Agree on this point. Seems like there should be a progression of yellow before green again.
1 Like
RLGL does not show a graph. So, we don’t really know how far into the red we are when we get a red. It does not follow that every red day should be followed by a yellow, regardless of recovery.
It’s surprising and disappointing that neither @Nate_Pearson, who just did a long RL/GL introduction on the Ask a Cycling Coach podcast, nor anyone else from the RL/GL product team has weighed in on the many valid questions and mixed experiences discussed in this thread. We’re trying to figure out whether and how to get the best value out of this new capability, but we have many unanswered questions and unaddressed doubts about RL/GL recommendations. What say you Nate and TR team?
A lot of times when I’ve done long outdoor rides there’s been a lot of unintentional intensity in them. I used the slider in my account to bias towards less aggressive. Personally I’ve found the that one step towards more conservative fits very well for me in my currently somewhat detrained state. In the podcast they have said that yellow days are not necessarily required to be endurance or recovery days, just that if you consistently always train hard on yellow days you are more likely to have adverse effects in the future. They used Hannah as an example stating that in her career she has had multiple yellow days in a row on which she trained but often followed by a recovery period and its worked well for her.
If you have a lot of experience training I feel this feature is just an additional data point, as HR or HRV would be. Its usefulness will obviously vary greatly from person to person.
oops! i’ll fix that, thank you!
1 Like
So far seems pretty good, there isn’t actually a green light though right? As far as yellow light, my adaptive plan gave me a workout to do and I chose a harder alternative Z2 ride than suggested, 1 hour instead of 30mins so then it gave me a yellow day again, took that off and then it suggested I was ready for intensity again the day after or well it didn’t give me another yellow… I guess what I would like to see is further clarification for what I can and should not do on a yellow day.
I am not taking much notice of it to be honest, it feels a little gimmicky. I guess if it had a couple of red lights in a row it might give me food for though but I think I know my body well enough to back off if required.
1 Like
I agree, when all the information and biometrics are all joined up it will be a formidable beast.
Had an easy day yesteday, 60 min Z1/2, 2 hours endurance today and it’s given me Amber for tomorrow yet Whoop shows a 98% recovery and says go go go!
Love all this data!
I get the same experience between TR (RL/GL), Intervals ICU, and Strava F&F. Someone commented on YouTube that AI only inputs your data if it is an indoor or outdoor structured workout, and not a generally unassociated ride.
If this is true, we’re going to get very different results across Intervals ICU and TR.
In my case, I can’t always trust my perceived exertion. I am on a cholesterol medication that has the side effect of creating soreness in large muscle groups. Once I warm up, it’s all good - - unless I really am on the edge.
I believe that is incorrect and AIFTP and RLGL take all rides into account.
6 Likes
FWIW, a few more weeks into this, and RL/GL is working exactly as it should. It’s a great addition to my training. The few times I’ve overridden the yellow or red warning, my legs let me know I was on thin ice within the first 20 minutes or so of the ride. Bona fide rest days are hard, but I love hitting the next workout at 100%.
5 Likes
I’m dealing with a slightly different issue. Following the LV 70.3 tri plan and my last 3 long runs have been yellowed and reduced to 20 minute easy jogs by the system. These occurred the day after planned rides where my most difficult rating was moderate. TR support responded quickly saying that as my fitness improved, the system would abort these measures. Here’s my question–If I’m nailing the workouts prescribed by Planbuilder with relative ease, why would the system change anything. And if the changes are indeed the correct response based off the previous day’s ride, then why are these types of activities coupled together so early in the plan. Seems like a better plan would be for planbuilder to space these out a bit more.
Adding some feedback for the TR team: Since March 1st, 30 calendar days have flagged yellow for me (and 3 red). I am not going to dispute the Reds; they were spot on in suggesting that I knock it off. But the yellows, at least for me, seem overly aggressive. Over 50% of the last two months populated as yellow, and I have simply ignored it nearly every time and have been recovering well. Maybe 3-4 of those yellows would have been the right move to follow (giving an honest appraisal of the last two months in hindsight). One of those was yesterday, but I was so desensitized after ignoring so many that I was too overconfident to accept the adaptations.
I am following a high-volume plan and have the capacity for high volume. My adherence to the plan has been over 90% (100% achievement on cycling workouts), with pre-ride surveys being neutral to positive and post-ride surveys maxing out at very hard (for high VO2 max intervals). After two months, the algorithm doesn’t seem to consider my acclimation to the volume or adherence to the plan, which I believe is oversaturating the yellow prompts. I also am open to (and leaning toward) the idea that the running and swimming components of TR don’t play as well with the algorithm that is obviously cycling forward.
However, I struggle with this a bit, as while I appreciate the intention to prevent overtraining or non-functional overreaching, if I had listened to the recommendation to trade down the yellow workouts to the recommended endurance/recovery alternates, I would not be making progress in my plan. To be clear, I realize that there wouldn’t be as many yellows if I didn’t ignore them all, but calling down key sessions in a very general sense is something I am seeing too often. For example, I am currently in the last week of a training block with a de-load week coming next week. This week should be a push; by design, I should be overreaching over the coming days to facilitate physical adaptation, with an entire 7 days to shed fatigue and reset next week. Prompting me daily to back down my plan is not necessarily the most constructive. Of course, yellows can go either way, and the intent is likely to have the athlete sense check against how they feel, but I think, for some, the permission slip to not do the hard thing may be too great to pass up. On the flip side, the type 1 error rate frequency in the yellows may cause desensitization, and athletes may be quick to skip the suggestion when it might be the right approach on the day.
