Since many of you are interested, I checked in with the product team to verify how TR Outside Workouts and unstructured rides (outside or from other platforms) work. I don’t have a ton of experience with this yet as I was injured most of the winter, and coupled with the weather I have only ridden one structured Outside Workout using AT. It was so seamlessly incorporated into my normal training adaptations that I didn’t even realize until this moment that it was something of note. Anyway…
Structured TR Outside workouts work just like indoor rides, and this functionality is enabled in the current closed beta. Adaptive Training assesses your success on the workout just like it does for an indoor workout, and you’re given the completion survey next time you log onto the app. Currently, I think the survey for Outside Workouts is only enabled on TR web, but it’ll be added to the mobile apps shortly. Naturally, the system is much more forgiving for deviations from outside workout structure, as you’re going to encounter stop signs/ turns etc outside.
Unstructured rides (or structured rides from outside TR) are analyzed by the system and assigned a predominant training zone, and assigned a difficulty level to determine how they affect that zone’s respective progression level. There will be the option to assign the ride to a different training zone if you disagree with the ML’s conclusion, and will likely also be an option to exclude the ride from impacting your Progression Levels. This might be useful in the case of a race, where an all-out effort might be equivalent to a 10.0 workout, but you would not want it to bump up the intensity of your training from day to day that dramatically. Unstructured rides will not trigger a success/fail judgement, since there’s no structure to measure that against. They also don’t trigger a post-workout survey.
Unstructured rides are not currently being analyzed or classified in the beta, but this is our number one development priority and will be coming very soon- definitely by the time full AT rolls out to the public. The system for this works very well right now, we’re just tweaking and refining it to be as useful for the real world as possible.
One open question is if something is changed/recalculated for an unstructured outside ride if you set “associate with” for that ride to a workout (after the fact) or change an existing association.
as I understand it, associating a ride with a TR structured workout always makes that planned workout’s zone and structure the operative one by which the ride is judged. Adding an association after the fact would likely trigger a “missed survey” notification, which is the same thing that happens if you skip a survey now. You could then answer the post-workout RPE survey for that ride and it’d be factored into your progressions.
Kudos to the whole team and the massive move ahead. It makes total sense to do a slow and deliberate roll-out of something like this. What’s funny is in today’s world of “order it today and its on my doorstep tomorrow,” it’s easy to want it now. So far I really like the “Train Now” feature as a smarter way to adjust or add to a scheduled plan. Of course, I believe I am a bit on the fringe of the bell curve.
I’m 100% with you in general. However I don’t see how my cycling performance/need for adaptations in my training plan has anything to do with my ethnicity?
Is there any way you guys can either publish or give us access to the TR workout library with the “AT Rating”?
I see a lot of people asking @mcneese.chad and others for screenshots of the workout library to understand the rating of the different workouts.
This might help while we all wait for our turn to take a stab at the AT.
I am sure the more talented research hounds in here could dig up anything else that might be applicable. In any event, TR likely lacks that info and will need to obtain it first, then use the ML magic to either see and/or apply any learnings related to whatever lies within this question.
The same basic issue exists for many of the other metrics or attributes TR mentioned. Aside from gender, TR will need to get new info from users that is doesn’t have at all right now, before any meaningful analysis or adjustments are possible.
Imagine there is an Ethnicity A that is more likely to have good sleep, good nutrition and less life stress overall than Ethnicity B.
Now if the overall plan design and general progression rate is derived from user base of 90% A and you happen to be B…the baked in assumptions might not be optimal for you (of course you can counteract through the questionnaire at the end and I would assume you would be more likely to need to do so).
But I’m just assuming as this is absolutely not my expertise.
And I guess including the root determining factors such as Sleep, HR, Nutrition etc. in the first place (as is planned) will be the best option.
That looks pretty good to start with!
Sorry, this thread is already quite long, so I hope I am not going to ask what has allready been answered - if so, I would appreciate a hint, because I did not find it
Basicaly, if I would do another sport like Xcountry skiing or skating (althought I doubt it this season, another severe lockdown in my country ), will the AI take the training load into account or will simply ignore it? In the past it was a bit frustrating to see a drop in the training in the calendar when I was skiing instead of riding…
P.S. I know you have to be asked this question millions times per day but… do you have an idea when we could hope to be in the Beta when we subscribed quickly after the announcment? Could this be days? Weeks? Months?
I am fully expecting reductions from Leconte and Write Peak -1, since they are notably above my current PL. Maybe they could happen with a “Stretch” rating, but I expect these to get cut to something “Productive”.
I think Lamarck will remain, since it is basically equal to my PL.
All this might change based on the results from my Ramp test tonight. If I get a different FTP (currently 251w), it will change my PL in some or all of them (if I understand correctly).
So this screen capture will be compared to whatever I have tomorrow morning after my test. I will grab that and report back to see what happens.
ETA - The potential curve ball I have is good weather on the horizon for the weekend, so I may be switching one of those rides to “Outside” and putting that to the test as well.
ETA - Link to my post with Part 2 (post Ramp Test) below:
@rizzi got it pretty right. One of the other industries I was referring to is face recognition. When you train face recognition mostly on white people, it can have poorer performance on black people (even when skin tone isn’t a factor).
The effects for physiology, I’d presume, would be lesser than for race recognition, but there could still be subtle differences in response to training across different ethnic populations. And not just ethnicity, but age, gender, etc. Gender and age (masters racers) was mentioned in the podcast, so it’s on TR’s radar…
I’m a new small-business owner myself, similarly bootstrapped while working another full-time job, AND training, AND raising two little ones. I’ve talked with my wife about some of your insights previously and always love hearing you talk about the business end.
Wife and I came to the same conclusion - we aren’t hiring people that won’t commit for XX hours per week for the exact same reasons. We’re nowhere near the scale you are and never will be, but this is an awesome little nugget - don’t sell out the commitment and “team” principle in the name of faster growth!
Kudos on ML, from a side-gig coach. If it’s as good as you say (I’m signed up for the beta), I won’t mind one bit if I never get paid to coach anyone because of it!
Something that might be a future development would be the ability to create a custom workout from either a fit file or gps route.
Then you could associate the unstructured ride with a ‘known’ workout? It would also give good insights into how hard different attempts of the same course were…
No need. As said…in the (prio 1 near) future there is no need to associate any outside ride with a workout. And to compare your similar outside rides you then can just look at the score that was set for those.
Multisport is really complicated to program for, let alone periodize. If someone like Chad doesn’t have the experience in working with it they would have to seek advice from elsewhere. It’s not something that a cycling coach could just “wing”. Every Tri Coach has their own opinions on how it should be done so there’s a lot to cut through before TR could ever offer something on their platform. I don’t think any platform has ever done it properly, have they?
), will the AI take the training load into account or will simply ignore it? In the past it was a bit frustrating to see a drop in the training in the calendar when I was skiing instead of riding…
Just as the weather turns nice and we can ride outside again and…lockdown!
