In my experience, ftp detection is pretty wonky. I’m riding it out since I’m paying for this product, but my faith is somewhat stagnant on actually increasing my ftp. My next detection is two weeks out and the prediction is close to what Garmin is suggesting and what I feel I can do. Will it stay there is the question. Overall I feel good, strong, durability and TTE are solid (mostly because I add about an extra 8 hours into the plan TR gives me). So there is that!
The prediction tends to be accurate in terms of the AI FTP at the end of the block.
In other words, predicted starts at 300W, gradually drops to 250W over the course of the block. At the end of the block AI FTP is 250W. So turning off prediction might remove the daily frustration. But the frustration of 4weeks of following a build program with zero gain would persist.
Of course you can just do a FTP test after the block. You could also just go back to TR 2020 with a fixed 4 week plan, no AI and ramp test ever 4 weeks.
Same here!. The it dropped 4W after doing a prescribed endurance workout rated easy. Go figure.
Yes, it’s been stated several times that it’s with Prediction. Outdoor workouts in your plan inflate the Prediction. I have not set any Outdoor Workouts since the bug was announced, I only have Group Rides and Solo rides in my Custom Plan so I can assume I am not affected.
Whether it is a big deal or not is in the eye of the beholder - it’s certainly frustrating a lot of forum posters.
Hey Team,
do you know if there is a solution coming sometime soon? apart from it being very frustrating to successfully complete workouts and get a massive predicted ftp drop i think i’ve reached a point where ai is prescribing workouts that don’t really fit anymore. i really like trainerroad and i liked the ai, but atm its borderline unusable for me…
All the best (love what you guys are doing!)
Lukas
In my case, this is more than just seeing the prediction drop after outside workouts. This also impacts my upcoming intensity days, as the AI no longer has my training progressing. For example, I am currently a 6.2 PL in VO2, the last VO2 I completed was marked easy, however it was done outdoors. Then my next VO2 workout will be in the PL 5’s, and if I try to manually progress and pick a 6.4 PL, it will say “Not Recommended”, and of course it predicts my FTP to massively drop. I am ok with the prediction being off, however my upcoming workouts are impacted and it is not progressing me even when marking workouts as easy.
So currently I am manually overriding the AI and picking my own workouts on intensity days in order to get my progressive overload.
Based on my experience, most of what I’ve read here, and what has come from TR, this is a prediction issue not a detection issue. But maybe @eddie can clarify?
My experience is that the new system as much better at picking the right workouts and detecting my FTP. The prediction IMO is just a distraction for most people in its current state. My $0.02.
In the past few months, I wasn’t happy with the workout selection (it was always giving me workouts I felt were one or two levels harder than predicted RPE), but it seems like it’s gotten much better in the last few weeks. It could just be coincidence. Having said that, my understanding is that the ALs (PLs) are no longer accurate, so although it progresses from a 5.4 workout to a 4.5, that doesn’t necessarily mean the 4.5 isn’t actually HARDER than the 5.4. It may be, but it may not. All that to say check the structure of the workout, it MIGHT actually look like progression even though the number went down.
I agree. the workouts being picked are easier than they should be so it appears both prediction and detection are a little wonky with predictions being noticeably worse
additionally It has me LOSING fitness at the end of a period of pretty strong training since the last reset
i really wish @eddie could bring us an update although i know he is just the messenger
Hey there, the component that selects workouts doesn’t do it based on PLs or predicted FTP.
Also, using the feedback to try to get harder workouts rather than describe the effort isn’t using the tool properly.
Self-selected workouts are more likely to trend to poor outcomes…have a look at some of the recent YouTube releases for more.
If the AI is basing workouts on its metrics of what your FTP is it matters. No sense training in the wrong zones and going backwards… Mine went down from 254 prediction this morning to 240 (I’m currently 243) after I nailed a Z2 outside workout. Report sent in! I’m in rest week after last block build phase. If the 240 lowering is true which it isn’t (I’m stronger after build and likely around 250) this would mean despite doing all the workouts as directed in build and not failing any, my FTP went down. I’m meticulous about following the plan. Now my specialty phase workouts for next week say I’ll fail them
. I notice the Vo2 max workout is hardly a build on what I was doing last month so it makes no sense I’d fail. I pinned them before this happened so I’ll just do them anyway. Previously I’ve just followed what AI gave me even when it seemed odd.
If it’s true that my ftp went down the app is a failure. That said when I was only riding inside it worked fantastic. I’ve loved it so this is disappointing and potentially disruptive to training. .
Not really, which is why it doesn’t.
