Today, TrainerRoad announced comprehensive changes to the platform they called Adaptive Training. This means that they start using machine learning algorithms to change workouts based on your progress towards a given goal. An exercise is recorded when you complete it, and then the rest of your scheduled workouts change accordingly to adapt to it. Although it was announced today, it will not be on most desktops’ desktops for some time.
What’s on your desk today is their new TrainNow feature, which gives you one-time workout recommendations in three workout categories, based on your previous workout history, and the option to choose the duration of the workout. TrainNow uses the basics of Adaptive Training machine learning to generate the training recommendations. So this is your first window into the bigger changes of TrainerRoad.
While this may sound basic, the reality is that with far fewer people having a fixed goal of training or racing by 2020/2021, ad-hoc workouts have become increasingly common for what are otherwise strongly structured athletes. This feature is fully aimed at those who use TrainerRoad, but without a specific TrainerRoad plan, of which TrainerRoad says that there are more users than you might realize (it does work if you have uploaded a plan and prefer the plan to ignore).
The most important point is that many of us want to do a structured workout but honestly can not decide which one we should choose. Whether on TrainerRoad, Zwift, Xert or others – you often get hundreds or thousands of choices without any real guidance, or even basic suggestions. Actually, I would suggest that it is true that platforms like Peloton have done so well: they bubble up the suggested workouts for you and remove all the indecision blockers.
So let’s jump into it right away.
How it works:
To get started, first update your TrainerRoad app and once it’s done click on ‘Workouts’ on the left, followed by the ‘TrainNow’ tab at the top:
You can then choose a specific duration for the workout, from 30 minutes to 120 minutes. This of course corresponds to the specific duration in the TrainerRoad library.
In my case, today I will choose the 45 minute option, and then you will see how the workouts below change to different options. As usual, the thin line in the middle of your workout is your FTP level, so you can quickly measure how much sucking this workout will involve. Or those who are more scientifically focused can simply look at the TSS and IF (Training Stress Score & Intensity Factor) scores listed next to it.
Do not like the options offered? Not your blue style you prefer? Do not like the name of the workout? No problem, press the ‘Refresh’ button to clear the options and get three new options. It’s like pulling the slot machine on your arm, suitable for a business in Reno, Nevada.
Eventually I pressed refresh a few more times until I got ‘Goddard -4’, which seemed interesting. If you look, you see the intensity and the TSS scores for all these workouts in each category are almost identical. This is not by accident. Nor is it something for today’s report.
After deciding on a workout, you can tap on it to open it. You will get a preview with a little more descriptive details, just like usual:
And then you continue to adjust any coaching devices and do the workout as normal.
And then simply go ahead and do the workout as normal. There is no difference at this point – it’s like doing the usual TrainerRoad workouts.
And then you get the summary information, as well as the results in the app and online for analysis there (and as usual synced with other training platforms like TrainingPeaks and Strava):
See … easy peasy!
Exit:
This is a smart move for TrainerRoad, and one I’ve been hoping to see for a long time. As I said earlier, the idea of ad-hoc workouts is not much different from what Peloton does, or in fact – even Zwift. For example, on Zwift when I open the companion app, all the upcoming group rides starting soon are offered. If I decide to take Zwift, I will choose one of the workouts in the next hour 95% of the time, based on the small portion.
However, Zwift does not have the same concept for structured workouts. I mean, if one of the group rides happens to be a group exercise – then yes. But there is currently no recommendation engine in the ever-expanding library of structured workouts in Zwift. And it ignores the fact that it’s a nightmare to find something in the Windows 95-style nested folder scheme that is the user interface. There may be a lot of gold down there, but I just give up and choose the short or long form of Jon’s Mix half the time – since I could find it in the folders at least quickly.
But what’s probably more interesting on TrainerRoad’s side is actually the otherwise harmless text just above the TrainNow workouts, which reads, ‘Let’s suggest a few other options, based on your recent training history.’
This is noteworthy because so far TrainerRoad has not actually recorded historical data except for RAMP / FTP test details, and has done anything with it that is recommendations or advances in nature. Thus, the single rule basically refers to the larger changes to the platform that are adapted by machine learning through Adaptive Training. But that’s another thing or two for another day.
With that – thanks for reading!
[Can’t decide on a trainer app to use? Ask no more, and hit up this massively long and detailed cycling trainer apps guide.]