be what it says it is. Doing short VO2Max blasts while calling it a Threshold O/U feels like a different workout and it seems like that’s not what the overall plan intends.

I’m 49, cycling for 25 years (mostly triathlon). I haven’t ridden much at all in the last 5 years though – started again in October and have been going great. Motivation is back, losing weight, loving every workout until the AI fed me to the woodchipper. I am all onboard for the new system and want it to work – no desire to fight the tool, I just think something went wrong when my new prediction came out. If I manually set it to 265W, I can see the predicted workouts look tough and are more than I’ve done so far, but they seem realistic and achievable. But 20W higher like the AI has set it – there’s just no way. The upcoming VO2Max workout is 15x1’ at my 90” max power, and that’s my first VO2Max workout. I was briefly at ~4W/kg at 305W FTP back in 2014, but broke a femur in 2015 and I’ve been pretty casual since then and nowhere near those numbers.

My own personal hypothesis was that the issue might be because I was going up through the PL’s too quickly. From October until the end of the year, I didn’t have a TR plan – I was just trying to get back on the bike and doing random Zwift events, a TR TrainNow workout here and there, and easy Zwift Robopacer rides. Was riding almost every day, but mixing the intensity and felt great. Had a bunch of time off around Thanksgiving and the holidays and started a TR plan after New Years. Made it to the 3rd week of Base before the AI version turned on. It had upped my FTP to 253 at the start of the Base period and I was going up 1+ PL on each workout – it started me in the 5’s and the first two I marked “Moderate”, and then it moved me through the 7’s and 8’s and those were all Hard. I was thinking the ramp rate might have played into it somehow. My Threshold workouts in this were 4.4 (Hard) and 6.2 (Hard).

The other interesting thing is that it set my new PL’s for VO2Max to 4.3 and Anaerobic to 4.0 – those must have been 1’s at my old FTP since I have done any of those workouts, so I’m not sure why they got promoted. But I don’t even want to imagine what an Anaerobic workout at these numbers is going to look like.

If the answer is to just set a manual FTP for now and suffer through until I can get a new prediction in a few weeks, that’s fine – but it would be nice to get some feedback from TR that this should get corrected when it does a new calculation. I’d like to be able to just let the tool run and do its thing, but the current numbers are just way out of line with my abilities, and I really don’t want to go through another Cloudripper experience.

[lifeintheslow]

@noahhudson

everyone was talking about how perfect their new workouts are and I’m thinking “yeah, that’s not quite what I went through”.

I don’t want to guess the goals of other athletes or their experiences. But if I say “This is not a sweet spot workout, this is a threshold workout” then “try it, you’ll be able to make it” is not the answer I am looking for.

As you said: I know that I will be able to finish the workout. But it’s not a sweet spot workout anymore. So if your goal is not to complete the workout, but to create a specific stimulus to generate a specific training adaptation, completing the workout simply is not useful anymore.

To me, some of this looks like people talking past each other.