While I’m being controversial:
I guess there’s also the SYSTM ftp test, which is available with the free subscription. It uses an initial ramp test, followed by a long flat stint where you maintain a fixed HR.
Of course TR’s AI model is trained against TR’s FTP metric, so even if you doubt it’s accuracy, you should probably still use it as your TR FTP test, to benefit most from the statistical (AI) model.
Don’t. If you FTP is wrong right now, no “AI” will suddenly somehow make it correct.
Do a proper 40Km time trial and enter you measured average power.
After(!) that, you maybe could rely on future AI FTP updates.
Ouch. 40k might be a bit much depending on how much extensive threshold work someone’s done (or not done).
Anything over 40 minutes should be fine. Get a number and round down to the nearest 5 watts.
They eliminated the free workouts and tests a few months ago. You can still do it during the 10-day free-trial period without having to provide payment info.
Thanks @Slowwithouttwitch. If I ever need a more accurate FTP measure, I may give this a try. Such precision is beyond my current needs.
My point, though, was not that AI can somehow correct for a badly measured FTP to create an accurate one.
Rather, I’m saying that there’s no advantage to giving TR a “better” FTP metric. That will not result in a better training plan recommendation for most users. Why? Because TR is statistically optimised to convert ramp test FTP (or AI predictions of ramp test FTP) into the most beneficial workouts. You might actually expect TR to make worse recommendations on average if you provide a “better” FTP. They didn’t have that data when they created the model, so they can’t necessarily use it as input to the model prediction.
How should that work while people get unpredictable results from a ramp test randomly on either side of their true FTP?
What you maybe refer to is the adaptability of the workout difficulty prescriptions to accommodate for a wrongly estimated FTP. While that for sure works for millder offsets, the forum is full of people in trouble before any adaption had a chance to fully kick in, especially the population where the ramp tests rather severely overestimates their FTP.
Ok, I have no dog in any fight here, but happy to outline my understanding. My professional background has overlap with this kind of thing (like other users here), and if that helps I’m happy to chime in, openly admitting that I have no “inside track”, and could be quite wrong.
All measurements (of anything) have random variations on either side of the supposed true value (often expressed as a “confidence interval”). FTP measurement no different. FTP is even harder, because it’s not a directly measured quantity, but rather inferred from some proxy measurements (such as a ramp, or 40min test).
TR’s recommendation model accounts for this by the fact that they have many measurements in their dataset, and many workouts and resulting gains that they can fit their model to.
The resulting model is the best statistical fit for the data of many users, and predicts what workouts would, on average, benefit a subgroup of users with similar parameters (“Ramp FTP”, “PLs”, age, …) to the user being evaluated.
However, the user being evaluated may differ (in workout performance, or resulting gains) from the average user in that subgroup. Then, the difference between their observed performance and the anticipated performance can be used to predict such discrepancies in future for that user, compared to the subgroup.
I don’t believe TR is trying to accommodate “wrongly estimated FTP” using adaptations. I think TR doesn’t really care about “real FTP” at all. They care about making good recommendations based on the available data. And the available data is Ramp-test “FTP” (plus other parameters like PLs, and observed performance data). So when you provide TR with input parameters for prediction, they should be the same ones TR used to construct their model, imperfect as they may be. Those will deliver get the best predictions from the model, and will minimise the amount of “correction” required, on average.
Some unlucky users will initially get poor recommendations, because they happen to fall in a particularly poorly represented subgroup. They have two options:
I’d recommend option 1, with help from TR support as needed, but if you’re having great success with option 2, obviously keep doing what you’re doing, and kudos!
I can’t add anything profound here. Others have made amazing suggestions. But, I understand your plight. At 58 I have to prioritize sleep. I need it for recovery more than ever. Interval and strength work days knock me out. I love doing them but if I get 1-2 bad nights of sleep I start feeling awful and everything suffers.
I also fuel a lot and probably over eat on the bike. I kinda of like doing it. Depending on the day, I’ll eat a ton of oats or some carbs-based thing before a ride.
The only other thing might be tweaking your training. We don’t all respond the same way to the same stuff. Whatever you land on, just be consistent.
