I just wrapped the speciality phase of a gravel “maintain FTP” plan. Over the course of this phase the AI FTP prediction has been slowly inching up to about 10w over my current FTP… until a few days ago, when it suddenly dropped to 1w below my current FTP. So, not surprisingly, when today rolled around and FTP detection was available, it came up with the 1w under current value.
In fairness, I had selected a “maintain” plan, and that’s what I got. But to see the number creeping up only to have it reverse course at the tailend of things is kind of demoralizing.
I would have a lot more faith in the product if it could explicitly tie decisions I made back to this number. Was it that missed recovery week ride? Was it that one sweet spot workout that I rated very hard instead of just hard? Or that time I swapped out a threshold workout for a nice long outdoor ride?
I realize the prediction feature is still in beta, but to be honest, I wish I could just hide it. It’s a distraction.
Think the reality is that most people who have been training regularly and for a while, won’t be able to increase their FTP within four weeks. And especially not if their training isn’t targeted to increase FTP. That doesn’t mean you can’t increase it at all, but It’s a long process, and the 28 day prediction window is often too short to see it.
All 3 of those probably impacted your FTP going down. Threshold likely the biggest impact, then sweet spot, then last your recovery ride.
That said you still got an increase over the month on a maintain FTP plan. If you are just trying to maintain I think that’s a big success.
The FTP prediction is still in beta so it’s not a completely finished product yet. You can hide it if it bothers you.
Mine always starts high and then life gets in the way and it settles somewhere between where I am and where the AI predicts I’ll be. I like using it for guiding my training decisions when I need to move things around on the plan but really don’t pay attention to it.
Did my first group ride last night in months and lots of people were shocked with how fast I was riding. And I wasn’t really pushing above tempo on the ride. Good to see the training working in the real world.
We are using ostensibly using TR to increase performance, and FTP is the accepted benchmark for comparing performance, not necessarily between different athletes, but within an individual athlete’s training history. So it’s no surprise that seeing declining, or stagnant FTP as a result of the TR AI update is concerning for many athletes here. The concern is well warranted.
There is a narrative from TR that FTP should be seen as more of a training calibration number and less of a performance metric. There may be wisdom in this, but just listen to Jonathan’s Leadville strategy podcast and hear him use FTP, and FTP prediction, as the performance metric used in targeting his goal time. We all do this.
Leaving aside that “entire self worth” is hyperbole, can you blame anyone for caring a lot about this? It’s the single number by which performance is measured on the app. After you put a ton of effort into all your intervals, and see negative results, I would say it is (at the very least) frustrating.
Spot on. Takes us back to “what really is FTP?” Anecdotal, but I was in slump following the holidays and needed a boot camp to get me going again. I also hadn’t also taken an FTP test outside of AI FTP detection in ages so was curious to see where I was at. I completed the Zwift 4 Week FTP Booster as I’d had lots of success/wattage gains in the past with this as a beginner before I started using TR. Made me realize a few things and I might not be stating this eloquently: First, I saw FTP increase as I worked on specifically increasing my FTP. Lots of work at threshold or above current FTP. Second, some short FTP building plans will get you really good at putting down numbers to prepare you for ramp tests and 8-minute protocols. I was hitting lifetime 5 to 10 minute power PR’s. I became good at the test format vs the actual material, and while that’s great if my races are ramp tests, they are not and training to ramp won’t help me train to win as I don’t think I could have the percentages of power for an hour. Lastly, made me realize how much better TR plans are than Zwift! There are rest days, better warmups, sensible recovery intervals. I completed the training block but it was terrible! I’m back on a TR plan now and will see how the numbers look after a few weeks.
When I went to bed last night my predicted FTP was 299 (up from 297) in 11 days. I woke up and it’s now predicting a decrease to 292 in 10 days. I made no changes to my plan in my sleep. I’ve been an advocate for TR and fan of the new AI stuff until now. Seeing the number is not only demoralizing - it’s now got me questioning if TR is actually the most effective way for me to get faster. For the first time in a while, I’m thinking about ditching it and just riding my bike.
(My calendar is public if TR wants to take a look.)
Same thing happened to me. In fairness I did stop a workout early but it dropped like 10w after one failed workout which was pretty demoralizing and kind or frustrating since it doesn’t explain why. Itd be cool if they could add something that kind of explains why it dropped, I think that would make it a lot less frustrating for me. I am considering hiding the AI prediction for now so I don’t obsess over it lol.
Ditto, And even though I’m upping all the prescribed workouts to harder workouts overnight my predicted ftp dropped 8w. But I’m ok with it. It’s in Beta. It says that in very big bold letters. I’m sure they’ll tweak it to be more conservative.
I just completed my four week block and Ai ftp detection was spot on. I completed all workouts with the caveat that I do all recoveries after hard efforts at 102 percent. I do this because I think there is a few second lag at the start of each hard effort and cumulatively that adds up over a four week block.
FTP for my block jumped the predicted nine percent and the first workouts if the new block are pretty hard but doable. The next predicted ftp jump is also expected to be nine percent and I really want that to happen so I am motivated
