Listening to this week’s Ask A Coach podcast, my understanding is the post ride survey “Easy/Moderate/Etc” will be adding some sub-divisions within each main RPE to fine-tune your post ride assessment. Something like:
Easy (Low/Med/High); Moderate (Low/Med/High), etc
My initial thoughts:
I’m not sure it was a problem needing fixing. For me I either immediately know which bucket I want to label it IE: Moderate / Hard, or its on the bubble and as mentioned many times by Coach John i need to “be consistent for ME”. For on the bubble workouts, the issue IMO is literally that I’m struggling to decide "was this Moderate (High) or Hard (Low) - actually having those options doesn’t help IMO. I decided early on to use HRmax as my tie breaker.
The biggest distraction its going to make for me is now EVERY workout is a “bubble” decision. Even if I immediately know its “Hard” - was it Hard (low/med/high). If I just always use “med” when I know the major category, then it doesn’t seem any different from just having 5 options in the first place.
I’m sure there is more going on behind the curtain trying to help us all benefit by the AI learning better, just not sure this will help and it may end up being harder on all of us doing our post-ride survey.
Also thought this new RPE scale was interesting given the noise the 1-5 scale has - which isn’t that hard. My first thoughts were this is a chance to align with TP and Zwift with a 1-10 scale and kind of makes the most sense.
Perhaps the idea is the finer scale will dial in AI workout selections. Obviously open to testing either way.
They have a model that is built on a 5 point. This is “one” of the logical approaches to move it off that scale and retrain the model over time, because they cant just switch the scale to a better 7pt scale (or 10 for platform alignment).
Its a mid-term approach to fix a problem they have which frustrates users and loses subscription.
I’m with you on this. What is the difference between MH and HL? The answer is probably what degree of harder or easier workout you get the next time, but it’s just more confusing for me, and doesn’t seem to really solve the problem.
This reminds me of a project I once worked on where I was working with a group of managerial end users and I asked them to rank the backlog items as High, Medium, or Low priority. They came up with a list of something like 20 high priority items, 2 or 3 mediums, and zero lows. I pushed back and they wouldn’t budge, so next I had them rank the High items as HH, HM, or HL
For TR, will we now have:
Easy (H,M,L)
Moderate (H,M,L)
Hard (H,M,L)
Very hard (H,M,L)
All Out?
For TR, I personally would have preferred moving to the “aligned” 10 point scale.
I’m wondering if more options on a perceived exertion scale is going to make the AI better or worse. Surveys on subjective things (which RPE is) have a sweet spot, probably some subconscious thing in our brain, where 3 isn’t enough and more than 5 (and especially 10) starts to produce a difference w/o distinction. I suspect if it goes to a 10 point scale, there will be numbers which are used so infrequently the “effective” scale is still around 5 predominant selected RPEs. AI query says Zwift users aggregate into 3 blips - 2/3; 5/6 and 8/9, indicating that “having” 10 subjective options isn’t used even when we say we want the granularity. Will be interesting to see what they roll out.
I was thinking experientially on surveys with 10 scale how there are just numbers I never use, and wondered if Zwift’s 10 point scale has “rarely used” numbers. AI query says users generally never selected 1, 4 or 10 and the workout RPEs clump in 3 blips. 2/3 “easy/Z2”; 5/6 “temp/sweetspot”, 8/9 “very hard to all out”. So despite having, and maybe even ASKING for 10 options - IMO we will all tend to self-regulate into a 3 to 5 point scale functionally anyway.
I believe you’re right that the end-state is how the AI builds/modifies future workouts and the problems Coach John discusses on the pod seem to be more realted with people not being internally consistent so the AI gets confused.
I really like the idea of a more granular scale - thinking more of it as a sliding scale with a few notches between hard, very hard etc so it’s less binary between the ratings. I find now, if I rate a workout that I find “pretty hard” as very hard, AI knocks my next week of workouts down to levels below me being challenged, but if I rate it as hard it gets them just right, so I personally feel like I’m adjusting my rating to get the right future workouts.
I also get the comment around prioritizing and adding more buckets doesn’t help but I think AI has the capacity to manage and small differences in a future workout produce big training gain differences (I’m up at least 8 watts comparing similar workouts from 6 mos ago).
A separate issue is sometimes a workout is hard because of leg fatigue but not because the workout itself would be hard on fresh legs. I wish I could delineate between the two without having to get to a failed workout survey to do so.
I agree with your point on people not using the full scale, but I’m all for consistency and integration, and since others use 10, that makes the most sense to me. Just like I wouldn’t have called the new training number FTP. I just want more clarity and less confusion.
Your last paragraph would probably provide the AI more usable info to help understand the existing RPE scale. I lift legs a couple times a week, and play ultimate frisbee on Thur & Fri. It -definately- affects my RPE and has nothing to do with how my cycling fitness is going. These non-cycling workouts are in the system, and I have a consistent Estimated TSS i’m using, but the AI really struggles with Yellow/Red day recommendations. Despite logging about 50% of my 6 wk avg TSS on those Thur/Fri ultimate frisbee nights, the system often just gives me a yellow day. If I did the same TSS biking, I get a Red day. Not a complaint on TR - its a cycling program, but having ability to add meta-data to my RPE may help the AI better understand my data.
I seem to recall research that people aren’t good at scales with more than 7 responses.
I think it makes sense to have an option to split the difference between any of the current RPE choices. Halfway between medium and hard makes sense. Some rides it’s hard to decide which fits best, and choosing the middle would be great.
I don’t think much more detail than that would be helpful.
I do think there needs to be some way for the user to flag rides that feel way off, either far too easy or too hard. Something to adjust more quickly what the model thinks you should be doing. Feels bad when assigned rides are in the wrong zone because you hit an edge case.
The issue is the TR scoring doesnt really stretch and align well with the words Zwift uses for their 10 point scale.
Even I dont align with the descriptions TR uses to express with their words. It isn’t until you use the description vs the word that it actually works. When they first released the integration, I was always rating through Zwift and it was 2 months until i started using the Specific descriptions of the 5 point scale
I find the complete opposite. I frequently don’t know which bucket to put it in when the only option is moderate or hard. Those words to me feel like very different perceived efforts. A lot of workouts in Base phase feel like they are right in between moderate and hard. I actually like the Zwift descriptor of “slightly challenging” - like it was kinda hard but not “hard”.
Thank you for saying what I’ve been thinking. My mind has truly been blown by the amount of digital ink that has been spilled in the last few years discussing the simple question, “how did this ride feel?”
Kinda funny watching folks get all bent out of shape over the idea of more granular RPE ratings between the usual 5. Just don’t use them if YOU don’t need them. No one is taking anything away from you, its just adding the option of more nuanced ratings for people who would like it.