Good luck!
WRT to the Wahoo Systm FTP test, I like to advocate doing their 1 week test prep plan. It includes two days of warmup, their “Half Monty” ramp test on day three, a day off, two days of warmup, and their “Full Frontal” 20 minute test. The HM results are used to suggest power to maintain for the FF test. In the FF, I try to just beat the suggested power levels and it’s killer. IMO, it provides a very consistent lead-in to the final test, and is a very solid week of work as well.
Even a solid 25-35 minute effort will get one in the ballpark. I just did my first FTP test of the year (KM stytle). I knew what my FTP was by minute 25. I carried through to 35 minutes and stopped. I could have struggled to do another 5 or 10 minutes but there was not point because I had felt out my FTP and TTE.
I think that’s BS. It may work on statistical averages but it will work better with an ftp that is bang on accurate. Otherwise, you have to go through a period of workouts that are too hard or too easy and wait for the system to adjust. If they are too hard, you might burnout during the adjustment period. That will probably lead to more time off the bike to recover. It makes no sense to go through any rounds of that.
Just go do a real test and get your training on track immediately. A simple way is to just ride 30 minutes at the current FTP. That should feel like a 7 or 8 out of 10 effort. If it is a struggle FTP is too high. Lower power during the effort until you find the edge of a long sustainable effort.
TR works well within its bounds; it tests, it analyses, it responds to user feedback, it adapts. FTP has never been less important, it’s just a training number now, one that fluctuates, you don’t need to argue any more or bend over backwards to find “truth”. It couldn’t matter less. The important thing is getting the right, consistent progressive overload training …over time.
Id bet that if I added or subtracted 50 to my FTP the system would rectify itself in a few workouts.
Hard agree, but that’s assuming a well-calibrated RPE. My assumption was a moderately-calibrated RPE in order to be on the safe side. 
sure_jan.gif 
Trust me it doesn’t, even half that amount.
I agree with @AJS914 if you are one of those unlucky people at the edges of the bell curve it makes zero sense to blindly follow what is clearly wrong and wait and hope for it to adapt.
Its very simple to correct it from day one, I cant understand why anyone wouldnt do this (unless they were unaware of the problem, or had no clue how to address it, which has been explained many times.)
Its very easy for people in the middle or close to the curve to say trust it… dont if you are not one of those.
Everyone thinks they are at the edges of the bell curve, the reality is that almost everyone reading these posts, if not contributing to this very thread, are slap bang in the middle and absolutely covered by the rules that apply to most of humanity.
Search gluten intolerance, good drivers, and human intelligence for more.
If thats the case there are more people that should do a test and get a reasonable number.
You cant argue it both ways.
Everyone doesn’t think they are on the edge of the bell curve. A major are close to the middle and dont think about it at all (so they wont be in threads like this.)
A majority in threads like this question and think they might be because there is a good chance they are.
I’m not sure that I follow the logic?
The app has a test, and AI analysis to create an anchor number for training. They analyse millions of these tests and analyse the performance in the consequent workouts, and refine accordingly.
Why would any manually set number work better with this system for any particular cohort? And how can you objectively know which cohort you are in?
Then, given the PLs adjust immediately after the first workout and feedback, what is the priority on manually ‘fixing‘ FTP whatever cohort you think you are in?
Thats all fine and good (I would almost say that its too much…)
But the question is what you eat before the ride. Carbs on the bike doesnt matter if you come in with low glycogen in muscles and liver.
If you felt like the first sessions felt fine, then as the plan progressed it felt harder and harder, I would bet that you don’t eat enough off the bike, don’t recover properly, and then try to make up for it with 100g/hr on the bike (which won’t work long term).
Because some of us have been doing this for close to 20 years and KNOW where their theshold is.
Others don’t mind testing, even use it as a good threshold session, why not use this number they know works for them.
My point was people that are on the edge, not ones that think they might be. If it is a marginal thing maybe see it out, see how it goes but dont struggle for weeks.
So should I or anyone use a number I know to be wrong or one I/we know to be correct.
It honesty isnt as hard as people think it is, it is in fact is very simple.