Agreed. I have done some different testing etc and for me it seems like it was too big of a jump. I get that the AI then does not use FTP to choose your workouts but I would like an accurate FTP for pacing races etc. I just don’t believe the number.
What you see are threads on the forum complaining about the issue. People who are happy with their AI FTP and current training aren’t going to say anything. Just because you only see negative things, even what you consider a lot, doesn’t mean a large percentage of the userbase is unhappy.
The outdoor workouts having affecting the predicted FTP probably does affect a large number of people - but I don’t think any of us have the stats as to what % of the TR user base does workouts outdoors.
When there is an AI FTP positive thread, lots of users show up to share their happiness and success with it. If you’re satisifed with the product, you aren’t devouring every thread in the forum about the topic and commenting “well, I’m happy with it”. That would be unhelpful and waste of everyone’s time.
Personally, I find the predictive FTP pretty useless. If I stick to the planned workouts enough, my overall fitness is going to stay the same or increase. And it really amuses me when people see a predicted change of +/- 5 watts when their FTP is 300+: we’re talking about a 1-2% change, which is within the margin of error for (nearly?) every power meter out there. (In this thread/case, it’s large 20W drops, but that seems to mostly be related to the outdoor workout bug for predicted FTP).
You have so many red and yellow days you are overdoing it.
Some of these red / yellow days were TR being too aggressive because my FTP was set too low. Getting back into training I don’t think the 1 month cycle of FTP increases makes sense. I went something like 200, 225, 250. The 225 felt really easy the last two weeks of the cycle. 250 feels “right”.
My calendar has always looked like this from a yellow / red day perspective. My best “season” was 2022. Still so many red and yellow days. Stacking days was how I helped build fatigue for an 8 hour race.
I have been exercising like this since early January and it’s been “fine”. With only two hard interval days a week, and usually one outside ride. I typically have enough flex in my schedule to feel recovered. I have occasionally moved an interval day because I didn’t feel ready yet.
I also have more red days then the calendar shows. Literally every Wednesday is a red day due to a martial arts class.
It’s likely this is interacting with the prediction being wobbly because it will see it as excess fatigue. I don’t think there’s any way around it unto your amount of work normalises. It needs data to train from and can only assume no work if it’s not shown. I had similar for the first two months where I really ramped hard to get past where it thought my fatigue should be and then it’s been good since.
Keep in mind that your FTP Detections will still be accurate each month. It’s just the predictions that are affected.
Any fitness gains will be accounted for during each FTP Detection. 
As much as that might make you feel better (and maybe me sometimes 
), that’s not the case.
There is something that is negatively affecting FTP Predictions when athletes have their workouts set to be completed outside. That’s the bug. It’s not bigger than that.
We’re seeing expected behavior for athletes who train mostly inside, and FTP Detections are solid for all athletes. It’s just predictions for athletes who have their workouts set to outside.
The fact that this is a smaller, more isolated issue, rather than a general issue with our model as a whole, is definitely a better scenario from my perspective.
5 Likes
But it’s not new, really. We’ve all been using it before it was launched.
The awareness is new, and unlike anything else on earth, our awareness affects behaviour and outcomes.
Personally, I find the predictive FTP pretty useless.
We can’t even agree exactly on what FTP really means. Then we guesstimate it with a ramp test, and then try to predict what it is based on a series of future workouts, and then never even attempt to verify any of this by doing an actual long effort. Out of all of the metrics available, it’s the most obscure and the least important. To build an entire training platform around it is actually wild. And now 75% of new threads are about it, as if it’s the only thing that matters. I feel like we have regressed as a community because of this.
While it may seem staggering, from a production side, it’s basically isolating to “only” that variable.
The system itself is working, but something about the outdoor logic is causing an issue, which is a much more isolated variable to solve.
I don’t think it’s meant to shun or diminish the population of people that ride outside a lot ( I am one of them ) but a discussion on the backside logic problem solving for the issue.
It’s not. Every cycling forum always gets into long-winded discussions about what is and what isn’t FTP. Is it 45, 60, 90 minute power? You can look at it from a macro level (measured power) or a micro level (lactate levels).
Then you can get into arguments about how to measure/test it. Ramp/MAP test, 8 minute, 20 min, 60 min, Kolie Moore, which one is more accurate…
FTP discussions have been around on the internet since Coggan and Allen first came up with the term.
The issue that’s being discussed here (and many other threads) is an issue with FTP Prediction that’s affecting athletes who have their upcoming workouts scheduled as Outside Workouts.
It’s as simple as that. We’ve done a TON of internal reviewing of dozens of athletes’ plans who have had concerns about FTP Prediction volatility, and this is the only major issue that reflects an issue with the product so far.
FTP Detection is still accurate. Any discrepancy between indoor and outdoor FTP is a general, lingering issue that has always existed based on cycling biomechanics, cooling, motivation, etc. That’s not something we can fix at TR that I’m aware of.
We all know that you aren’t happy with FTP Prediction/Detection, but let’s please not muddy the waters with your general dislike and disagreement with the fundamentals of the product when we’re discussing a specific issue. That doesn’t need to be spread across every single thread on the forum. We hear you, and we’re open to that discussion. There are plenty of “What is FTP” threads, and that’s a great place to have that debate. 
5 Likes
@eddie what is then the solution?
Got same issue here, I do all my workouts outdoors, and I’ve noticed that my FTP prediction keeps going down over time — even though I’m completing all the prescribed workouts, they feel manageable (even “easy” towards the end of the 3rd week), and in many cases I’m actually pushing slightly higher watts than the targets.
At the end of a 3+1 week block, the FTP gain is very small, and after each workout during easier endurance week, the FTP estimate drops quite significantly. Now I am starting the next block with with very low ftp gain, which doesn’t really feel accurate.
So I’m wondering:
-
When doing workouts outdoors, should I be aiming for higher power than prescribed (higher end of given wattage interval)
-
How should I rate workouts if they feel easy — does marking them as “Easy” negatively affect FTP detection?
-
Should I be doing an FTP test after each block to “correct” the estimate?
I thought the point of AI FTP detection was to avoid frequent testing, but in my case it feels like it’s drifting away from reality.
In my first block, the diff between 1st prediction and after 4weeks of training was not that big. In this block it is very low ftp gain.
Do you guys realease some new version of model? Or smth like that?
Why is that?
Why only outside workouts are affected w this?
Thanks!
edited to move into a diff thread
Welcome to the forum! 
Yes, it sounds like you’re running into this bug. 
Do your best to follow the power targets as closely as you can. If you’re feeling good and think you can overshoot a bit and still get the same intended stimulus, then feel free to do so. You can always use Workout Alternates to find harder workouts if you think you’re capable of doing more than what we’re prescribing.
Always rate your workouts as they felt. If the work you did felt easy, then mark it as easy. If we didn’t think that workout should have been so easy, we’ll use that data to help find you a more productive workout next time. Always rate the work you did as it felt. You don’t need to consider how we predicted it to feel, what work we prescribed, etc. Just rate what you did.
You shouldn’t need to test your fitness. This bug is really only affecting FTP Predictions, not Detections, so your fitness should still be on track.
If you feel like your FTP is significantly too high or too low, let me know, and I’ll take a look at what might be going on.
Let me know if this helps!