Love the AI FTP detection but wonder if it’s initially optimistic 28 days out and then converges to a more realistic number? I had a 239 estimate which landed on 222 (though this was when I first joined, so based on limited data). I now have a 241 forecast (currently 222 FTP) which would be another chunky improvement. I love the innovation though and it’s great to have something which reacts to each workout.
One suggested improvement - a fuelling strategy for each session e.g. x grams of carb to be consumed on bike.
My experience with the AIFTP prediction has been that it seems to be fairly accurate (=matches the eventual AIFTP detection) as long as I am able to complete all the workouts that were initially scheduled for the following 4 week period as intended and I rate each workout in accordance with the most likely post-workout survey answer expected by the AI.
(The model may also do some extrapolation based on HR which may impact the outcome but this is just my suspicion.)
However, the prediction seems very sensitive to any deviation from the AI expectations particularly during weeks 1 and 2 as these often have a cascading effect on the following weeks and can lead to a significantly different outcome at the end of the 4 week period.
I agree with you that the AIFTP prediction seems by definition optimistic as in order to achieve it everything has to go as the AI expected it to go (whatever this includes).
I don’t think anyone really knows how it works. Lots of observations but no real conclusions. We know, not following the plan exactly, reduces the prediction in most cases. I look at it but don’t take the numbers it produces as something set in stone. Example: during a build block I missed a short 30 min sustained workout. The next day AI had predicted a 9 watt increase in 28 days. The day before that workout it predicted a 4 watt increase. I had to laugh.
If you’re doing any of your workouts outside, you’re likely running into a bug that’s causing what you are describing. @TKilian this may be the case for you as well.
It honestly doesn’t really matter all that much to me, but if you think it would be helpful for others you are more than welcome to post my Calendar here.
For my part I’ve just learned to ignore the FTP Predictions since they are simply not reliable in my experience.
@TKilian, I think it’s helpful when it can cause misinformation/confusion to someone who doesn’t know about this bug and comes across claims that may or may not be true.
Here is your Calendar for your last set of FTP Detections:
I am not going to dig too deep into this one. I’ll just show an overview of what I am seeing with plan adherence that may be causing your FTP Prediction to be one thing on day one and the things that are happening in between that causes the Detection to be something else at the end of the 28 days:
While you may be completing your Interval Workouts Indoors, you are completing your Solo Rides rides Outside, and these are part of your training plan.
No training for an entire week (Germany Trip).
Manually changing workouts (hence the no Calendar icon on Wild Snow, 8/21), or pinning workouts like Phoenix -3 today.
Adding extra rides/changing things around like on Week 8. Looks like this should have been a Recovery Week, but since you took the Germany trip maybe you changed it to be a training week?
All the things above are examples of things that impact the Detection. They are neither good or bad, just normal life dynamics. In your case, I do have to say, that since you started training with TR in May you’ve been pretty consistent, your FTP has been growing, but most important of all is that you are rating your intervals as Hard so you’re getting the right training stimulus. So nice work!
On a side note RIP Tacx Neo, sorry for your lost, it looks like the trainer had a good run.
I had my Kickr since 2020 and I have a feeling it close to giving up the ghost
Is the argument here that once a person does ANY rides outdoors the FTP Prediction is not reliable because of your outdoor bug? I was told this only applies to workouts and it would be a strange argument to make that any outdoor riding makes it ok that the FTP Prediction gets reduced directly after a completed indoor workout
So? This was planned in before, so the Ai was completely aware of it at the start of the month. And as I said: the point here was that the prediction got lowered several times after properly completed workouts indoors. The week I was away for business didn’t come into effect on this at all.
Yes of course I change workouts. I adapt to my life and make sure that I move a workout from one day to another when needed, always checking for the impact on the Ai Prediction. Again I don’t think you are making a coherent argument here. I have not deleted any workouts or reduced them. And with every change I made sure that the Ai Prediction stayed the same or increased. I actually put quite a bit of effort into my training calendar management
See 3
This is my main problem currently in this forum. A person makes a valid point, in this case: Even perfectly executed workouts can lead to an immediate reduction in Predicted FTP, even at the very end of the month. And the response has nothing to do with the actual point but tries to excuse it with things that don’t matter.
As I outlined above: any change I make still keeps the overall structure, keeps the workouts and ensures that I maintain the prescribed intensity. And with every change I have to make I also double check the potential impact on the predicted FTP and adjust if there’s a drop.
How do you know about that? Did I post in here about it somewhere? And thanks for your condolences My new Tacx Neo 2T has taken over and is so far doing a fantastic job!
The forecast is useful as a rough direction, but it has enough moving parts that I would not chase a single four-week number. I would use it as one bit of context, then look at completed work and how the efforts are landing. Session fuelling guidance would be useful too, though that needs more context than a predicted FTP.
This is why I been turning it off during the 28 days and just turn it on for AIFTP day to see what the following 28 days may bring, then it’s back off again.
All the moving number was making me a bit nuts.
This is how I do it every 28 days just to see what was the difference between Prediction and Actual AIFTP and have and idea what predicted for the next cycle.
I have no problem with the Prediction feature but found the numbers always moving a distraction an had a feeling it was going to influence my RPE during workouts.
Did you actually read her post? You are doing 2-3 workouts outside a week, have made numerous changes to your workouts every week, and then decided the FTP prediction doesn’t work when both of those previous things impact the FTP prediction lol.
Changing workouts will affect the FTP prediction. Riding outdoors will likely involve the outdoor FTP bug. This is not a hard concept.
I’m sorry you have a reading comprehension problem, but none of what you said is correct.
First of all I don’t do workouts outside, which I mentioned. I do ride outside, but all my workouts are done indoors. So that point is completely irrelevant.
I also don’t change workouts every week. What do is sometimes swap the day of a workout or in VERY rare cases adjust a workout (usually to increase duration of the workout). Every time I do this though I check the change in the predicted FTP in the calendar and make sure to manage the potential negative impact. Nobody at any point said that a change in workouts should not impact the prediction. I can explain it to you, but I can’t understand it for you, sorry.
So maybe next time try to understand a post properly before you jump to conclusions. I have absolutely zero issues with FTP Predictions changing when changes to the workouts or the schedule are being done. What I DO have a problem with, is if completing a prescribed workout perfectly leads to a decline in predicted FTP. This is not a hard concept.