What do you think about adaptative training? I have heard a lot about it since like ever but never understood it
In this context, adaptive training is just software doing the job of changing the plan to the athlete rather than the coach. Very easy to say, very difficult to do.
Looks like they are starting with running first. Definitely curious to see where this goes.
3 Likes
On another related note I hardly seen any updates to TrainingPeaks since like almost last February (cf: What's New with TrainingPeaks: Feature Updates | TrainingPeaks), I am definitively switching over TR calendars which is more than enough,
2 Likes
Yeah and those “updates” from them are almost all big fixes going back to 2018. Not a great strategy to hold on to your premium subscribers.
I really want to get on TR’s calendar (I’ve tried a few times), but I need runs to be imported from Strava or Garmin Connect before it becomes a useful enough tool to replace TP.
8 Likes
I think Strava and TrainingPeaks have the same product development philosophy: just say No to new features 
5 Likes
Hey now, gotta give TP some credit that at least they fix bugs and add a few things from time to time.
Strava’s strategy is to just keep removing features, copy anything remotely good built by an outside developer and then revoke their API access.
11 Likes
Not to mention that expecting a steady stream of features is like saying don’t work on anything that takes more than a few months to get right.
1 Like
Usually when that happens when they are in maintenance mode, it means or that they work on a brand new version or that they are in “Due diligence” mode to get bought (altho i doubt that goes over 8 months period)
The next generation of training plans will be adaptive, taking into account lifestyle, work, family and overall athlete availability rather than a “one-size fits all” approach.
…
Over the last two years we have been doing the R&D on an adaptive engine to drive the future of your training plans.
Kinda sounds a lot like what the TR podcast has been hinting at for months.
The desperate attempt to frame it as “we’re trying to help coaches, not replace them” had me giggling a bit:
New technologies will provide the coach with more coaching tools to help them make faster, better decisions. We want to embrace the machine to do things that make the coach the hero.
…
However, we are not forcing coaches to create solutions to solve the needs of hundreds of athletes if they don’t want to…Coaches should be given the ability to spend less time on the computer, and more time with their athletes doing what they do best—listening, understanding, empathizing and strategizing. In the future, TrainingPeaks will function like your own assistant coach, freeing up your time to do just that.
1 Like
I know it’s not auto, but just checking you know that they can be added manually. It’s super-easy to do and takes no more than 30 seconds.
1 Like
Brilliant, looking forward to seeing where they go with this and how powerful it becomes. Ultimately it’s a machine learning coach that I want. Something that can crunch years of data, power, HR, performances etc and create interesting plans for the weeks ahead.
Xert kinda does it but it’s dull as dishwater and it can’t really generate a great plan for the month ahead.
Fun and entertainment is also important to me Neo, more than before. I’m more more likely to smash it up Alpe Du Zwift instead of a prescribed 2x20 sweetspot or threshold workout.
I can do greater workload and actually enjoy it that way.
As for someone developing the next gen of adaptive plans in a polished way… I think TrainerRoad will nail it first.
1 Like
TR is still struggling to get a calendar with features TrainingPeaks has had since day one and it currently has the most rudimentary data analysis tools of any of its competitors. As far as i know, while TR has computer programmers galore, it has no actual credentialed sports scientists on staff or as paid consultants. That is way more important than programmers in coming up with an adaptive training program tool.
TrainingPeaks, through WKO, has the gold standard in data analytics and data modeling and has direct relations with the scientists and coaches who played a big role in developing and applying the standard data analysis and modeling tools and theories. WKO already gives data that can be used to modify training plans. That is what it is used for by coaches now. They already have their iLevels a/k/a individualized power targets. That stuff the hard part - how are you going to modify the plans and why? - automating the output is next. TP as a business cut its teeth working with professional teams, including most of the World tour teams and its primary business is serving coaches, including many of the best coaches. It is directly seeking assistance from this brain trust for its development of adaptive coaching tools (See the end of the press release).
So, between TR and TP, who really has a better shot at this adaptive coaching thing?
