I tend to agree, if you dont like the ai FTP then take a ramp test. Mine was wildly low at the begining of the season and it suggested i take a ramp test and just have not had the time to do it (that ramp is a taxing big deal for me, i def need a rest day after and treat it as a 1 of 2-3 HIT days). I think that if you ride outside a lot and do not have clean data (not all bikes having PM,s HR data) i think the AI FTP will test on the low side. As it stands right now TR says mine’s at 236 and just this past weekend i did a 50 min crit with a NP of 316, it keeps suggesting i do a ramp and will do soon. My road bike is my only bike with a PM so i think thats whats going on but, still makes me loose a little faith in the platform. link to my calendar for ref: TrainerRoad
I think it’s not stressed enough that the athlete is the coach and the software is a tool.
One shouldn’t lose sleep over an ai ftp prediction fluctuating. It’s just a software prediction and if one’s FTP hasn’t actually changed, it doesn’t matter if the software fluctuates around. But people seem to get all bent out of shape by the number. Just do the training and the numbers will self correct over time. If you think, it’s a problem contact support. Or, you know what just do a test. Go ride for 30 minutes at 95-100% of your FTP. If it’s easy, you learned something. If it’s impossible, you’ve learned something else.
There was a topic recently about a rider choosing the bare minimum 3 hour plan and then complaining that it wasn’t appropriate for their 5 hour gravel event. TR asks you how much you can ride. You have to be honest and we all know that more time in the saddle is better. I don’t care if they said in some marketing blog that less time with higher quality is better. Everybody knows that pros train 20-30 hours a week and top local amateurs are putting in 10-15+. Don’t think you are going to hang with the big dogs at 3 hours a week. It’s like wishing for miracles.
I do realize that they say on the home page - far down: “Your Personal Coach”. I personally don’t take marketing slogans literally. But if one wants that, then follow all the prompts to the letter and don’t try and second guess it constantly. You can’t have it all.
I mean people do that and then get told that they should have been doing more or adjusting their training when the FTP improvement plan does not improve their FTP or even regresses it….
Or if one does second guess the plan and acts as their own coach the forum and TR says they should have been following the plan to the letter if they wanted to improve.
This sounds like the strategy my spouse uses when we have an argument. 
I’m not 100% sure that the AIFTP lines up directly with , if you like’ traditional’ FTP. My AIFTP has been consistently higher than the method I’ve been using with good success for years, WKO5, but it has now come down a bit closer and I do need to do a few PD Curve efforts, so the two might align even closer once I do.
I’m quite a difficult person for the AI to deal with I suspect, I ride quite a lot of Audax and more so this year as I am chasing 50 AUK points for an award (you get a point for each 100km ridden), as i result I couldn’t really generate a plan via Plan Builder that I would’ve stuck to, so I opted in the end for one of the ‘stock’ low volume plans, sustained power build, and I’ve been quite impressed by how it’s adapted to what i’ve been doing.
One feature of the AIFTP prediction that I’ve found useful, is using it to gauge the effect making changes has, for example, after a big weekend last weekend, I was yellow until Tuesday, so I moved Tuesday’s VO2 workout to today (Wednesday), deleted Thursday’s Over/Unders, which cost me 4 watts on my prediction, personally think that’s quite a bit for one workout, I did put an hour long ‘endurance’ workout in it’s place, but it had no effect what so ever on my prediction, so I deleted it and will now have an extra rest day tomorrow as a result and will probably be fresher at the weekend for another big ride as a result.
I do have some doubts, for example, I firmly believe that your FTP doesn’t suddenly change overnight after 28days, it’s a constantly flux of a few watts gained or lost, and I have used other services that adjust on a daily basis, which means you just get used to doing the work and keeping an eye on the general direction of change rather than having to endure a whole month of watching a future number fluctuate (in some cases it seems wildly). But overall, I’m quite happy having TR outline a plan for me, but also give me the tools to make informed decisions on how and when to adjust my training, I see it a little bit like a colouring book, the outline is there for me, exactly how I choose to fill it in is upto me.
Thanks!
