I believe that the bug only affects Outside Workouts.
Keep in mind that unstructured/unplanned riding will likely still affect your upcoming training, so it doesn’t surprise me that your prediction might change after logging an unplanned/unstructured activity. 
It sounds like Zwift workouts are not treated like outside workouts.
They should be handled the same way as indoor TR workouts.
I just replied to your DM. 
I’m having the same issue. I’ve come to accept that it is what it is and there’s not much I can do to change it. It’s certainly frustrating. For instance, I’ve had perfect adherence and rating of workouts, but now that I’m missing one sweet spot workout at the end of my first block of the plan, I’m dropping 5 watts (not from +5 where it was initially predicted, but an actual LOSS of 5 watts from my FTP), which is absolutely ridiculous. This is an FTP building plan.
That’s not real. One workout doesn’t make or break a whole block and if it does, then there’s something really wrong with the training. Over the first two weeks of the block, I was setting power records and the workouts weren’t even hard. The training was working great. But now all of a sudden, I’m down 5 watts from missing a single workout? Nah. That’s wrong.
Feels like a RNG doesn’t it 
Hi @eddie is this outdoor bug more thoroughly documented in another thread or elsewhere? I suspect it’s affecting my experience also so I’d like to know more. Thanks.
If you’re doing a good chunk of your riding/workouts outside, then you’re likely experiencing the bug.
Let me know if you have any specific questions.
Yes, I already understood that it’s likely impacting me. What I’d like to know are more details about the bug please? Is it all outside rides? Or is there a difference between doing TR Outdoor workouts vs. unstructured Solo Rides or Group Rides?
What known impact is the bug having on our training plans, predicted FTP, etc.?
Suggested workaround? (Other than doing everything indoors!)
As far as I know, right now, athletes who do their training outside are seeing issues with FTP Predictions.
Predictions for some are too high initially, and then drop over time.
Training should still be accurate, though, and the main problem is with predictions.
How is your training feeling? Are your workouts appropriately hard?
So, do I take it then that the actual FTP recommendation on the scheduled day will be accurate, despite the fairly wild swings in prediction?
My predictions each month drop quite a bit in the final week leading up to the FTP detection - despite doing the prescribed workouts and rating them appropriately. That’s been the case for several months.
This time around I began the block at 226 with a prediction of 233/234. It stayed around there until these last few days and now it comes down a watt or two with every ride (even the prescribed easy rides).
Training feels good and appropriately hard, yes.
Thanks Eddie!
I can see how the code predicting FTP and the code evaluating FTP can be different, so a bug affecting one doesn’t necessarily affect the other.
Does the bug affect workout selection? Trying to think this through, I suspect it does.
The following is my understanding of how the AI works. Some of it is logical extrapolation to fill in gaps not explained by TR:
If the above is roughly correct (it might not be), then I suspect the bug with outside workouts is affecting either the actual PD curve update (so the AI isn’t properly “seeing” the work done) or it makes the PD curve bumpy (so the AI has trouble interpreting it / deciding how to properly expand it). Either of those scenarios wouldn’t result in proper progressive overload. It might be a subtle effect if it’s just a few workouts but small errors compound over time.
Where this kind of goes off the rails is when someone does a workout (even if it’s indoors) and the graph shows perfect erg compliance, every zone type score matches exactly, and still the prediction drops upon completion. This isn’t just an outdoor rides thing and there is no information on why it happens.
Well it’s very likely that there’s more than one bug in a big, complicated overhaul of such a complex system, especially when there’s one or more AI black boxes involved that muddy the waters of QA testing.
I’m not so sure it’s a bug, but I digress and don’t want to take the thread off course.
You’re, of course, welcome to share your experience. If you’re having issues, I’d definitely like to hear about them.
If you don’t think this is the place, feel free to start another thread or send me a DM.
Hey @eddie, I’m skeptical. You’re basically saying this bug manifests as FTP predictions are over-inflated for the first 3/4 or so of a 4-week cycle and then it’s adjusting down in the final week to where it should have been for the 4 weeks of planned workouts - correct? So in the end we get to the right number?
If that’s the case, then why is the Ai giving me 4 weeks of work that will only amount to a 1 watt increase? My last FTP detection was April 3 at 226. Prediction at that time was 233/234 for a while. Now, just in the last few workouts of this final week leading up to my next detection date it’s dropped all the way down to 227. In fact, with still one more planned ride to go I’m worried I’ll be at 0 gain by the detection date. But I feel like I’ve nailed the work and rated the key workouts with moderates, a hard, and one very hard for the VO2 workout that was back in the first week.
See my point? If what you’re saying is true, then why is the Ai giving me a plan that when followed is only pushing my FTP up 1 watt over 4 weeks?
If you’d like to look at my calendar, TR username: dbf
Thanks
I’ll take a look. Can I share some of the details that I find here?
It’s important to know that it is possible to follow a plan and only see a one-watt gain by the end. There are many things that go into your training and adaptations.
Are you on a plan that brings the right type and amount (not too much or too little) of training stimulus to elicit gains? How is training compliance/execution? What about recovery and nutrition? How trained are you coming into this block? Was your training progression linear without many setbacks?
I don’t think that what you’re describing impacts our understanding of this bug, but I can see how someone in your situation could have concerns..