I’ve been having real issues getting it to show reasonable suggestions.
I try to set the daily availability but it never saves.
The system for choosing a workout that matches the required XSS etc is still clunky.
Have you reported it ? It is in beta
I simly go to a day wth forecast on it, set the availability (although the detected has been pretty close) set the amount, set … set as default for that day, and all good
I can only set it for 2 minutes to 20 minutes.
Disappointing as I wanted to set it for two hour power.
1 Like
I don’t think that’s how the Xert modeling is supposed to work, 20 min is GC in order to deal with 2 minute climbs, but the rest of the training is supposed to be defined by your athelete type, but like I said before, not a Xert user, just liked what I saw so far, you’d be better to ask in the Xert forums, unless TR is already doing for you what you want
My understanding is that they removed anything over 20 minutes as a target.
Apparently very few users had that as a target.
I always had it set 1 hour plus… Grand Fondo or Triathlete (3 hour power)
1 Like
I am using the Xert Forecast AI together with TrainerRoad. I do all the workouts in TrainerRoad as the workout player is much nicer (e.g. syncs with Garmin, runs on MacOS, doesn’t require a combination of phone, and a web browser to see it on a big screen).
My thinking is to utilize Xert to manage my load, especially when going into the outdoor season. I like TrainerRoad as it makes it very easy to do workouts and create workout plans. However, the plans are not adaptive even if TrainerRoad brands them as such. It will not adjust to your availability or which workouts you have done. With Xert you will be asked to increase the duration if you are behind the plan, take a rest day if you are ahead of the plan or too tired. It will adjust the recommended workout type based on what riding types you have done so far. So if you are ahead on the hard workouts it will ask you to do endurance workouts instead.
I set up my calendar for the coming seasons with my planned events, vacations, and work trips. Xert then adjusts the workouts around these days so my overall load targets are met.
2 Likes
This is the baseline for Xert
Yeah I was disappointed they removed the only targets I care about. But I guess 20 minutes is a good enough proxy power wise. From what I can tell you can still set your athlete type after using the forecast ai, and if you set it to ironman, et al, you’ll still get your relevant metrics on the training/progression screen. (threshold, hie, pp, 3+ hour and lower threshold.)
Cool, so probably around 2030 then based on PL2.0?
3 Likes
Yeah, I know it’s more complicated than this but logically I’m thinking.
- If you can predict my ftp based upon what I’ve done…
- Project into the future 3 rays of confidence about what I’m meant to be doing.
- Allow me to select how conservative I want to be in my confidence.
re-predict my outcome based upon changes (for example, if you raise my levels and adapt my training, change my target).
voila. take my money pls. :throws_money:
It’s really not too complicated to try to model yourself and estimate individualized parameters. I always like to share this paper and spreadsheet.
Clarke and Skiba provide a rational DIY method for doing so…and for FREE.
You can use this spreadsheet, from Clarke and Skiba…tableS1
You can also use Golden Cheetah for Banister modeling.
The workflow is like this:
-Establish CP (FTP) from 2+ time trials. Verify CP (FTP) with RPE. No goofy approximations.
-Perform regular time trials during training to compare modeled performance to actual.
-Collect enough time trials (6+) to tighten model parameters and get SEE of the model
-Once the model parameters are optimized, you can then forecast future performance by your planned TSS, XSS, scoring units, etc.
I’ve been running more this year, and the model has been surprisingly useful.
Is this method going to be able to estimate to the watt or three? No! but it gives an idea of how the future training could impact future performance.
3 Likes
If people thought Xert had a learning curve, GC is like quantum physics. I tried so hard to learn how to use it and it’s just too complicated for me.
2 Likes
I was curious and signed up for xert (again) to play with the forecast AI.
I’ve put a goal and a time of about three months, and let it make a training plan. I’m not quite sure of the plan - I haven’t put in any time constrains, but the result is a bit weird. The weekly hours are reasonable, as is the length of the sessions, but its grouped them quite weird. Sometimes there are four days of training in a row (usually 1-2 endurance), other weeks there’s four rest days in a row. I’m a bit out of touch with xert, so not sure if that’s normal.
I might try and follow the plan, but think I’ll have to give it a month or so, as I’m still in cross season and the super-high intensity workouts xert wants me to do are not on the cards right now. I might also need to feed it some more data, because the fitness signature it has calculated from the data I gave it has resulted in too high an FTP.
I’m going to keep an eye on it and play with it some more.
Edit: I’ve recalculated the forecast…14 days of training in a row, then 2 days off, 1 day on, 4 days off…umm.
@splash That sounds … thrilling.
2 Likes
I’ve done it again and it’s come up with something completely different yet again. I’m not convinced it works, tbh.
It also keeps showing my target power and signature wrong in some places.
I have a deep curiosity for programs that give new results every time a calculate button is clicked.
Crunch the Numbers (youtube.com)
4 Likes
@JustinDoesTriathlon …same
so if you look at the planner, what do the stats actually say about the weeks themselves?
I know that xert ‘fatigue’ is a rolling window not a gmt kind of thing as well, like you might be red after completing your workout and ‘blue’ 3 hours later depending on the workout. Depending on the recalculate timing you might be giving it different starting conditions. Not defending it, just curious.
I’m personally fascinated by it’s output. the way it calculates xss, the ramp rate, and hours. It isn’t exactly what I would expect, but it does follow something that I would call an intuition about myself.
I can’t really compare between the different versions, because I didn’t save or screenshot anything before recalculating. The version that started with 14 days continuous training definitively made those days short and easy (30 min endurance) at the end. It could be that the total ramp rate was similar to a version with less, but more intense days? But still as said above, you expect some sort of similarity between calculations, even allowing for AI fuzziness and the fact that there might be multiple solutions.
I’ve been reading on the xert forum that you’re supposed to create a forecast, then set your time constrains, and adapt the forecast. I haven’t set any constrains so maybe that’s why its varying so much. It’s still a bit concerning, lol!
1 Like
The other thing I just noticed is that it is ‘soft-pedaling’ until the critical point where it’s trying to build to my goal. 
Guess I wasn’t aggressive enough in my goal setting. haha