Thanks. Just zoomed in on my own Zwift workouts and can verify it’s actually there 
just much easier to see it when doing those workouts directly in TR app.
2 Likes
If you unpin the next work out it should go back to being able to be modified by the ai assuming it wasn’t something you manually put on the schedule.
If you delete it then it won’t just add something in.
If it was something you manually put on and therefore won’t become something adjustable by the ai. You could just add an ai chosen workout of the similar type of workout and it will give you the next appropriate workout.
2 Likes
I don’t think its failed that bad. It was however performed under the target power so it should be expected that it didn’t give you full credit. If you went that much over it would have given you extra credit. So it goes both ways.
The drop in FTP prediction is mostly just because you are now very close and the system is more accurate the closer it gets, less time for variations and it has more data. Dropping 10 watts sounds bad but if you are still going up that is still good and again, not a fail of this workout.
If you had just worked less on the rest intervals and hit the target on the main sets you’d likely still have seen a drop in both because overall workload from this workout would have been lower. But I think working under on the last 2 main intervals is the main thing working against you, against you isn’t bad.
1 Like
Without seeing your calendar it’s difficult to say for sure - but I’d bet it’s not just that workout being below expectations, it’s also because you have a ‘not recommended’ workout pinned. The model thinks you’re determined to do something it doesn’t want you to do…. unpin it and let AI select a recommended replacement, my guess is your prediction will go up. Not necessarily back to where you were, but back up nonetheless.
1 Like
It shouldn’t. Warmup and rest interval power shouldn’t really factor in to workout compliance.
That reminds me of my workout like a week ago. It was threshold with like 9 minutes in between intervals which was way too long for me, so I cut them down to about 5 minutes. AI asked why I cut the workout short? Literally made the workout harder by getting less rest and it says that I failed. Lol.
3 Likes
That’s a big difference than not being able to be remotely close to the prescribed wattage during the rest intervals. If I physically could not hit pretty close to the rest interval wattage, it would probably mean it was an “all out” effort, because it was so hard I couldn’t even maintain on the rest periods.
In your case of cutting your 9 minute breaks to 5 minutes, I’m not sure why you didn’t just select a different workout? I’d imagine you’re in tune enough with your fitness to make that decision before the workout.
1 Like
Not really. And they’ve said it countless times on the podcast that rest interval power isn’t important. Just keep pedaling to keep the legs moving but the power targets aren’t actual targets for you to hit. It’s a rest interval. They’ve even said that some workouts extend the rest interval for no other reason than to make the workout fit a specific time, like 1:30. That is, the workout itself might get you to 1:24 so they extend out the rest intervals to give you a round number of 1:30 for the workout. The intensity intervals are what matters.
Because it doesn’t exist. If my workout is 3x10 at 100% or whatever. Sometimes those don’t exist as a 1:00 workout. So they make it into a 1:30 workout and they use the rest intervals as fillers. I don’t need the fillers, I just care about the 3x10 intervals. Whether you noodle along for 5min or 9min in between isn’t that big of a deal. And like I said, I actually made the workout harder. So it’s funny that AI says it was a fail when TR for years has said the rest intervals aren’t that important.
2 Likes
Try This (note this only works if your FTP has not been updated since the workout)
Note down the day you did the workout and the workout name. Find the workout in “Browse workouts”, select it, then scroll down to the power record and create a custom filter for the day you did the workout.
Here is an example of mine, this is a VO2 max workout done on the trainer using the TR software in erg mode (i.e. not zwift)
You can see that ive exceeded the “perfect” workout (in NP) for the entire workout by 2~3w which is about 1%
Try this analysis for our workout and see how close you really got and post the screenshot.
The Ai is looking at what you did. So did you really get close to a perfect workout or did you undershoot the power by a reasonably percentage?
12 Likes
Fantastic use of that tool, love it!
3 Likes
There must be a threshold to cutting the rest out (usually just a minute or two), I have never noticed a negative grading for cutting back on intervals. Usually when I see recoveries over 5min I just incrase the warmup and scrub some of the midworkout rest. (Unless it is a gnarly workout (vo2 or on/offs), where i am thankful for it LOL)
Here’s the workout in question. I cut out the rest between intervals 1-2 and 4-5 because it felt too easy. So instead of 5x10 it was 20/10/20. Same TSS for the workout. Still rated moderate, really wasn’t that bad at all. Even upped the intensity at the end. Total time was short by 13 seconds and it asked why I cut the workout short. I picked Time because I remember them saying that selection would override the fail so to speak.
As far as I understand it, the AI is looking at the power duration curve of the actual work done in the workout. Note that this would be average power not NP because the AI doesn’t use FTP.
So having lower power in the rest intervals means lower power of any duration that is longer than the work intervals and this includes time from the rest intervals.
So yeah, the power during rest intervals matters, at least for the AI. Same with pausing the workout, because it has the same effect since the power duration curve is counting your power during the pause as zero.
Now, you can argue that rest interval watts do not matter from a physiological standpoint. Maybe they do, maybe they don’t. But from a mathematical standpoint they absolutely do. And AI and anything done on computers is all math under the hood.
2 Likes
Seems like a roundabout way of saying the AI is garbage. TR themselves have said in the past that recovery interval power doesn’t matter. And from a physiologic standpoint I don’t think it does; or it’s minuscule at best compared to the actual intensity interval. If I’m doing 320W for the work, whether I ride at 150W or 100W for 4 minutes in between is inconsequential to the goal of the workout. If the AI is actual AI, that is, has intelligence, it should see and know that. If a human can look at a workout and say, “yeah those 20W during the rest interval don’t matter,” then the AI should absolutely be able to do that. Otherwise it’s pretty useless. If the point of AI is to replicate a human coach, failing a basic function like that is a massive red flag in my opinion.
Same with the workout I posted above. I made the workout harder by removing rest periods. The AI should have the capability of seeing that it’s not a failed workout. Any human coach could see that in 5 seconds. Thats a huge problem if AI is having difficulty with a slight deviation to the planned workout.
1 Like
The AI did in fact see you did more effort and gave you credit for that effort. The Fail was due to time being cut, that doesn’t actually mean anything at the end of the day, you not only got credit but you got 20% more credit as far as workout levels are concerned for the main zone of the workout.
2 Likes
I don’t think recovery power matters if its a true recovery target, but the OPs look like Z2 or 3 target step downs. I think you can tell in the app what ones matter and those that don’t, the ones that matter have a actual average circle displayed in the player above or below the target.
3 Likes
Those pesky 13 seconds ruining my workout. Lol.
Again, a real human coach would be able to see that and wouldn’t have labeled it as cutting the workout short. Just AI being AI I guess.
100% agree with this. Some workouts the “rest” periods are part of the workout and definitely matter. But in my case it was a rest period where the power target doesn’t matter.
1 Like
But it doesn’t matter, doesn’t matter one bit, you got not only the credit but more credit for doing it. the system is only going to look at the power you made and the time you did it in for choosing the next workout.
Yes at the same time that makes it weird that its picky over 13 seconds but it doesn’t matter.
1 Like
Just as a heads-up, we are looking into this case.
I’ll let you know what I find when I hear back from the team. 