I agree with everything that @Torneng has said above. ![]()
@Niterider4 let us know how things go moving forward, and shout if you have any other questions along the way. ![]()
I agree with everything that @Torneng has said above. ![]()
@Niterider4 let us know how things go moving forward, and shout if you have any other questions along the way. ![]()
Bon chance! You get stronger during recovery, not during training. ![]()
My favourite is when you do a ramp test, create an increase my FTP plan, and the predicted FTP at the end of the plan is lower than the test.
Just add a button to turn the “AI” off.
I’ve been doing the TR stuff for around 8 years using a Wattbike trainer and Wattbike atoms, I mainly use TR in the colder and wetter winter and spring months so that when it gets warmer, I hit the ground running with a decent level of fitness. I don’t ride to power outdoors, it would suck the fun from cycling for me, I usually ride some of the tougher UK cycle sportive but I don’t race, I am quite dedicated to my training regime and I know what a hard workout is, if I’m struggling I usually know why, if I’m due a raise in FTP I can gauge what I am capable of, like many that have done this for a long time, I am aware of my limitations. I’m 57 years old for the record.
Lately these AI predictions seem pretty wild, a recommended bump of 6.5% in FTP seems excessive given my recent workouts, HR data and the fatigue that I have felt, I am aware that underlying illness can cause issues, but given my relatively weak showing recently since the last change in FTP, the data suggests that I am closer to my true FTP now, and if I’m not, the other metrics should be picked up by something with the word ‘Intelligence’ in it’s description, instead, it’s predicting a further 11 watts increase on top of that in three weeks time if I follow the current schedule.
I know that everyone is different, but I know my own body and I have been above a respectable 4 W/Kg most seasons and monitor not only weight, but composition, therefore I am experienced enough to say that I believe that the AI is flawed and seems to be way off with predictions. If I trained at the current suggested FTP, I would duck out of some of the workouts for certain.
Have a quick read on the forum here & you’ll find many others in a similar boat.
Broadly speaking. theres two schools of thought on it as far as I can tell:
The TR plans, both before and after AI, always struggled to serve exactly what every individual user needs in every situation. This is to be expected as its a big population with a wide variance in many different dimensions. This means that it’s always been somewhat necessary to be a bit self-coached, and to adjust things as you go. I think most people would agree that it’s preferable still with the AI to do this, although there are some who advocate a hardcore stick to the AI plan approach no matter what (“FTFP”)
My own view is that it’s become far too confusing to do that self-coaching adjustment with AI. Everytime that something is adjusted by the user, the AI reacts by making further (unwanted) adjustments in a cascade effect. In my case I found it to be overly sensitive to feedback surveys, and would dramatically alter a plan based on a RPE response of hard v very hard, which could have been borderline. That the FTP prediction was also so obviously crazy ebbed away at my motivation but also my confidence in the AI.
But others seem to have had success with it so it’s hard to give advice one way or the other.
In a previous thread I found the most reasonable (but depressing) response to be - maybe sometimes a product takes a direction that no longer aligns with what you value & it’s time to find a replacement that’s a better fit.
YMMV
Trainerroads ftp prediction is toward a best case maximum. So seeing an unrealistically high prediction seems to be a choice.
Before the new AI system TR FTP prediction was always a watt or 2 different from my ramp test result. The workouts were hard but do-able. Since using a (human) coach my workouts come via training peaks for me to do on TR. This has affected how my FTP is predicted, I assume because I am not on a plan, and I am not doing the workouts TR would give me. I’ve kept my FTP of 142 but the predicted is 126. Doing the workouts at the lower number would be much too easy, so I just ignore the prediction (or I’d be pretty demoralised). I will stick to doing the ramp test ,or similar in future. It’s embarrassingly low ,but I’m a half pint sized 70 year old female ![]()
I love when it’s always the user that is the problem in these discussions. Has the outdoor ride bug been fixed yet, over 6 months past when it was identified? For people that say to not look at the predicted FTP, that’s not the point. The point is the team thought it was accurate enough to put it in front of their customers when there is consistent feedback on how the execution of it isn’t aligning with customer expectations. To put it back on the customer is mind blowing to me. Explain why seemingly always having it declining is an appropriate behavior despite so many people questioning it. I’ve used TR since 2013/2014, and the constant point back to the used in this forum is so tiresome to watch again and again.
I’ve seen exactly the same thing over the last couple of months. At the start of a 4 week cycle of recommended workouts, the AI Predicted FTP at the end is around 2% LOWER than I’m starting with. Then, miraculously, in the last week of the cycle, my AI Predicted FTP starts to climb, and ends up with a decent increase over the starting FTP. I guess this is a “feature” of the AI, which I’m thinking of disabling! (If possible).
Yep, after being back on TR for a week now, I’m putting the predicted AIFTP functionality in the gimmick category.
