I read one of the blogs from Marco ( IIRC?) that talked about that point and using it as a guide for ‘readiness’ to train.
Spend a few mins during the warmup looking at the dfa1 numbers and if you get to 0.75 at a relatively ‘low’ power then avoid a hard day and rest or go very easy.
I rather like that approach as I find RPE can be very misleading in that regard. I’ve dragged myself onto the turbo many times when thinking I wont even survive the warmup based on how my legs feel, only to deliver excellent numbers. If I’d gone on ‘feeling’ and ‘rpe’ walking to the garage I would have simply bailed unnecessarily.
I agree that this is useful, but don’t think you need to look for a specific number as it is more a drop from what your number is when recovered. Also probably less of a drop and more at wattage x it will be y alpha 1. Do an endurance workout you may not get a low alpha 1 number but if its lower than what you normally would get maybe you should do more rest. Same goes for higher intensity intervals.
So for all your progression levels in TR you should have an expected alpha 1 reading for the work part. If the current reading is below that you may want to bail and rest. You should also be able to fail early by just seeing how you do during the warmup part. This could easily be added to the red/green functionality @Nate_Pearson said is coming as a way to let the user bail early in the workout. Plus if the number is low but not low enough to bail could influence how much rest or lower work load adaptive training should give you. (or increase your ramp rate if the alpha 1 number is higher than expected, but maybe wait till the end of the workout for that measurement)
This would seem to be something that AT and trainerroad could look at - they have control over what the warmups are like, and there likely aren’t that many variations of the warmup. (and if so, that could be changed.) This is likely already on @Nate_Pearson’s radar.
Quite honestly data collection for this work could be added without users knowing about it - the HR monitor is already paired, just more complete data needs to be recorded. Hopefully this happens sooner rather than later - this should also enable using the TR recordings to do offline HRV analysis, which would be another benefit.
Personally I think using dfa1 during warmup may not be that useful, and I’ll explain why. I’ve been using Garmin/FirstBeat Performance Condition for about 2.5 years, here is the description on FirstBeat site:
in my experience the Performance Condition metric is primarily useful as a trending tool, versus during a ride it doesn’t add value. During a ride it typically ranges between +3 and -3. Even when it starts dropping, and falls to -3 during a ride, I’m able to complete my intervals. So I don’t show it on my computer, and only look after a ride. On the other hand it has confirmed when it wasn’t my day, but that’s only happened once when I started riding and 10 minutes into the warmup felt absolutely terrible, didn’t need a computer to tell me it was time to turn around and go home, and I looked and Performance Condition was -7 or -10 or some really high negative number.
Back to my point, Performance Condition is most valuable for me as a trending tool, if I’m trending +2 or +3 workout after workout, that usually indicates a small bump in fitness is coming.
Trying to determine “readiness to train” ahead of a workout has never made sense to me. However I’m not trying to cram 4 hard interval days into a week, with each interval day pushing hard to drive up work load and training stress in a minimum time-crunched amount of time.
Instead I’m doing a sensible amount of hard work, leaving something out on the road, and watching my fitness climb because there is adequate balance of endurance and hard work. Completing workouts is rarely an issue. If I’m deep in a workout and feel I need to back off power, or drop an interval, well that happens and I got some good work done and will take it for what it was. Better to try and push thru because its usually a mental barrier. Finish up and ride home. I’m a math major with an engineering degree, a real data nerd, but at the end of the day there is an art to training, and knowing what data is important (and when), and ignoring the rest. Sometimes we train our mind AND our bodies.
For everyone on the machine learning ‘feed it more data’ bandwagon I’ve got news for you, Garmin has some solid ML training analytics, been around for years, uses HRV and HR and Power, and if you don’t have Garmin you’ve been missing out on a pretty interesting tool. Just a little FOMO for you, instead of posting another wouldn’t it be great, add it to the TR enhancement wish list.
It just got better, here are part of my post-ride notes “Woke up needing more sleep. Didn’t feel good and took longer warmup. Higher RPE but thought ‘Coach Isaiah says sometimes we train the mind’ and then proceeded to get it done without drama.”
I was mentally ready to give up, the last thing I needed was a readiness score to support giving up before actually attempting to do the work. Just got stronger as the intervals went on, and that’s pretty common for me.
If I paid attention to morning HRV or early ride RPE and/or Garmin Performance Condition, I’d be spending a lot more time either off the bike or doing easy rides. IMHO that’s a formula to reduce performance.
This. On quite a few occasions I have sat reading my HRV at 5am and then got it into my mind I’m going to have a bad workout. I then proceed to get on the bike and smash it out without any problems. All about long term trends with this type of data for me.