The workout was a simple 3x10minute @ 95% of FTP to establish the level of aerobic decoupling to check the lower bound on my FTP. The details of the intervals can be seen in the screen grab below. The ride was outdoors. The 10 minute garmin stop between the warmup and first interval was time to allow me to get into position to ride for 10 minutes continuously (I built the workout without a lap press):
I also did McGregor-2 yesterday and let the power push in the final intervals by feel (the target middle was 261). The HR drift up a little but it’s commensurate with how I let the power drift.
Sure, I’d appreciate that. You can contact me by DM.
I understand what you are typing/saying. I’ve started taking my completed ride data (and fully completed weeks of rides) and having other Ai platforms analyze them to explain to me what is happening with the TR Ai model and its scoring. For the most part, what Ai has interpreted about my TR workouts (and me adding some context to Ai interpreting my rides) has given me some reasonable explanations for why my estimated FTP dropped or improved post ride, if I’m looking at separated from my emotion.
It is frustrating that TR Ai estimates fluctuate so much. I’m leaning on my progressive levels and other metrics to tell me if I am maintaining, improving, or losing fitness. I feel strong and my fitness feels improved, so I am trying to be patient to see if something comes out on a podcast episode sooner than later.
I think what is probably required is an equivalent of power smoothing then. Some people like to see their watts change every second when they are riding, but most of us just need to see if we are in the right area, so set our power to a 10s measurement.
The same should probably be true for the AI FTP calc. For most of us it’s not useful or helpful to see that number swing wildly around every day. Just give us a weekly average that helps us understand we are heading in the right direction.
Lack of trust here…
The entire mistake here IMO is that the software was 100% ftp based, and always has been. 20min tests, 2x8 min tests, ramp tests, any way to get “accurate ftp for training plan planning”. Then they released something, keeping FTP as “the metric” in the forefront of the plan planning, but now try to explain it away. It either matters or it doesn’t? Show it and make it right, or don’t build the interface around it
I don’t know what to believe. And I’m still a believer in the software as a tool, I’m just waiting for an exceptance of a botched attempt and an explanation of steps to fix this AI roll out.
That’s great. Thank you. I don’t know how I missed this.
I’m pleased with it so far, finding that the one month FTP prediction keeps me motivated and on track. I nearly always finish sessions without a break now as I know that the AI punishes any non-compliance. It takes some time to get used to, but I’ve become adept at juggling my programme to accommodate changes without being overly detrimental to predicted FTP outcomes. My overarching approach now is to finish every turbo session and tell it that any outdoor workout was “easy” even if it’s a 100 mile sportive with 3000m elevation!
Sounds a bit scary… please Mr AI, I’ll be good. I’ll finish my workout! Don’t hit me!
You can only game it so much. Eventually, you’re gonna start failing workouts if you say it’s too easy. The system will be self correcting. Just work out and don’t worry about the number. If the workouts are hard, then you’re probably getting faster.
My experience has been great. I’ve definitely noticed that when I do a workout and rate it as very hard it adjust the next upcoming workouts down a notch. Makes a big difference in me not having to worry if I was going to complete the next workout. I did have a month where I was being overly generous to myself on the difficulty and I paid for it the next month when it over inflated my FTP. I failed a few workouts and it brought it back down. I finished the next month and I was back to where I was before. I also just find it exciting to see how my performance feeds into the next prediction after every workout. It is super motivating and keeps me from failing workouts becuase I hate it when it predicts me down. I was also nervous that a future week vacation was going to ruin my fitness but seeing the AI still give me an increase post vacation (I scheduled it as my recovery week) made me less nervous about that and now I don’t have vacation anxiety about losing my fitness. I would be interested in if they were gaining more customers or losing more customers after the update. I can’t imagine ever canceling my membership now because it is such a big part of my life as a middle age man with riding being my main hobby.
I don’t like the tendency for it to undervalue outdoor rides. They might not be structured, but if they’re long and hilly I don’t expect to see my predicted ftp drop like a stone.
Really impressed with the accuracy of the prescribed training sessions and steady progress. Interesting the AI ftp number is consistently close to the intervals.icu eFTP.
That’s an interesting observation. Mine is not even close to my Intervals one. I wonder how common that is.
How does your intervals.icu eFTP compare to your AI FTP. Mine is always very close +\- 2.