OP did give it data.
4 Likes
starlight 2 seems decent enough. I might prefer a 2 hour version of it being the weekend and having a bit more time. Or I could do the starlight then join an easy group ride for another hour or so.
True, I guess I’m too much of a skeptic to take those statements at face value…
1 Like
Yeah, I am making the assumption there is some good data in their rides. They have said they are doing decent volume for a few months now so it should be enough to get a reasonable idea of their fitness.
Hey there,
First off, sorry to hear about the experience – improving how everything starts off for users with more training history is something we are actively working on.
I don’t think, however, that your assertion that TR has “a lot of your data” to work with is quite fair:
To me, it looks like you don’t actually have that much on TR at the moment… Which would likely explain why you’re getting served low-level/low-volume workouts.
That said, though, your approach to boost your PLs up is what we’d typically recommend. More experienced athletes normally find that sessions at Workout Levels of 3.X-4.X are good departure points when they are starting out with TR. Workout Alternates is a great tool to use in this kind of situation.
Once you successfully complete a workout at a given WL, your PLs will update accordingly and Adaptive Training will slot in workouts at your updated levels. Any workouts in that updated zone already on your TR Calendar for the future will be subbed out for more appropriately challenging ones.
As a bit of a coda, I’d also add that while you do say you have a strong training history (and I believe you!), we’d recommend approaching your training and plan with a bit of caution. We’ve seen a lot of athletes champing at the bit to really get after it – and, well, the bit often winds up champing them.
If you haven’t done much training in the past year, we’d recommend increasing your volume/intensity gradually and conservatively. I promise you won’t be “wasting” any training time, and your body will probably thank you for it.
I say this because (at least from what I can see on your TR Calendar), you’re now on your 9th consecutive week of training without any obvious recovery weeks, and you just started a new plan (which won’t have a recovery week until 3 weeks from now). Remember that rest and recovery are just as important as the actual training is!
Hope this helps clear some stuff up – and I’ll pass your feedback along to the rest of the TR team. Feel free to let us know if you have any additional questions!
34 Likes
I just noticed that training history is actually missing a few weeks of workouts. Something didn’t sync right. My first ride back was on Nov 5 where I did a 22 min ride. The next day I did a 40 min ride. See I started slowly 
so sounds like the system gave you appropriate workouts then. Glad that got cleared up. 
18 Likes
When you look at the yearly graph it does look weird doesn’t it. From 0 to 100 in like a month. But I’ve been going as my body responds to the training.
How long of a history would the ai need to start right? Im definitely an outlier because the starting point makes it look like it thinks I just got off the couch jumped on the bike and don’t know anything.
My aerobic system kicked in pretty quick once I started cycling but the ai should be able to see that my hr is in the right zones but I don’t think it takes hr into account.
If I was here saying a PL2.0 easy ride was easy but my hr was 160-170 or higher than ya you can say I’m over reaching on my easy rides but I’m here sitting at a 100bpm avg. any coach should be able to see that and clearly see that level of effort is not enough. My body tells me that.
I understand the rest stuff and if I was telling my wife to start cycling I would tell her the same thing. She doesn’t ride so starting her out slow and learning what power zones are how she responds to different efforts how it feels to ride at ftp what sweet spot is etc she’ll need to learn. I’m like let’s skip all the intro stuff and get to the meat and potatoes. And that’s where the disconnect from the marketing and actual plan the ai builds is.
“## Your Personal Coach
Just Sync Your Data, We’ll Handle the Rest
We analyze your training history and build an adaptive plan tailored to your fitness, goals, and training approach.“
In reading the front page I don’t see it highlighted anywhere that it needs 10 rides to detect my ftp. Even after signing up it just says ai detect will handle it. Which it did. I’ve done enough 30+ min hard efforts to get a good picture of what my ftp is. Then I jump in and I’m on ground 0. At least I know enough to see the workout and say ya that’s too easy.
Some people definitely prefer to start slow as I saw a post today of someone saying a PL 3.2 workout was too hard of a start for them on a polarized plan. I see that workout and immediately want to change it seeing it as too easy. Seems like TR needs a different system than just plan type of setting it to balanced or aggressive or those in between.
1 Like
because that is what you told the system……this isn’t a TR issue, it is a “data in” issue, compounded by you being an outlier.
