I never look at my HR in a race.
3 Likes
HR drift is always good to observe and HR is lower or higher than your expect it can be a sign of overtraining or you have an illness coming on.
6 Likes
No. You were told it wasn’t part of the AI FTP algorithm. But you as a human could always gain some value from it, such as HR drift decreasing showing improved endurance or pacing yourself better in the heat or… or… or…
2 Likes
Ok so I’ve had a pretty big bump in FTP with the beta which has been a very short ego boost, lol.
The next few workouts were a bit of a TrainSmash, though, because they stayed at the same progression levels (around 4–5), so the jump in FTP made them feel properly spicy. I ended up backing them off about 4% just to get them done.
After completing the post-workout surveys, I can now see the next round of Build workouts after this recovery week have adjusted down to level 2–3, which feels much more manageable, so it looks like the AI is doing its thing.
Really enjoying the beta and it’s cool seeing the new features roll in. Loving it team TR.
4 Likes
Yeah, that’s why none of the pros are wearing heart rate monitors 
3 Likes
I wanted to share some data around the new AI FTP and how people’s FTP will change overall.
There’s a sentiment on the beta group that the AI will give people an inflated FTP compared to the old way. This is based on anecdotal reports in this thread.
Here’s a histogram of non-beta athletes’ levels who have done a TR ride within the last 30 days.
The really big jumps in FTP will be for the people at level 7 and above, which is 1.4% of athletes.
We’re trying to normalize people around a level 3 threshold, so you can think of it like this:
< 3: FTP Decrease
Around 3: About the same
4-7: FTP Increase
7+: Large FTP Increase
So what we’re seeing in the thread is survivorship bias. The “survivors” in this situation are the ones with dramatic FTP changes, are they are more likely to post. The ones where it’s about the same, or decreases are less likely to post (based on what we’ve observed in this thread).
I just want to mention this because narratives can run out of control when a few recurring posts don’t represent reality.
That being said, we’re still looking at how people’s workouts go after an FTP change to validate it. For the threshold, the failure rate due to intensity or training fatigue is 2.4%, which is in line with our acceptable predicted failure rate.
30 Likes
This makes a lot of sense. Thanks for the clarification. It’s definitely confusing looking in from the outside as someone who hasn’t been added to the beta yet but browses the forum regularly. Looking forward to trying it out!
2 Likes
Hey Nate!
I just switched my Training Approach from demanding to balanced and well…
This does not seem correct. My FTP prediction was 363w (much more reasonable) with the demanding Training Approach. Any insight?
1 Like
TR had my FTP 313W and yesterday I rode one hour outdoor at an average of 320W, which is 2.2% higher.
Your AI detection states
In the meantime, we’ll automatically monitor every ride and workout and let you know if we detect a significant change in your fitness.
I would have thought that a real world one hour power record, that is 2.2% higher would trigger a FTP update. I’ve had similar experiences early in 2025 where I have ridden a 337W 20min climb which would equate to an FTP of 320W based on 0.95 and TR did not recognise that as a significant bump in FTP.
3 Likes
I think you’ll progress faster this way, probably because you’ll recover better.
We really need to add Monte Carlo simulations too. It’s pridcing the most likely path, but if something is 55/45 it could go another way.
This is still a good way to see how changes impact your fitness!
8 Likes
I was curious about how many simulations you were running. So if a workout is predicated 65% Hard 25% Very Hard 10% moderate, is that reflected in the next workout AI simulation, so it’s the workout that would occur if you selected the most likely RPE, then the AI simulation is the most likely improvement for each of those most likely workouts?
Did not expect this. Wouldn’t that chart hint that AI FTP prediction was too optimistic for these athletes or many of these athletes train with inflated FTP?
Good to see that you will normalize athletes towards level 3 threshold.
I think that would depend why the 1s are 1s. Is it due to workout failures or level decay.
6 Likes
I’m jumping in the forum and reading all the cool stuff you mentioned in this thread. I really like the approach of dropping the FTP but in that case, how should we configure our bike computer for outdoor rides other than trainings like sportives for exemple to help us pace our efforts ?
Thanks in advance
3 Likes
Or time constraints
I thought that only the new Beta(not yet released) was supposed to incorporate data from outside rides. I had been riding outdoors almost exclusively for about 3 months and I knew my FTP had increased. One week before I started up on my winter TR training plan I did a SS workout that wasn’t too hard. I only got a 4W bump even though the ride wasn’t hard. I new that it was going to bump it up but I didn’t want a huge bump to start my training so I didn’t pick a very hard SS workout. I guess If i would have picked a hard TH workout I probably would have had a big increase in FTP.(from what I have read in the forum) After I did some some Z2 TR workouts this past week as a recovery week before I started my TR SS workouts. I had a big adaption on PL for SS workouts(even before I started the TR plan) they went up 3.2 this was based on only doing Z2 rides( think it took in the lower HR with a slight increase in power). Now that I finished today’s ride I got another adaption and increase in PL for SS workouts. So I am thinking that part of that Beta program is slowly leaking in to the reguar TR program. I know from my workouts that my TR AIFTP is low as it didn’t account for the outside workouts. This new release is supposed to fix it. But for me the PL have incresed significantly without an increase in FTP all due to riding outdoors.
TR AIFTP and adaptive training have taken outdoor rides into account for quite some time. This is not related to the beta. Progression Levels do not take into account unstructured outdoor rides, but they take into account structured ones.
2 Likes
The new beta is supposed to take into account unstructured outdoor rides, which is what I was doing from about May. That is why I did a TR ride to make TR Increase my FTP as I knew they didn’t take it into account. I was doing the Garmin coaching outside workouts with 20min repeats and 15 minute repeats, close to one hour of total work over FTP consistently week after week with the Garmin coaching and TR AIFTP would not even have given me an FTP or PL increse if I didn’t decide to do the TR rides on my trainer. ( I know this because halfway through the training plan I just decided to use the TR AIFTP and it actually lowered my FTP). This entire time my Garmin VO2 went up by 11 in 12 weeks and the increase in watts and the percieved effort and lower HR made it clear that my FTP should have been increased significantly. This has been an on going issue as I usually only ride the trainer during the winter. The wierdest thing is that before I switched over to the Garmin training I was actually doing the TR workout outdoors, Just not sending it to my Garmin Unit. I would memorize the workout and go do it outdoors. Sure it wasn’t perfect as it may have been a 2:30 TH ride when the TR called for a 2hr TH ride. But I have climbs nearby where I can do the set intervals at the set wattage that the workout perscribed. The overage of time was always a longer warm up and warm down and of course slowing down and stopping for traffic lights. Yet it was the same workout that TR had on the calendar. Yeah I actually didn’t want to send the actual workout to my garmin as I am not riding on a road without any interruptions.