11 Likes
Having seen how wildly off-base TP / WKO5 has been with its supposedly super-smart tools that don’t work unless you dedicate your life to a never-ending series of timed power tests, my hunch is that neither platform will be able to provide anything close to what a personal coach (or even yourself with a little education) can provide.
It’ll either have to be a one-size-fits-all system that may require human intervention to further individualise (like the TR ramp test), with anything more smart / personalised being a massive over-promise.
3 Likes
I don’t know why the simple use case needs to be that hard, why don’t they go for the very simple use story :
I am off travelling/sick for next week please adjust my plans,
by changing the intensity the week after or adjust my goal race,
automatic adaptative training is niice and all but I think most people would be happy to be able to have some smartness that goes beyond a static training plans of x weeks
2 Likes
If only there was a podcast that could empower individual athletes to evaluate their performance and off-the-bike factors to make informed decisions about adjusting their training…
19 Likes
Adaptive training plan “taking into account overall athlete availability” sounds like nothing more than a calendar that will shift sessions when you’re too busy to do them. But, first of all you’ll have to tell it that you’re too busy, so you’ll either have to input tons of extra data into TP, or maybe they can pull it from your google or outlook calendar. If you’ve entered it there.
I used TP for a bit to train mostly be TSS, but imho while it was easy to plan how much work I was doing, it didn’t take into account what type of work I was doing. So I’m not that convinced they’ll have some sort of AI coach working soon. (I know that wko has analytics that would help with that, but there’s currently still quite a big involvement of a human expert to look at the data and deduct what to do from there.)
Agree. I’m probably a bit too hard on Training Peaks / WKO - I do recognise that the opportunities for data analysis and visualisation they offer are second-to-none.
But I also believe that the coaches looking after the top athletes are spending a LOT of their time doing their own analysis to adapt their plans and respond to individual athletes. Someone correct me if I’m wrong, but I’m pretty sure that Quickstep’s coaches aren’t just automatically prescribing intervals based off whatever iLevels and mFTP WKO spits out.
Agree also. If you want true individualisation, pay for a coach or learn how to coach yourself.
1 Like
There is a ton of low lying fruit in this space. Folks jump to fine tuning the daily workout but that’s a huge ask. And probably doesn’t even matter much to most of us who ride bikes and try to be good at it.
Two posits:
Most coaches stink. Many do little more than prescribe cookie cutter workouts and send encouraging emails. The set plans like TR crush most of these “coaches”. Of course there are amazing people training athletes, but those cats are very rare and often very expensive. Out of reach for Joe athlete and Joe athlete probably can’t maximize value of those interactions anyway.
Many folks get the major aspects of their training wrong. They increase AC work when they really need long Z2. They add volume when they need rest. Far too much focus on small gains when the broad strokes are wrong. It’s like worrying about painting your home’s trim a new color while the foundation crumbles.
If we start there, a “less dumb” system that can make “less bad” recommendations on what type of training to do next for say 4, 8, 12 weeks will add big value.
Look at the current success of trainer road. The platform gets the basics right. If bike riders simply cycled SSB1 to SSB2 to General Build they would do better than most. Now add in a bit of lift by having an ML/AI engine suggest tweaks, say a 4 week block of short power build to boost up some lagging metrics on the power curve, and you have a good iteration of an already good training platform.
A good engine with enough data might even suggest what training effects to expect. That could help a lot with managing expectations and driving consistency.
Am certain that ML/AI coupled with some expert curation is absolutely going to provide value. But I don’t think it will be at the minute granular level that most folks focus on. Rather, it will help people get the basics right.
Making up some numbers… if a current system gets 90% of potential performance from an athlete, increasing that to 95% is big lift.
$0.02
11 Likes
Suggested amendment: most coaches’ cheapest training packages stink.
Definitely agree with this. Interesting you leave out Specialty, which is my main gripe with TR - I think it’s just too much intensity leading up to races.
4 Likes
I would bet on strongest development team, not the one with the most sports scientists. Its not easy to turn spreadsheets and scripts held together with bubblegum into something deployable for random athletes to use. So if TR has a strong development team I think they would have a better shot.
1 Like