The mild inconvenience of having to constantly skip the adaptations (and potentially seeing the algorithm miss other adaptations to alternates at varying progression levels) is not that big of a deal, but I think the model needs some tweaking. I think the recommendations in this post are great, but most of those items certainly fall into the future/long-term opportunities list. I have to wonder if, in the short term, the logic can be modified to more heavily weigh the last 4 weeks of load/adaptation and training plan adherence to provide a better balance for those yellow triggers.
I’m thinking the following from the OP could be considered for short-term adjustments: 8-1 - training blocks/camps, 8-2 - history of successfully overreaching (instead of overtraining), 9 - benefit of intensity in a fatigued state could be somewhat considered in the strength of the logic against key sessions, 10 - future training/calendar activities could also be integrated, with the caveat that the athlete needs to be diligent with adjusting their calendar at least one week out, and an emphatic YES PLEASE on number 11.-suggesting lower progression levels rather than endurance. I had a yellow for a VO2 max session last Tuesday, and I had to reach for every last pedalstroke to get through, which definitely came from fatigue. Doing those same intervals at 115-120% of FTP instead of 120%-125% of FTP would have actually been the perfect adjustment. I was too stubborn to adjust the bias once I was in the workout, but if the alternate had been served up initially, I probably would have (and should have) taken it. I would still have gleaned the primary benefit of the workout without some of the sloppy form. I will absolutely raise my hand to say being too stubborn to adjust the bias is short-sighted and, frankly, dumb, but when endorphins and adrenaline are in play, dumb decisions are frequently made.
In any event, keep serving up those red lights that I affectionately call “hey, idiots,” and I am looking forward to how the treatment of yellow days progresses as more users provide feedback.
2 Likes
If you move your RLGL slider 1 notch to the right it should be more appropriate for you. RLGL errors on the side of caution for the general public. You, like me, can handle a bit more so need to tweak the algorithm.
2 Likes
Definitely use the slider and set to aggressive. That will likely drop some of the yellows. Listen to to your body. If you can handle the stress and recover, go for it. I know for me though I tend to do too much and don’t give myself enough time for recovery. You can also just turn off RLGL.
1 Like
I did have the Training Approach slider set to aggressive, but I had to manually install an update recently, so it may have reverted to balanced. I’ll keep monitoring to see if that improves things a bit, and I’ll have to keep an eye on that slider to make sure it doesn’t revert - I also have TR on multiple devices, which could have caused an override of settings. I am not quite ready to turn off RLGL as I did want to see how it evolved, not minding the extra workarounds in the interim. In the 8 years I have had TR, I have yet to see a feature they haven’t hammered into a good, refined, and usable product.
With all that said, my definition of “aggressive” and someone else’s would be relative to training history. It almost seems as if the “balanced” approach uses too broad a brush. I think that slider is a great feature, but it may need to bump against rider history, which it may not be considering heavily (if at all) through the Adaptive Training model. I’m sure more refinements are to come, at least some of the noise should be cleared up for now.
1 Like
I’ve had my slider set all the way to the right for about a month. So far so good. I still get a yellow after a 300 TSS day, which seems about right.
1 Like
The way I’ve come to think about RLGL is that it is analogous to a screening test for a future medical condition (in this case, probability of future long-term fatigue/burnout), not a current condition (probability of current/short-term burnout/failure to complete workout). Analogies might be bone mineral density (current) for risk of hip fracture (future), or LDL cholesterol (current) for heart attacks (future).
For any screening/diagnostic test or algorithm, you can make an ROC (receiver operating characteristic) curve, which shows how good the test is at predicting the condition if you have the condition (sensitivity) and how bad the test at falsely telling you you have the condition when you don’t (1 - specificity). For any test, these are inversely correlated–as you increase your ability to detect true disease, you also increase false positives, and vice versa. The concept comes from WWII and the early days of radar, where you had to optimize the trade-off between sensitivity (launching fighters in response to bombers coming over the Channel) vs specificity (avoiding launching in response to a large flight of geese crossing the Channel).
There are two additional sources of uncertainty here–the inherent uncertainty of predicting the future (which increases the farther out you’re trying to predict) and the quantitative uncertainty resulting from any model based on sampling, so the ROC curve has confidence intervals
We don’t know what the RLGL curve looks like, but presumably it’s somewhat similar to this qualitatively. My bet is that the “balanced” point in the slider is the upper left hand corner of the curve; moving it towards “aggressive” moves it down and to the left (lower false positive, but also lower sensitivity) so if you do get a yellow or red light, you should definitely pay closer attention to it than on the “balanced” setting.
3 Likes
This is good to know, thanks! I believe that my personal use case throws some challenges at RLGL.
- 62 years old male; 205 lbs./93 kg; 6 ft. / 183 cm.
- My world: I live on the edge of the Blue Ridge / Appalachia. Every outdoor ride is 100’ per mile.
- Goals: Modest. I like to ride fast(er) but I don’t enter races.
As a somewhat new TR user, I am coming to the conclusion that I’ll get enough (intense) exertion in my unstructured outdoor rides, and the main bonus with TR is going to be in my Z2 rides. It’s pretty much impossible (for me) to hold Z2 in the mountains, but I do perceive the muscle endurance improvements.
@Sarah Can you share a screen shot to give a visual of what you are experiencing?