That doesn’t really make sense with how workouts are generated by TR. perhaps forgetting ftp and zones would help you train to the prescribed power.
I find it incredible how many people think inconsistent and interrupted workouts are normal/good.
There was one guy in spring outraged that he hadn’t got “full credit” for a one hour workout where he’d taken seven breaks. Including getting off the bike and going to the toilet for five minutes. Physiology don’t work like that bro!
So I’d say there is a lot of AI-blaming going on on these forums, most of it reveals a misunderstanding of TR and training, and it’s discouraging, not encouraging, people to train well.
I had Mills +4 on Tuesday night, I am currently 251 FTP, the TR AI recommended FTP from last week is 267 and predicted in 3 weeks is 278. Had I accepted the 267 last week, I know that without doubt I wouldn’t have completed Mills +4 on Tuesday as I would be trying to generate another 32 watts sustained for the first part of each interval and tapering down from 32 - 16 Watts on the second part, my HR was already maxed out as the workout got to the business end and my legs were hanging off, my mind was even more of a problem as that was telling me to forget this training lark and go and have a beer or do something normal people do. ![]()
![]()
The forward prediction can seem fanciful but it’s just a prediction, the AI prescribed FTP however seems excessively high, I’d generally expect 1.5 - 4 % improvements on a trained set of legs depending on the level of fitness.
As I said previously, it may be that I have an underlying illness, it’s the cough and cold season in the UK and I think that I am actually fighting something off, so my respiratory system may be slightly compromised.
I’ve never had an issue with the FTP recommendations of the AI system, until recently, it can be a little flattering but after 8 years I know what’s realistic and what feels like smoke being blown up my rear.
I’m being light hearted and serious, not having a go at you:
No you’re not currently 251 and no you wouldn’t have had to generate another 32 watts! Because it wouldn’t have given you Mills +4. It’s trying to give you the optimum power targets and you’re fighting the system because you think you know better.
Maybe you do, but I just don’t get it. Pride? Fear of failure?
I just got back from ParkRun, Garmin wanted my first real run in months to be 4x4mins at 179bpm. I gave it the Paddington Bear Stare, but what the hell, I gave it my best shot and couldn’t do it. I’ve learned something, Garmin has learned something…And so we move forward. Nobody died. ![]()
Ps I am so heavy and crap at running.
Hey bro, that’s a cool story but what’s it got to do with this athlete? I find it incredible the amount of posters on here that want to find fault with the athlete rather than with the fundamentally changed approach to 1) FTP definition and 2) training plan formulation that TR has rolled out.
There’s a huge number of posts saying some version of ‘I’ve been using TR for 6/7/8 years successfully but lately it’s giving weird FTP/plan/workouts/predictions etc.’
That some posters seem to respond with ‘the problem is you the athlete’ is unconscionable really when it’s clearly a problem with the app.
But, you know, let’s discuss the 5 minute break that some guy took or whatever…
Quite. Good luck to you, I’m out.
Clearly in some cases. Dude need to share calendar, maybe like many cases it’s not only bug to blame for the lack of results. Nobody can deny that TR under delivered, but the scale isn’t 110% of the athletes.
While it may not have given me Mills +4 because my levels would all be adjusted down, my point, obviously not made clearly enough, is that I couldn’t produce that power to exercise at 120 % FTP in a VO2 max workout for even half the time (which is within the scope of a lesser workout), I am saying that a 6.5 % increase is excessive, that’s the bottom line and the projected 11 W in three weeks time is fanciful.
I do agree with you to an extent that it the system will select a less challenging version of the workout with the higher FTP, but the part that is missing is the higher FTP gets propagated to the other zones. I posted about repeatedly last year about having a too high FTP where the workouts could be completed for say over unders, but the SS workouts become threshold and the zone 2 rides are too hard. This adds way too much fatigue and isn’t sustainable.
I’ve been back on TR for a week and it appears to be doing the same thing. It’s clear that TR isn’t listening to its users.
Funny you say that cause I missed the part in the responses that indicated that problem might be on TR side. According to you, how much do you think is a TR issue of the issues raised? 5, 50, or 95%?
If the user says she/he has been doing structured training for 8 years and knows that the FTP estimate and prediction are both way off, then I don’t need to see their training history as I believe them.
Just for info, nobody has ever exactly followed a plan, both before and since AI. Amateur cyclists also like to ride with their buddies outside and have real lives that have to sometimes take precedent etc. We all need to juggle things around from time to time. Critiquing their plan adherence is just deflection from the fact that the TR FTP estimates and prediction are wrong in this case.
I get that you feel that’s being constructive (and that I’m not), but if you’re not first acknowledging the very experienced user’s feedback, it comes across that you’re trying to find fault with their training (or understanding) and discounting the under delivery on TR side.