AI FTP initially optimistic?

No one has ever completed a workout perfectly

So, you do a workout* and rate the RPE and the prediction goes down.

Either:

  1. The RPE was higher than expected and the workouts leading up to the prediction date were changed, and so the prediction changed

  2. The RPE was as expected, but the upcoming workouts were changed because the AI decided there were better ones for reasons that probably no one can explain

  3. The upcoming workouts were changed because of fatigue detection (if in use)

  4. Something else

Bottom line: any workout can and likely will cause downstream changes. If you don’t have screenshots of all workouts btwn that workout and the prediction date, there’s no way of knowing what impact it had.

*for workout, read workout/ride/lifting session etc etc

They may not be outdoor workouts but you are riding outside in between your indoor workouts. So you have training fatigue that would limit your ability to do the indoor workouts as well as could be done if you went in rested. The prescribed workout probably factors in the extra fatigue to give you something it thinks you can do well.

So if you didn’t do those extra outdoor rides you might have had harder workouts to do leading to a bigger increase. Plus extra time off bike means extra time to recover which is where your legs get stronger.

Don’t upcoming workouts always get impacted from what shows up as activity on the calander? You may not be fatigued to get a yellow day but still not fully fresh to do a harder workout You rate the workout as hard so it was correct for what you could do then as it doesn’t try to inform you that you should have come in fresher. (maybe a feature request? strip out all non planned workouts and tell the user if the real workout done is harder or easier then that so the can know what impact outside efforts had)

This is the sort of thing that I was thinking of under my “2.” option. :+1:

People like you will find an excuse for absolutely anything. Ai is perfect, ergo the fault can only be with the user. There’s literally nothing I could say here that would convince you otherwise.

Workouts that get completed as prescribed and rated as expected should NOT lead to a reduction in Predicted FTP at the end of the planned out period. That is clearly not intended behavior. That is all that I am saying here and said from the start.

And I’m not in the habit to take screenshots before and after workouts. I’m not the TR QA staff. They are welcome to check my profile as I said before. But the problem above has now happened several times and every time it’s raised as an issue the apologists in here go crazy and claim stuff that I already refuted. And yes, I have pointed out several specific workouts where this happened to Eddie in DMs by now.

The workouts are done indoors, not outdoors. The prediction does not go down because of calendar or workout changes I made. They change directly after a completed workout (ERG Mode) that’s rated exactly as the Ai predicted (RPE). This has nothing to do with adjustments to the calendar or individual training adjustments. People keep strawmanning this because they don’t seem to be able to accept that a BETA feature is not working as intended at the moment (or better: still has some issues that need to be figured out).

Train drivers? I don’t see how that’s relevant.

I was just endeavouring to explain how I think it works and what people are seeing, not to excuse the results.

FWIW I think it’s a massive distraction and gimmicky.

Which kinda makes your straw-man argument that I think the AI is perfect and can’t possibly be convinced otherwise kinda moot, no?

Anyhow. Good luck, man. :+1:

Riding outside introduces the bug my friend. If you are riding outdoors, workouts or group rides, you are affected by the bug. She literally said this in her post.

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)

Swapping workout days or adjusting workouts is, in fact, changing the workouts lol. Anything you modify has cascading affects on all future workouts through the next 28 days.

What I DO have a problem with, is if completing a prescribed workout perfectly leads to a decline in predicted FTP.

You have outdoor rides. This is part of the problem due to the bug. She already explained this.

N=1 and all that, but as someone that has struggled with the bug since spring I do notice that in my current block that has no structured workouts, only group rides scheduled, the bug has suddenly disappeared.

yeah, I ride as much outdoors as I can. I found that building them my plan as indoor, then switching to outdoor workouts helped. I think now that changes to the prediction are driven by either obvious changes I make, compliance to the workouts, and/or HR and surveys. Of course, it’s hard to pinpoint.

Having said that, what makes the system good is that it’s training as it’s testing, so you’re actually working with the system to narrow the accuracy of your zones as you do more work and you get a much better working sense of what the range of your zones should be.

I have a little take that I think has been causing me the same issues and I have a little resolution that seems to check out, well for me.

I have noticed that my erg doesn’t exactly hold the prescribed power. Let’s say 200watts for 10 mins and my erg bounces my power 195w to 202w for that duration. Let’s say I end up with 198w for the 10 mins as an average.

Now this is all hypothesis, I am guessing that the AI is saying I didn’t complete that interval as prescribed. I am guessing that this compounds across intervals in each workout and I am guessing the ai is adjusting the next workout to be easier (prescribed watts) due to the failure. (again this is hypothesis). The reduced workout load compounds the issue as the training block goes on over the 28days, resulting in a lower reflected aiftp

So with that in mind (I use Zwift and TR) I’ve been upping the percentage like 2-3% for the erg to hold at the start of the workout adding a few watts to the prescribed workout. So now my 200w effort bounces between 200-205w on the erg.

For me this has seemed to resolve the issue of aiftp going down after “successfully” completing a workout.

I’m just putting it out there for something to try.