I would just pick a few workouts that seem reasonable and start loading the system with good data and it will adjust quickly.
8 Likes
For sure – I can totally understand that and I’m glad to hear you’re able to leverage your previous training experience to your advantage.
Here is a sneak peek of what we’re working on for situations like this straight from our CEO, Nate Pearson, if you’re interested. Basically, we’re working on a big update that will help get athletes “locked in” to the right workouts even faster than AT is currently capable of doing.
Since your Threshold workout for tomorrow seems a little low based on how you’ve described your plan, I used our RPE Prediction tool (which is incorporated into how TrainNow functions after recent updates) to find a replacement – I went ahead and replaced your original workout with “Warlow,” a Threshold 4.7. It may be “hard,” but it should still be doable. 
Once you complete that workout, your PLs will update and keep you on track.
12 Likes
Yup read through the whole link. What I don’t see is how heart rate may or may not be incorporated into a workout selection. I don’t see anywhere to enter my resting and max heart rate either. You may not be able to talk further about what’s coming but I hope that’s integrated.
I definitely like that warlow workout much better. Lots of peaks make me want to go after that workout.
I appreciate everyone’s insight.
1 Like
Heart rate is such a risky thing to include, because it is impacted so heavily by variables that TR cannot be aware off. The data is pretty poor generally (huge variety in HR monitor quality), plus temperature, hydration, fatigue (TR knows a little about this), sleep, etc. Overall I think not considering HR at all for adaptive training likely results in a better quality workout provision overall for TR users than if it did use it.
The simplest approach would probably be a data upload plus survey that seeds you to a certain set of PLs right at the start, then adjusts down as well as up. I’d have been best off around 3-4 PL rather than starting at 1, so that was also a frustration for me. That said, the new ML model sounds like it’ll improve things.
I’m not really sure what the point of PLs is for endurance and tempo TBH. The intensity of those sessions is better determined by your level of fatigue.
3 Likes
Just completed the 20 min ftp test. Result was 300avg and ftp 285w. I prob lost the 3 watts in the 5 min effort before the 20 min. So the ai detection was spot on for ftp.
One thing I noticed in this integration is that erg mode does NOT turn off in Zwift for any workout segments nor the 20 min effort. Have to turn it off manually.
1 Like
Curious what your weight is (w/kg FTP) as your FTP is impressive as is for a new cyclist, but not super surprising given your athletic background.
83kg so 3.47 w/kg
1 Like
Very solid. Sounds like you’ll be picking it up pretty quick with a lot of potential. Don’t overlook the cutback weeks. It’s easy to feel like they aren’t as necessary vs running since it’s low impact, but still very much needed.
6 Likes
Disagree. I think TR has already said they use HR data, but not to what extent. Sure there’s variance, just as there is with power data. Also, even if my HR data isn’t accurate, it’s consistency I care about anyway. Those other externals are factors, but “it’s hard” seems like a bad reason to not do something.
Think I read somewhere that FTP tests should be completed using the Zwift version for now
It’s not consistent. Power is a measure of output, 300w is always 300w. Heartrate is a response. The reasons for any heartrate value could be extremely varied.
Don’t get me wrong - it is useful to track as an individual or as a coach, and you build it in to the overall picture to help inform decisions… and therefore logically it could one day be considered by AI/ML, but we’re a long way away from that.
Would be good to get some insight from TR as to how if at all they use it, but I am very confident that adaptive training doesn’t consider it for any decision making.
1 Like
Just think about it for a moment: TrainerRoad would need evidence that you have consistently worked out over the years and performed well. If you look at the workout history, do you see that reflected?
Your current numbers are good, but not exceptional in any way. Also here there is no sign (yet) for the algorithm that you have tons of experience.
How should TR know if you don’t tell it? IMHO you should still take it more slowly than you want and start with a bog-standard plan first, slowly ramping up intensity and volume. That’s the advice to cyclists that get into running: start ridiculously easy. In the converse direction you don’t need to be quite as cautious (for reasons mentioned already), but still.
Heart rate is a secondary metric in cycling (unlike running). TR’s algorithms do include heart rate to my knowledge, but in cycling the primary metric is power. Heart rate is a metric that needs context, much of the context is not contained in workout data.
6 Likes