What you’re describing is something that is not common at all. Care to provide an example from published literature? Only one I’ve seen is in elite marathoners, but never for cyclists.
Yes, it’s not common. Only elite cyclists. But that’s the goal standard we (ppl who ride long distances) should aspire to. I’m sure you understand the constrains of an academic setting, which guaranteed something like this will never get studied. Q
Which is moot since they’ll be getting full metabolic testing.
1 Like
Yeah I see what you’re saying about LT1 (not its importance but all the back and forth about how to measure it).
I think a lot of ppl (myself included a few years ago) aren’t actually counting breaths as much as trying to “feel a significant change in respiration”.
Truth is, I never actually FEEL the change in real time until it gets well beyond LT1……like panting level. Not very helpful, and as a result I gave up on the breath test.
Once I plotted breath rate (by actually counting breaths) against Pw:HR it was very clear. I was measuring lactate then and it matched up. DFA matched up as well but counting from one to five for 15 secs and multiplying by 4 is easier than all the other stuff
Now I just count my breath, triangulate with HR and (to a certain extent) power and feel smug. It’s all very reassuring. Lol 
Like all the things we measure, it’s one thing to know it, but quite another to know what to do with it. Train just above, right at, below? How often? How long? What will the internet think?
1 Like
Quick hint. You only need to know breaths vs cadence, unless your cadence is all over the place that will be directly correlated to breaths/min.
Like in ERG mode? To clamp HR and power? Otherwise there’s no correlation
200w at 90rpm, my breath rate is different than 250w at 90rpm.
I mean you probably use AI all day every day and don’t realize it. I bet you use Google Search and translation services all the time. And live transcription…there’s a lot.
On the TR side are we not categorizing your workouts correctly after the workout is done?
But, I TOTALLY agree with you that the term is overused and it’s often a bit of marketing thrown on something that isn’t using some version of AI.
I should have used the quote feature, just talking about an easy way to remember breathing rate besides counting for 15 sec and multiplying by 4.
Yes, I would definitely expect rates to be different in those two examples, but I have a pretty good idea what my rr is for most workouts just be gauging against cadence. 6 in 6 out, 3 in 3 out, etc…
1 Like
We’ll probably just record one HR stream with timestamps then line them up with a workout after the fact. But yah, we’re on the same page.
1 Like
I gotcha now. Yep. 
I was explicitly referencing the service AIEndurance, that after loading my data, populated a calendar with generic workouts 6 days a week for 2 months.
On the AI thing…Im a user and creator of algorithms. I prefer when people use the proper names to call things instead of umbrella labels. For instance, check Deepmind’s description of Alpha go:
We created AlphaGo, a computer program that combines advanced search tree with deep neural networks. These neural networks take a description of the Go board as an input and process it through a number of different network layers containing millions of neuron-like connections.
One neural network, the “policy network”, selects the next move to play. The other neural network, the “value network”, predicts the winner of the game. We introduced AlphaGo to numerous amateur games to help it develop an understanding of reasonable human play. Then we had it play against different versions of itself thousands of times, each time learning from its mistakes.
I wish at some point, you let a lead engineer talk about your system, at a high level of course, It’ll be super interesting!.
The hard part is they aren’t “media trained” and might give away secrets. Maybe we could do something one day pre-recorded and then edit anything out we don’t want. That could be cool.
But even being aware of some of the systems that we use would give too much away.
As people here probably know, ML is a combination of data + technique. The actual “ML” part of it (mostly) uses off-the-shelf products. There are some companies doing their own crazy custom ML stuff (google, tesla, etc) but that isn’t us.
The real trick is what you feed into it and how you chain it together.
3 Likes
While I’d love to better understand how your system works I don’t see how you could explaining it without giving away too. Unless you’re so generic that it doesn’t really mean anything from a technical standpoint.
I feel like feeding in the raw ride data in would be too noisy to train the model well so the interesting part is how you process the ride first as that would have the greatest impact in the model training. A 19x19 matrix where each position can only have three values is much simpler then a ride that has 14,400 data points per hour where each data point is around a byte in size and has to scale to all users. (Well you could be paying for lots of time on NDm A100 v4 machines…)
1 Like
Another approach is to have a real model, say like WKO5, Xert or any other derivative of CP, W’……Without a model you end up with the problems of medicine where the stats of the population don’t apply to the individual.
I’ve tried that but the actually act of counting seems to focus my mind on it and I start to try and breath “better” ie slower and deeper. And then I forget, or something distracts me and I start to breathe normally again and then I start counting and the cycle begins again. Not conducive to good data.
Someone (LOL) Wrote: "I did this exact experiment a couple of years ago with the same kind of results. After 6-7 weeks of this training, I was breaking Strava PRs left and right and riding stronger than ever. I continued this type of training for 13 weeks, peaking at 13 hours per week, and unfortunately didn’t see further gains.
I’m happy with the experiment but really wish I had switched gears to a sweet spot or threshold block after the 7 weeks.
This kind of training definitely increased my durability and overall endurance"
Quoting the larger amount of text as we have a few different topics now in this thread (all enjoyable).
This observation is interesting. As a generic interpretation for discussion, 7 weeks productive, next five perhaps less so. Can’t say I did the same experiment, but by chance, after a few months, I decided to add some VO2max along with some harder “race-like” days. In effect, I went from the LT1 focus to more of a twice a week hard focus.
It worked pretty well. By end of summer I was hitting historically solid 20 min power numbers and had a very solid MLSS / FTP test that went for about 50 minutes at low mental stress (RPE) for the wattage.
So…
Since that time have been wondering on how to be more purposeful and how to block periodize the work. I’m in Mid-Atlantic, we have winter. It might be great to do the LT1 stuff on trainer, but too boring. Am looking at something like this:
January + February: SST based indoor training
March + April: LT1 focused aerobic build
May: VO2 block (Empirical Cycling style)
June + July + August not sure what I’d do next
Probably into the twice a week hard approach and more riding than training. Meaning Tuesday Night Mountain Bike throw down, and either a really stout group ride on Saturday or a race.
Curious what you guys think. Should comment that my riding now-a-days is a lot more about being fast enough to have fun when the weather is good as opposed to peaking for an A-Event.
Anthony (AJS914) wrote that, not Brian!
That was my first LT1 experiment. All that training was actually 5bpm under LT1 because I chose 120bpm back then.
The other thing I noticed was that when I first started, my 120bpm speed was around 12mph. Painfully slow because I was used to cruising around at maybe 17-18.
That 120bpm speed eventually climbed to 17-18-19mph.
I’d think about doing VO2 in March (maybe just two weeks is plenty) after your Jan/Feb extensive TTE work. Then do your LT1 work with some weekly maintenance intensity thrown in.
1 Like
Simply structure your training in such a way that you enjoy it the most. It won’t make a difference (for your goals) if you start SST and progress with LT1 or the other way round. Simply do some/the work. Consider progression.
2 Likes
Apologies - I confused the attributes!
Yes, I had considered the VO2max block earlier.
Sryke - Yes, agree.
It’s become more about attention span than optimization. But I like to have some purpose. As productive as longer LT1 is, during Jan-Feb its all trainer and I’m more interested nowadays less in optimization and more about 45-75 min of useful work and getting off the trainer. My days of 2-3 hour trainer rides might be in the rear view mirror.
FWIW… this season I’m doing LT1 and weights Jan + Feb. When weather starts getting good in March and April will start turning pedals harder.
As long as it can easily handle the r-r timings from a fit file too as garmin head units can record r-r timings. Would be useful to get this same info for outdoor rides. May as well try and use as much of the same code path for both data sources
2 Likes