The AI determines which power targets on the work you have actually done, and retrofits that determination to workouts in the library. Yes there are constraints on what the AI can select, but it will recommend increasing time allocated if the best workout is a longer workout than is available for your planned session.
Pinning your workouts is working against, not with, TR. The recent YouTube vids might help on why you are less likely to benefit from self selected workouts.
The reality of FTP has always been that it fluctuates data to day and within days, but the industry has pragmatically, but ultimately incorrectly, worked on the idea of a fixed hour of power for decades to prescribe workouts, which is a big factor in why so many people misunderstand the improvement TR have made.
Yes, I don’t want to work against the app either.. I’ve been doing what the training plan told me since January. Coming back with a season off from injury my FTP has build back up from 200-250. So I’m very impressed with the program. Then I got the flu and it went back down to 243. I let the AI keep it the same 243 last block as I felt it made sense to integrate more. Now I am finding upper Z2 workouts very easy. Yesterday for instance I easily completed the workout — Garmin saying 95% execution score. So obviously my fitness didn’t decrease. Never pinned before. It gave me an easier ride for Friday — after the glitch. I changed it back to the previous workout and pinned it. I’m not going to arbitrarily do easier workouts when the app is malfunctioning. It’s never had this issue on me before and it fits the glitch description perfectly.
I’m aware of all the FTP arguments. I’m using the term as TR uses it which is not my hour power, yet. Yes the app changes workouts based on history. This is my issue. I consistently put in the work and am noticeably stronger. If the app predicts I’m getting weaker and changes me to easier workouts it is malfunctioning.
I have also experienced many drops after outdoor rides. However it hasn’t been just one ride, more like a series of rides where it drops 1, 2 or 3 watts. I rate the ride as expected and usually hit the targets. Sometimes I have traffic related stops in the intervals, sometimes not. When we initially switched to the new AI FTP model, my FTP dropped 15 or 16 watts. This past month I have been riding well, hitting targets and feeling strong, but it is the first block where it was actually going to drop my FTP. It dropped 9w in the last week despite hitting my targets for the workouts I was able to get in, albeit I had to adapt the week some due to vacation. All time power PRs for some durations in one of my workouts in the last week where the estimate dropped 3w.
For block I am currently in, it estimated to move me from 340 to 360. I saw the new estimate before a scheduled workout and having it actually detect my FTP. I completed the workout, rated as expected, and the estimate went all the way down to 347. I then declined my AI FTP detection because it wanted to drop it to 336 and I am completing workouts fine at 340. My first workout this week I did inside. It was Pioneer -3 and I did it at +2%; estimated FTP went from 347 to 357. I just now updated my weight on the website and my estimated FTP went from 357 to 346?
I have assumed all the drops after outdoor rides were just the penalty for having traffic related stops. I have balanced it so far with doing indoor rides at +1%. Slowly have bumped the FTP back up from the initial drop when the AI FTP model switched. Its been 6 months though and I am not back to where I was. This recent drop of 9w in the final week when I am feeling strong and hitting targets has not been great for my trust in the predictions/detections.
I can look into your case and let you know what I find. Can I share that info here?
Same goes for you @Barbjarnagin
Sure, share whatever you need to. The strangest behavior to me is that changing my weight impacted the FTP estimate. I did send an email to support about that.
Weight is one of the many factors that go into FTP Detection. How big of a change was your most recent weight entry?
It was around a 1lb drop. I’ve been updating every 4 weeks or so around when I do the FTP detections.
Sure you can share, but I’d like to see what you will share beforehand. Thanks. I also had unavoidable traffic stops yesterday, but I usually do. I don’t have a screen shot of the 14 watt drop before and after the ride as I assumed Ftp prediction would remain in a reasonable range until next week. Last month it did the same (dropped then showed no increase in the final week), but like I said I was ok with it because I needed to integrate more after the flu. I also notice last week it shows me as doing more hours and TSS than expected. The expectation is way less than I usually do. This is odd since I was following the plan. It’s like it changed it after. I did cancel Sunday’s race last week due to weather and deleted it on Sunday. This week’s rides updated accordingly. So all working seemingly fine until yesterday morning post-ride.
So I unpinned the Vo2 and attack sessions next week, they shifted down in difficulty and FTP prediction went up to 250. This makes no sense to me, why the huge shift after yesterday’s ride? Then I just out of curiosity unpinned my z2 ride for tomorrow. It is now slightly easier. The sessions next week went back up to what I had previously pinned and now it is predicting 251. I get FTP is a rather flexible number in a range. Just not sure why one ride shifted it so much! Coach Jonathan says we can change rides depending on how we feel that day within reason. I like that the app has that flexibility.