Like this site? Help us to make it better.


TrainerRoad launches Adaptive Training designed to remove the guesswork from improving fitness

New software “uses machine learning, science-based coaching principles, and an unprecedented data set to train every cyclist as an individual”

Indoor training app TrainerRoad has released new software that it says automatically adjusts to external factors such as a missed training session, time away from the bike or an increase in functional threshold power (FTP).

TrainerRoad says, “Using an advanced, machine learning-based system, Adaptive Training analyses cyclists’ performance in every workout. It uses these insights to automatically determine Progression Levels, a real-time reflection of relative abilities across power zones.”

Generally, the process of setting yourself a training plan involves setting goals or an event to work towards, working backwards to determine weekly or monthly micro targets, and then drilling down into the workouts that need to be performed each day to take you to those goals. It takes quite a bit of time and ends up looking very smart on a calendar, but then life tends to play its usual trick of getting in the way. Those carefully laid plans can quickly unravel and it is this that TrainerRoad’s Adaptive Training feature is looking to address.

> Review: TrainerRoad Online Training 

Training sessions have all been ranked within their category - be that sprint, VO2, threshold - by TrainerRoad to give the rider an indication of the session’s difficulty.

TrainerRoad Progression Levels

Adaptive Training works, TrainerRoad says, by analysing the rider’s performance within the various power zones. Should the rider be struggling to complete a set of VO2 max intervals, for example, the technology would, it is claimed, suggest that the next VO2 max workout be easier.

When following a TrainerRoad training plan, the software will suggest adaptations that take into account the rider’s recent training, form and recovery. If the rider is improving quickly, then Adaptive Training will suggest a harder replacement session. This, TrainerRoad says, will be of equal duration to the one it is replacing, but have a higher intensity factor (IF) that results in a higher training stress score (TSS).

TrainerRoad Plan Adaptation 2

The rider can choose to accept changes to their training sessions or give the system full autonomy by always adjusting the training plan automatically.

TrainerRoad also says that they are working on outdoor rides being given their own difficulty classification, which should help riders who mix indoor sessions with regular rides as the system will take into account the rider’s long, unstructured weekend rides instead of assuming that they had two rest days. This feature is still in the planning stage though, with no release date set.

TrainerRoad says that individual aspects of the new technology will be rolled out gradually though in due course it will be available to all users in all training plans. Currently in a close beta phase, Adaptive Training will be rolled out to users in batches and TrainerRoad is clear that they expect some bugs to be found in the early stages.

Add new comment


PRSboy | 3 years ago

I suppose Zwift kind of does as it changes your FTP based on previous efforts, though this looks a good deal cleverer. 

I do TrainerRoad workouts whilst riding on Zwift for the best of both worlds.

Secret_squirrel | 3 years ago

Excellent.  Now can we have the same in Zwift and SUF please?

Jetmans Dad replied to Secret_squirrel | 3 years ago
1 like

Secret_squirrel wrote:

Excellent.  Now can we have the same in Zwift and SUF please?

Genuine question ... why? Why do we want all of the platforms to do all the same things?

I Love Zwift and have racked up around 7,000 miles on it over the last three Winters. I like the immersiveness of the gamified experience, I like riding through a volcano (because I am weird), and I like having the flexibility to do some racing and some simple workouts/training plans when the fancy takes me. The TR approach doesn't really interest me, but I know riders who swear by it and don't use Zwift at all. Horses for courses. 

As a Zwift user I would be quite disappointed if they went down the road of trying to emulate everything all the other platforms are doing, particularly when there is already long-standing discontent among users for promised feature updates that have not appeared for literally years. I imagine TR users would feel the same if they suddenly decided to introduce some virtual worlds in an attempt to do the same as Zwift. 

Surely it is better to have several different packages that have different focusses, with as great a user experience as possible in each case, than several packages all trying to do the same things and not doing any of them really really well. 

Secret_squirrel replied to Jetmans Dad | 3 years ago
1 like

Why take 2 training apps into the shower?

If there are themes that work with the unique parts of each app I would agree with you.   But these are universal features that would work for any Cycle training app, be it Zwift or even Training Peaks.    There are 2-3 running apps that already do this so its just a natural extension of existing features.   

I find Zwifts inflexibility when I miss a training session or want to reschedule one of its most annoying tbh.  YMMV.

FlyingPenguin replied to Jetmans Dad | 3 years ago
1 like

Because it would be a massive improvement on the (frankly lackluster) Zwift training plans and it's applicable to any app that has structured training? SUF has 4DP which captures strengths in a bit more detail than straight FTP, but no platform yet helps combine indoor and outdoor riding with ongoing analysis of form and performance.

If your indoor riding consists entirely of unstructured riding/racing then you'll never interact with it, but training plans exist in Zwift already (and I've done several of them end to end), but compared to the competition they're just... a bit shit.  

Also, from Zwifts perspective, they're still working on the structured training side (see the steady trickle of plans and workouts added), so they evidently see value in that market segment *shrug*

PRSboy replied to FlyingPenguin | 3 years ago

Just one simple change would improve the Zwift workouts... the ability to search for workouts per TSS/IF/zone for one, rather than time.

There are probably some more pressing Zwift issues though, such as being able to change 'world' without having to exit and restart the game!

@flyingpenguin... for what its worth, TrainerRoad does integrate outdoor riding data into your TSS history, and Strava does via the fitness/freshness score. 

Latest Comments