Ampel means traffic light in German.
1 Like
TrainerRoad OverLoad Limiter?
Keeps you from getting TROLLed?
@Nate_Pearson can you take a Look at my history for reference? I just started commuting to work again (30 min each direction) and I’m curious how it affects my training. Should be the perfect example of junk miles: too short, too much stop and go and then usually a little too much power output to keep going with the flow.
Also, I’m interested how RLGL would respond to the endurance rides on Sunday. They are planned indoors 2h, but usually I take it outdoors for 3-4h on hilly terrain with some spikes in power output and a good amount of coasting. No proper Z2-work…
I’d call this tool the train brain.
Check Engine Light
2 Likes
1 Like
Yes, I want to be able to turn it up! I like that the TR AI program suggests workouts, but sometimes it feel like baby steps even though I’m not only completing, but extending the workouts as well as adding an extra workout to extend the ride.
1 Like
I’d be curious what it says about mine @Nate_Pearson. I’ve been away from structured training for quite a while, but started up again in November.
Tonight I failed my first workout, Owl +2. Intervals.icu and Elevate both show my “form” as being firmly in the middle of the “optimal training zone”.
I’m on a medium volume plan but have certainly added in additional outside rides, because I felt easily able to at the time
Tonight’s ride should have been threshold, but felt more like vo2 
The classic TSB method of judging over training is just looking at TSS and time, so is going to underestimate your training load if your FTP is set too high.
I wonder how accurate RLGL will be for Paris that have their FTP set too high. For some people ramp test and AIFTP both set their FTP too high.
I know Nate said it’s using more variables than TSS. It will be interesting to see if RLGL can somehow detect overtraining through the other factors (presumably heart rate/power, post-workout surveys, etc), but I assume TSS is still the primary factor, which could lead to not warning the athlete soon enough.
Adaptive Training doesn’t continuously push PLs until failure.
It aims to “make the hard days hard and the easy days easy” in simple terms.
In more complex terms it’s the workout for your current goals/plans that has the greatest statistical chance of success (success is defined differently for different workouts/parts of plans).
5 Likes
Yes, we’re going to build it into plan builder so we can do dynamic volume like I showed on that CEO update video.
Basically show you what volumes you can handle and let the athlete specify how “aggressive” they want to be with their training.
Maybe something like:
- Maintain
- Progress
- Stretch
There’s another level to this that I’ve never talked about. It’s a brand new AI system that will allow us to run Monte Carlo simulations on different training plans/workout combos.
We can then find the most likely combination of volume, workout spacing, and workout choice that leaves you with the highest combination of FTP and PLs that take into account your cycling discipline and goals.
For those unfamiliar of Monte Carlo simulations I’ll give a brief/simple overview for our use case.
It would simulate execution of your training plan over and over (maybe 50,000 times?) and come up with a distribution of fitness outcomes for training plans for you.
Much like how they use it in stock market investing where you get a “most likely” outcome but also a spread of possible outcomes.
We can simulate your training and run AI FTP Detection on the plan and chain it out month over month.
So end result from plan builder would be:
- appropriate volume based on how “aggressive” you want your training to be.
- optimal workout spacing during the week (it’s gets tricky for short workers, fire fighters, er docs, etc.)
- predicted spread of FTP
The really cool part about this is it would predict skipped workouts, failed workouts, and outside rides.
Just don’t tell anyone. 
18 Likes
Well recommend rest in TrainNow on a red day.
RLGL should go into early access this week!!! It won’t have TrainNow support this week but it will before full launch.
17 Likes
There were some in early season base phase but not many.
They basically stretched themselves early during the season and then rode 
that fitness the rest of the year.
Makes sense too, better to push it during preseason when you can control more rather that right before big races.
6 Likes
Potentially not in this case. We have an update planned to recommend an entire recovery week for athletes like yourself who train consistently and never reduce their volume.
We know that’s a limitation of the current system but we thought it would be better to get this out and add a ton of value and then add that later.
3 Likes
No it is not. 
3 Likes
5 Likes
Time to lube up the F5 key.
3 Likes
Simulating an individual athlete with specific characteristics (age, sex, history, goals) and then using distributions from the larger population of athletes to predict the range of outcomes? Will you be able to account for correlations between those distributions or assume they’re independent (will have an effect on the width of the confidence intervals)?
I assume, once this is live, anybody can toggle this via Account → Early Access?
4 Likes
Correct
5 Likes
Good to hear of some progress. Could be ideally timed to make an appearance in the next podcast episode?!
1 Like