Athletica AI Coaching review - Part 1 (weeks 1-2)

A bit of background on my online training

Over the last 25 years or so I've found that indoor training is the key to improving my fitness and performance in time trials. Away from the vagaries of dealing with terrain and motor traffic, I can focus on the workout rather than survival. 25 years ago, power meters weren't readily available, and one's focus was on heart rate as a measure of exercise intensity. Indoor trainers offered adjustable resistance, but little else. In those days I found the structured plan in Pete Read's The Annual Manual (the Black Book), which presented a month by month plan for training on an indoor trainer, to be highly effective.

TrainerRoad has been the mainstay of my training over the last few years. I’ve found it pretty effective, particularly since it moved towards its adaptive training that uses machine learning to proactively modify upcoming training plans. I also like the way it estimates FTP on the back of recent activity rather than by an FTP test (which I always find difficult).

TrainerRoad has a very intuitive interface, and has apps for iOS and Android that provide a very basic interface of a histogram type depiction of the interval session to be conducted. It’s not a very engaging way to do indoor training, and as a consequence I usually undertake TrainerRoad sessions in parallel with another app such as Zwift, FulGaz or Rouvy to offer a bit of visual entertainment. Overall, while my experience with TrainerRoad is that it steadily increases fitness (at least as judged by rising FTP), it does have some downsides.

On the potential downside, while TrainerRoad gives you an indication of the type (e.g. Sweet Spot, Threshold, Tempo) and intensity of a session (via Progression Level and the TSS score of the session), it’s not terribly obvious what activities contribute to adaptive changes to the training programme. I’ve read, for example, that outside rides don’t make an important contribution to adaptation, though they do to the AI FTP estimate. Also, while there are thousands of different training sessions in the TrainerRoad library, many of these are pretty similar, often leaving you with a sense of of deja vu as you once again grind up a series of escalating steps. On top of that, I often find that some of the tougher sessions are just a bit… well… tough!

As a training app, TrainerRoad isn't strong on recording a variety pf physical parameters. Over the last few years, I've gradually added more measures to my records, using the excellent intervals.icu web app. This allows some serious data wrangling around my physical condition and the effectiveness of my training. If you like graphs and data, this is the web app for you!

So I have an inclination to have a bit of a refresh by trying another self-coaching app, athletica.ai

Athletica

I came across Athletica a few weeks ago, and I was quite intrigued by the way the system rested on the work of a sports science academic (Prof Paul Laursen) and lifted by the use AI or probably more like machine learning approaches in analysing workouts, revising training schedules, and providing feedback. The basis of the training philosophy is in the use of High Intensity Interval Training - Laursen has co-edited a book on the subject. 

Starting out with athletica.ai is a bit like engaging with other training platforms - you enter various bits of information, including your aims (what you are training for) and the volume of training. You can integrate with Garmin Connect and Strava. Athletica provide a detailed description of the platform via an e-book in pdf format - The Science & Application of Endurance Training Using AI-Platforms, which covers this in some detail.

The main screen that you’re presented with is the usual calendar, showing week by week, with your prescribed training sessions indicated. Athletica 01

The Athletica calendar screen (click to enlarge)

The workouts are defined by whether they are indoors or outdoors (sometimes you can choose) and by a brief description of the workout along with its goals. Clicking the hamburger menu in the purple bar brings up a menu, and when the workout is complete I select “Add Data”. In the dialogue that pops up I upload the fit file (because I haven’t linked my Garmin Connect), and fill out a few questions, typically RPE and a note on how I felt. Providing a commentary on how the workout went gives a bit more grist to the AI mill in generating a response. The rest of the data populate from the fit file.

Athletica 02

A completed Athletica workout (click to enlarge)

The workouts with a green panel are those where I met the objectives, you can see a summary of data showing how I did in the green panel, with a brief synopsis of Athletica’s AI response. I actually find these little snippets quite nice, even though I know they are computer generated.

After adding data to the completed workout, there’s a short wait while the system processes all this and generates a response. Once that’s complete, you can see the outcome by clicking Details & Analysis, which brings up a window like the one shown below.