I nearly always find the performance condition motivational. It always been positive every session this year (with the exception of two sessions when like you said it was clear there was a problem, ill and 3 hours sleep the night before) the number of time Ive felt like shit and you hear a beep look and its +5 makes me think… see feeling crap is in the head today.
I’m just hoping the news from a bunch of people saying alpha 1=.75 doesn’t equal LT1 (even though those studies don’t really conclusively show that) doesn’t mean recording this data gets dropped as even if .75 doesn’t mean anything there is still use for the data. Also the discovery of the garmin app giving proper alpha 1 numbers using Ant+ should make it even better for TR as they should be able to easily do the same with their ant+ code. Not depending on BLE heart rate strap pairing makes the population of users that this could be useful for much larger and easier when you don’t have to tell them how to pair.
Adding this functionality without users needing to know about it is why this could be very useful. It may be complex on the backend but no extra work for users.
Garmin and polar straps, thats probably a big number
Who cares? The TR app does the recording for indoor rides. For outdoor rides just post instructions. This isn’t just about impacting the people who’s HRV is being recorded as there may be other trends in data that this could make more clear and help train the model better to help even those without this data.
They may be trying to give the user the same data but that doesn’t mean they are based on the same calculations. The details here are important. Garmin doesn’t have a large data set of HRV info. By default no garmin even saves HRV data and if you do save it (hidden deep in the menu) and are paired with an ANT+ strap the data is poor (see discussions above on how the new garmin app gives good data with ant+) So if it records bad HRV data for alpha 1 calculation purposes (its very sensitive to small drops in data) why would you assume it can do those calculations in a useful way? Sure, it may still do some HRV measurements, but ones less sensitive to bad data.
Another thing to look at is that performance condition is from back in 2013. Alpha 1 is pretty cpu intensive for very low power devices like a watch from almost a decade ago so very good chance that’s not what its based on.
Also while there may be machine learning to create the algorithms that run on the different garmin devices, the devices are stand alone and don’t get their data processed by the backend. So their machine learning isn’t learned off of your data and all calculations on your data are relatively simple
blah blah blah sounds like a bunch of red herring arguments.
Garmin’s HRV/HR/power machine learning algorithms are working well for me, have been for about 2.5 years. Firstbeat has published papers with high-level algorithm details. I don’t care how much TR points to its mountain of data, coulda shoulda woulda, why does TR devalue aerobic endurance in its training plans? After all, we are in the LT1 w/o blood testing thread. And TR isn’t collecting HRV data yet, is still working on higher priorities, and while TR has stated that adding HR data is important for the future it is not here and none of us have any idea of when it might happen. A year ago they were going to support outside workouts, and a year later it morphed into waiting for Levels 2.0. I’m sure that will be a nice improvement, but bringing new software to market isn’t easy.
I’m more interested in living in the real world and using what’s available now. And I can tell you the Garmin stuff works, for me, no matter your attempts to cast FUD.
It often does but it’s complicated and contextual, like FB relationship status. Not sure it’s worth elaborating and the conversation has already moved on. LOL.
I can tell you this much, all this bitching about RPE is validating and vindicating for me. I struggle with it. I seek a metric or set of metrics, but I fear there is none.
In my previous life as a college runner, I always thought it was an asset to be able to lie to yourself. “Backstretch, you feel great bud…let’s go” <— that’s always a lie.
All the garmin straps. It doesn’t have to be BLE only if paired to TR’s app (which has been gone over a bunch of times above) That is already a large enough population size. As I already stated, sure the data may be the most helpful for users who have this data but as TR is based on building ML models, the extra correlations shown by this data in a large subset of the users will help lots of users
The TR app records straight off the strap. Why does the setup of a garmin head unit matter.
Garmin doesn’t have the data from their large pool of users as this data doesn’t exist. But somehow that is a red herring?
When did I say Garmin’s stuff didn’t work? I did just buy a Fenix 7 when it came out. All I’m saying is we shouldn’t be limited to whats available now as we should try to improve on it. You’re the one who said that Performance condition from Garmin doesn’t add value so say because that you don’t think dfa1 would be useful. You’re the one claiming garmin stuff doesn’t work. Performance condition is not alpha 1 so what applies to one doesn’t have to apply to the other. All I’m saying is TR should look into this data
These aren’t such easy metrics to use. HR changes do to lots of other things that have little to do with stress like dehydration. The advantage alpha 1 gives is that it appears to correlate much better with physiological stress.
a garmin is a very low power cpu. TR runs on cell phones and full computers. TR can do more very easily and lots of people have the needed hardware
