I personally like to use AI FTP only to track progress and then bump my actual FTP setting in 10W (~3%) increments once it’s pretty clear that the old number isn’t accurate anymore. That way I can progress PLs and time in zone.
The way they have it now, your FTP fluctuates pretty often and you kinda get stuck with the same workouts over and over.
I think this is key. Could I do a L6 z4 first day? Maybe. Would it derail my following training? Possibly, given I’m 44yo. Trying it on the final day of a loading cycle I’d definitely be struggling more.
I refer to the ‘FTP’ from a ramp test as Fake Threshold Power
So I just re-activated my sub today after a few years off. Last week I did a ramp test on my indoor bike (not on the platform, just a regular workout that went on Strava). I imported all my data and somehow TR AI detected a FTP of 309 when I maxed out my 1 min power at 376 making the FTP 282. WAAAYYY off!!
No chance I could do any workouts with ftp set to 309 lol
That certainly a big difference 
Probably worth a message to TrainerRoad support to see if they can explain it.
Otherwise just set it to a value you think is realistic for now and see what your next AIFTP says in 28 days time.
I’ve been doing longer intervals lately , AI ftp detection has been pretty spot on with an accurate ftp estimate.
Todays estimate 283
My ftp based on a “test” 2 weeks ago 285
With adaptive training and progression levels, having your FTP precisely set isn’t that critical in my opinion. If it’s set a little high, you’ll struggle in workouts and they will be adjusted accordingly. This is particularly true for any workout that isn’t around FTP because the % of FTP for those workouts varies significantly by individual to begin with. The only place I see it mattering much is when doing stuff around threshold (high SS, over/unders, etc.) and even that will self correct to a point based on workout survey response.
If you really want to know your physiological FTP (the point at which lactate is ~ steady state), I find over/under workouts to be better than any “test” unless you are in a lab having your ear pricked. With experience, it’s pretty easy to hone in on your FTP be feeling yourself coming in and out of threshold as you ride the fence of a sustainable effort.
For AI FTP, I’d say it’s good enough to get a starting point for workouts and then let the system do it’s thing adjusting progression levels over time. But I would not use the AIFTP number for pacing a 60’ TT (for example), it has no idea what my 1 hour power is unless I’ve given it a representative effort. But that doesn’t mean AIFTP is wrong, it just means that TTE varies greatly at FTP across individuals and levels of fitness.
AI FTP is very accurate for me personally.
I did an outdoor 20 minute test (just can’t motivate myself to do the traditional ramp test indoors anymore) towards the end of last year and it was pretty close - AI FTP 288 vs 284 on the 20 minute.
The most important thing for me - training levels feel spot on.
That is not “pretty close”, that is pretty much dead-nuts on.
More “accurate” than a test for me. I can have a good test during a lower fitness period, or a bad test during a period where I’m actually more fit, variance being down to how I’m feeling on that day. That’s just one example of subjectivity of testing for me. I’ll never test consistently enough (ie every 30 days) because life.
AIFTP takes away that subjectivity, and ensures regularity of testing. It’s also not adjusting FTP based on a single snapshot.
I’ve long been a proponent of long form testing and experimented with trusting the AI FTP numbers for part of last season. They were roughly accurate and certainly close enough for guiding training and intervals from threshold and below. So while actively training I think it’s a good model for me
However, I recently returned from a 5 week vacation with no bike access. Just for fun I used the AI generated value and it dropped me from 372 to 364.
I appreciate the confidence in my fitness but an 8 watt drop after five weeks off is kind of laughable. I think the logic around extended breaks could use a review
I’d be curious to hear how TR reacts as you start training again, like if you left the FTP and marked the workouts however they felt or failed them.
I left the ftp where the system set it but have been training exclusively by RPE. Like with any extended break I am resuming with a week of endurance, a week of tempo, then a week of sweet spot.
By the time I finish those three weeks and recover for a week I’ll likely be close to the 360 range, so we can see where TR puts me before I start proper training
My AI FTP detection recently suggested a bump from 214 to 220. I declined and then tried a 3x10 at 214 and failed miserably.
In my case then I think it overestimated by 15%. I am as sure as I can be that I am not tired, ill, injured, stressed or underfueled.
I’m prepared to accept that I haven’t ‘fed’ the AI much relevant info but I have ridden twice a week since September and half of that is TR workouts, albeit sub-threshold. I’ll reserve judgement until I have done regular threshold work but in my case it hasn’t worked so far.
Sounds like 214 is also an overestimate for you.
Your scenario sounds exactly like mine. Consider reading my posts above in this thread. I did a Kolie Moore test (you can find it as a workout on TR), came out 10 % below AI FTP (223 > 203). I accepted that and have done about a month of TH work since. I’ve rarely been able to complete over 8 minute TH intervals on TR. Yesterday I dd 3x15 at 100% ftp. It was tough, but manageable. Next week 3x18. I love the TR platform, and will continue to use it. However, I will use my own DOABLE ftp going forward. Hopefully AI will catch up down the road, I’ll monitor it and report back.
??? Where???
I should have posted that, it’s not easy to find. You have to go into teams and join “Team Gold”. Once joined, go to workouts, scroll down to teams and check team gold. I think @windwarrior put in on here (sorry if I’m wrong on that). I did the baseline test. The workouts have links to the trainingpeaks articles where the protocol is discussed in more detail.
I think I found it in the TH progression thread on here. Threshold Progression
I wish there was a way to save it to your workout library. I marked it as a favorite, but it won’t show up (because it’s a custom workout?). You have to go back into team gold to schedule it every time.
I agree - that was AIFTP too.
It was a reality check to see what my real off the couch FTP is, as estimated on a calibrated Kickr Core. Now to find some noob gains.
Ah thanks, that does sound a lot like my scenario. I read up for your other posts and agree with all of them.
I’m a big fan of the Kolie Moore test, I made one in the workout creator a while ago. I was indeed thinking of him when I dropped the power target today and tried to feel out what was sustainable for the rest of the intervals, which I finished. I’ll verify with another 3x10 next time and hopefully can enjoy it enough to progress the interval duration after that.
Good luck!