Athletica 03

Detailed report on a completed Athletica workout (click to enlarge)

The blue panel shows the detailed feedback from the system:

You executed your workout beautifully, completing 229 minutes with an actual load of 190, surpassing the planned load of 166. Your average power of 157W and heart rate of 145 indicate that you maintained a solid effort, aligning well with your goal of enhancing aerobic endurance. With an mWR of -24%, you pushed yourself to a maximal effort, showing significant improvement! Your FTI of 70% suggests you effectively trained at a moderate intensity relative to your threshold. Overall, your RPE of 4/10 and feeling of 4/5 indicate you managed the session well while still challenging yourself. Keep up the great work!

I get that this is produced by an AI engine but it does feel quite encouraging - a clear verbal construct wrapping the data metrics reported in the fit file.

My early thoughts on Athletica

So far, so good. I’m enjoying the explanations of the why and the how of each prescribed workout, and the textual analysis of completed workouts. Of course there's a bit of a learning curve when adopting a new system, but Athletica supports the user with loads of explanatory material.

Through the free trial period, I was playing with automatic downloads from Garmin Connect, but I found that was a little confused because I’m kind of phasing out the use of TrainerRoad but some TrainerRoad rides still get in there, along with recreational rides.

The training calendar is quite dynamic, and I’ve noted some last-minute changes - for example, the road ride workout above changed from a 3h30m ride to a 3h40m ride the day before - presumably in response to prior workout data. I really like the detailed explanation of a workout's aims and even better the detailed critique of the workout once completed. There's a real sense of an integration between indoor and outdoor workouts. In my locale, it's completely impractical to ride a HIIT session outdoors, and to have lengthy road rides decribed in terms of a broad power zone is actually quite helpful.

Riding indoor workouts presents a bit of an issue - athletica.ai doesn’t itself provide an interface to control a trainer but instead suggests partnership with a couple of other applications. I’ve chosen to stick with TrainerRoad for the time being as I’m familiar with the app and its use. To do this, I download an erg file of the workout and import that to TrainerRoad (via the Workout Creator app). That’s a bit circuitous, and clearly other options might be more practical - perhaps downloading the workout directly to my Garmin Edge computer might be one route.

Ultimately the proof that Athletica is going to be a useful tool in improving fitness will take some time - will my various metrics show improvement over a period of months?

Athletica seems like a good choice if you're interested in self-coaching, but I've noticed some websites advertising coaching services (for example Simple Endurance Coaching) that use the Athletica platform - the human part of that equation presumably offers a more person to person approach. 

Expect part 2 of this review when I've completed a month or two of training with Athletica.

×
Stay Informed

When you subscribe to the blog, we will send you an e-mail when there are new updates on the site so you wouldn't miss them.

1001 Albums - 221-240
Old school tandem refurbished

Related Posts

 

Comments 3

Guest - Chris (website) on Saturday, 28 September 2024 08:13

Nice write up. The book recommendation (Pete Read's The Annual Manual (the Black Book)) sounds really interesting but I cannot find it anywhere. Google / Amazon come up with your site + a forum post, but not much else. By any chance do you have an ISBN or some other way to track it down?

The little AI feedback snippets also jumped out at me. I thought they were nice, obviously AI, but nice. Sounds like they are doing it right if you thought the same.

Good stuff, thanks for sharing.

Nice write up. The book recommendation (Pete Read's The Annual Manual (the Black Book)) sounds really interesting but I cannot find it anywhere. Google / Amazon come up with your site + a forum post, but not much else. By any chance do you have an ISBN or some other way to track it down? The little AI feedback snippets also jumped out at me. I thought they were nice, obviously AI, but nice. Sounds like they are doing it right if you thought the same. Good stuff, thanks for sharing.
Robert on Wednesday, 02 October 2024 10:39

There were some bootleg copies of the Black Book around the internet. I don't think Pete Read's manuals were formally published - one used to write to him with a cheque! He passed away some years ago now.

There were some bootleg copies of the Black Book around the internet. I don't think Pete Read's manuals were formally published - one used to write to him with a cheque! He passed away some years ago now.
Guest - Chris on Tuesday, 08 October 2024 07:06

Ahh, OK thanks.

I have a friend who sold word processing software in a similar fashion long ago

Ahh, OK thanks. I have a friend who sold word processing software in a similar fashion long ago :)
Already Registered? Login Here
Thursday, 21 November 2024

Captcha Image