Another AI training app: Athletica.ai

From HIIT superstar researcher Paul Laursen.

First impressions:

  • A bit of basic interface
  • Strength training incorporated is a plus.
  • I like that the have links to videos for strength training.
  • Appears to use DF1a or HRV to validate VT1
  • References to VT2 for training zones
  • Has rider profiling tests. I like this very much.

I’m trialing for 2 weeks for free just to see.

Cheers

Edit: Don’t be fooled by the basic interface, this app and all the resources around it have very interesting ideas, for instance: a protocol to assess recovery embeded in the recovery ride.

Overall it has a feel of detail and coaching individualization, unlike anything I have seen out there.

3 Likes

Any ideas what hardware they support for assessing these?

Thanks for this. I just generated a plan but I don’t see any strength training?

For VT1 you can ride with me and tell me about your day. Keep your front wheel next to my front wheel. In about 40 minutes, I’ll be able to tell you about what your VT1 is.

1 Like

Not the question I asked…I know how I’d do it, I’m curious how they are

2 Likes

Garmin only

For this test, we’ll use progressive steps of exercise to pinpoint your lower threshold (i.e., aerobic threshold). For this test we use heart rate variability (HRV) during the test. Currently, our system is only compatible with Garmin for this determination, and in order to achieve this, your watch must be adjusted in the settings to enable this. If you have a Garmin watch (i.e., 945), go to settings - physiologic metrics - log HRV - turn on - then use the back key to get out of the settings. You’re good to go. Power will also need to be visible to you throughout the session.
Equipment: Power meter or stationary erg recording power, HR monitor set to collect HRV
Protocol: After a short warm-up, begin riding at 100W for 3 min, and increase by 25W every 3 min. Continue until rating of perceived exertion reaches 8/10. Don’t worry if you can’t get through all the steps written out - most won’t. Cool down.

Mmm. I selected MTB

Thanks - not shocking that they are only using the data from Garmin, but makes me wonder what specifically they are relying on in that data set that other HRMs aren’t capturing or perhaps are just sending in a different manner

Yes, interesting. The description is ambiguous to me. Maybe they meant Garmin watches only. The have running too. I can see a case there. Like no Apple Watch.

they’re going to use dodgy wrist pulse data and your dodgy shimano power meter to guess what VT1 is. Not based on your wrist-pulse-derived heart rate…nope, not sketchy enough…but based on your wrist-pulse-derived heart rate variability.

:smiley:

Trust me, my method is probably 20% more accurate. But I’m probably going to need an e-bike to do it properly.

(I forgot to add, you have to pay for the GIGO service.)

1 Like

That’s why I was asking. I’m very interested in the science and validity of their approach. At best I’m an hrv skeptic so I was hoping I could test their approach in the free trial, compare to what I’m fairly confident my lt1 is. Unfortunately no Garmin for me

The Blog on that site is really excellent! Well worth the reading.

They do have the slowest loading blog page in the history of the internet, though, so there is that. I remember the days when dial up was a thing…this is taking me back.

I’m guessing because Garmin x30 and x40 bike computers record HRV. All my 530 fit files have HRV data embedded in them, starting from when I bought it circa September 2019. Maybe a year ago I used a Python script to go back and analyze HRV data and roughly estimate DFA1 (roughly because I have no progressive step workouts at lower threshold).

2 Likes

Regarding the blog, loading fine over here and now and last night.

This

Is gold.

6 Likes

No issues for me

1 Like

Right now, based on perception, I’m around 200W.

So lets look at their model using this picture:

From this blog: https://athletica.ai/power-profiling-is-the-future-of-training-load-control/

I’m going to do the engineer/math major hand wavy thing and fit a couple curves to my own 90 day MMP:

I did a couple of ‘vaguely similarly shaped’ curves, and it appeared they fixed it to 5-min power on the left, and I wasn’t sure where they fit to the knee of the curve on the long duration.

That was perception and 90-day power curve, here is the steady endurance portion of a ride from 4 days ago:

  • steady and slow breathing
  • HR slowly rising and completely in sync with slowly rising power
  • could have been talking to someone

Three months ago my HR would have been trending around 138-142bpm at that same power, this one and last several have been on warmer days and trending around 133-137bpm.

Some interesting ideas that I haven’t previously used as a lens to view my data.

Isn’t omnipower very similar to Xert’s thingy?

Just looked through it. This is really a great source, thanks.

Looking more into it, perhaps Xert will be out of business soon or it will be even less relevant than today. They were really slow to develop and now they are left way behind.

I’m picking you’re trolling as much as anything, but the Xert people hinted at quite a few “features” imminent.

I don’t really know the in’s and out’s of how either of these platforms work, but it hardly seems like either shifts the needle that much.

There are already a bunch of platforms doing the “constantly adapting” thing. No one has the whole package sewn up yet.

1 Like