Not sure about the prescribed duration of the workout that you already completed, but remember that the workout level algorithm is very inaccurate across different workout lengths. Just look at the longer and shorter tabs of your upcoming threshold workout. They vary greatly for the same difficulty. Longer workouts are over appreciated and shorter workouts under appreciated. Better is to simply look at interval duration and power if you’re going to compare what it’s giving you.
Thanks. I replace workouts with the same duration but slightly higher PL than the previous week. Ride time varies due to the time it takes to get to a climb.
2 Likes
Ah, ok. The algorithm is much better for the same duration, but I have seen cases where something that is rated 0.2 WLs lower at the same duration has a higher AI projected failure rate, so I check that too.
And I’m very jealous that you’re able to do your workouts outside!
2 Likes
Why? How high is it, to be a significant factor for you/us?
I’ve never even noticed the failure numbers.
I was under the impression that the new system is supposed to look at your PLs etc for unstructured rides, is that not correct? (basically WLV2 as has been discussed ad nauseum for years)
I did a Zwift race yesterday and was excited to have a look and see what the AI thought of it (OK excited is probably a strong word 
) but I look this morning and there’s nothing different to before - just the trace and power curve etc.
Not correct no.
AI will take your zwift race into account when assessing your fitness and prescribing future workouts.
But it won’t try and determine what kind of structured workout your unstructured ride was like - that wouldn’t really make sense.
3 Likes
Yeah, I get that part - the system itself only looks at watts etc. etc.. But it is still displaying FTP and PLs as information to the user (essentially). I thought one of the applications of the new system was basically to cover off the WLV2.0 which people have been asking for/has been discussed on here for literally years, which I guess it is still doing “under the hood” anyway using watts, but I guess I was just interested to see what it actually thought of the Zwift races.
It’s not clear to me from your screenshots, but I would add in all of the other ten hours risings solo rides, have a stab at the rpe, for the next four weeks.
See what that looks like.
Then try removing rides and switching them to structured workouts. See the effect.
My guess is that part of the reason WLV2 never came to anything is because unstructured rides are rubbish structured rides.
I.e. if you tried to assign workout levels to a zwift race all the zone scores would be really low because you are unlikely to have spent much productive time in any particular zone - you’re just tired - but you had a lot of fun at least.
4 Likes
I’ve loaded a workout after a zwift group ride and it then attempts to categorize the ride. I used a sweet spot ride as the ride to match against.
Do you get the same categorization if you match it to a VO2Max ride?
What did you set your training approach to with this new plan? I think you were at “moderate” before, which could be part of the reason that you were seeing slightly lower progression. Balanced might be a better place to start.
Regardless, you’re expected to see fitness improvements, so I’d agree that’s a win!
1 Like
I haven’t tried. I have to have a workout in zwift to link it to by loading. This weekend I’ll probably do a race and some threshold work, so I’ll see how that goes.
1 Like
so it’s not possible to disassociate the workout from the Zwift ride and put another workout on the day and the associate that like you can do for outside rides?
I’ve been having an issue that I’ve sent to support, just curious if anyone else has been seeing it….
Came up on my FTP prediction day, going from 292 to a predicted 305 or something. Then the day of the FTP prediction, the detection it gave me was unchanged at 292, but figured it might be because the next month covers a specialty phase in my plan. And my next FTP prediction gave me an even greater increase.
But now every day since, it tells me I need to accept the FTP detection again in order to get an FTP prediction. So I accept it, get the prediction, then the next day the exact same thing. And each time I accept the still unchanged detection, and then get another prediction for 28 days later. I’m now on my fifth consecutive day of this.
Anyone else see anything similar?
I’ve definitely seen a post like this before. Not sure if it was you who posted it or someone else.
Thank you Eddie
I kept it at moderate
Help me understand this.
“pauses lowered your score”
Workout description:
“brief, occasional backspins are encouraged”
I extended the recovery valleys some to stand and relieve genital numbness. It happened for the first time last week so I still need to do some saddle tinkering to address it.
Where do I tell the AI this?? 
Also some of the pause was in the cooldown.
What a mess…
Not even addressing the fact that I was assigned this workout after a rest week, and it predicted I would be yellow and red for the next month
It’s not a significant factor. I think that once the projected failure probably gets up to 2.5%-3%, TR starts telling you that it’s too hard because the chance of failure is too high.
The point is that sometimes, even when the PL of a workout is slightly less than another workout, the AI projected survey response might actually be slightly more difficult. This means that the AI doesn’t totally agree with the algorithmic assignments. So when users see next week’s ride have a slightly lower PL than this week’s ride, it doesn’t necessarily mean that it’s easier.
I hope that they start using projected survey response to rework the workout level system.
4 Likes
Do you mean, add in tr rides for the next month that would roughly mimic what I’ve done historically (to see what it does to the structured rides) then go back and remove them?
Thanks