Sweet spot on the other hand looks like a piece of cake!
2 Likes
Funny that the endurance ride has higher TSS than the other two. Does a Score of 150 not imply that all the ride would be at Threshold, more or less?
Curious to what the TR team thinks about this one. I got sick during my last training block. I ended up taking 5 days off the bike completely. I attempted a sweet sport workout way too early, then did a few days of endurance and one day of tempo in the recovery week. Understandably, my FTP prediction went down to my current FTP level (351).
Now I’m back on a training block (Base), completed my two sweet spot workouts without any problem and AI FTP Prediction went down to 347. Today was the day for a new AI FTP Detection, but I decided to complete the Threshold O/U workout before detecting my new FTP. I completed Bounty without any problems (3.9 workout level), did the AI FTP detection afterwards and it came up with 349.
Now a 2W decrease in my FTP is basically a rounding error, so I decided to keep my current FTP of 351 to protect my ego a little. I thought I performed well enough post sickness to keep my current FTP, so I was surprised to see AI FTP detection take me down a little.
I’m curious to what the TR team thinks about how the AI is interpreting my performance. @Caro.Gomez-Villafane - what do you think? Thanks!!
Here’s my calendar: Log In to TrainerRoad
How well does this new ai FTP correlate to a ramp test and a more standard way of measuring FTP? That I can use a % of for pacing. I am doing the Mallorca 312 in April so looking to average around top end zone2 and do the climbs at tempo/low sweetspot.
I went from 286w to 298w during the beta phase of this new AI FTP. I just got my 28 day update today and it went to 313w! It’s now predictiing my next 28 day FTP will be 336w!!
The highest FTP I have ever seen using a ramp test was 309 and that was a five years ago! I’m 40 now, so not getting any younger. I have hovered around 270-300w for the last couple of years. I recently did an effort on a climb during a training camp where it was ~300w for 33 minutes and I wasn’t totally fresh so I feel like my FTP is probably closer to 300w but not sure it’s 313 and no clue how its going to be 336 in a month. Unless I have made some massive breakthrough in my training. I have been doing triahtlon for the last few years and switched to a pure block of cycling for the last few months but even so that seems a bit nuts to me but I will trust the AI to pick my workouts as it hasn’t failed me yet!
1 Like
According to tr aiFTP detection is estimating FTP. In theory then you should be able to use that ai number the same way you use FTP.
I think the reality is different. Personally the old model worked better for me.
2 Likes
I completely agree but that is what TR is saying so 
.
At this point they have seemingly decided on their model and I don’t think that’s going to change until the next one comes along.
2 Likes
Sorry to argue a bit here. But this is simply way to confusing or complex for the average user.
I just want the modelled power-duration curve so I have an estimate on how much power I am able or deliver for a given amount of time.
I don’t care about level 3,4,7,9 at threshold/sweet-spot/supra-threshold or whatever. That is to complex for me to get my head around. It needs to be easy understandable how to translate it into real world pacing estimates
2 Likes
The interesting part now is I read all these complaints on why does my FTP prediction change.
The focus is on the outcome not the logic to get there.
2 Likes
Maybe you fall into my theory here:
1 Like
I don’t think so. Granted I haven’t done a 20 minute test for years but I found those pretty accruate if not horrendous to do. I also found the ramp tests quite close to reality as well.
Looking at my numbers in my power ranking there isn’t really an outlier in the shorter durations they are all quite similar across the board. All time I am around the 80-85th percentile in 5,10,20 and 60 minutes. This year it’s pretty similar as well, 96-98th percentile across the board for 5,10,20 and 60. My 5 second and 1 minute are lower this year and all time as well but I guess that makes sense as I do triathlon and long sportive type events so have trained for that type of sustained power.
I did my 90 minute threshold workout today with my 5% FTP bump. It was a lvl 3.4, with 3x8 minutes at 310w with 2 blocks of 20 minute tempo at the start and end of the workout. I rated it as very hard but I finished it.
1 Like
The levels are a training measure within a targeted training zone that’s all. If you want a PDC, do some capacitive efforts out in the real world. This update is a training system, not a “how do I race given my expected power duration” but the platform does have the tools for you to get that too.
RE the capacitive effort testing, that’s the point of AI FTP and the level system is so that you still get accurate workouts that boost your fitness without the need to use a training day to test. And if you have a bad testing day for some reason now you lost the testing day AND probably the 1-3 days of rest you took in anticipation of the test. Or you end up using a block of training on a number that’s way too low because you simply weren’t feeling it.
The other method is to use the Trainerroad power analyzer in the desktop. Use a history of 90 days and it’ll give you a current PDC.
7 Likes
Looking forward to hearing from you.
If I may humbly ask for something:
I my opinion, it would help a lot of athletes (I sure would help me 
) if you would discuss the matter AND communicate it coming from two angles:
- How the AIFTP determination helps inside TrainerRoad (eg: chose the right workout)
- How it helps by comparing to the outside world for pacing and stuff (or just very clearly say “it doesn’t” so that you at least end discussions)
From what I hear the messages are just not clear:
Regarding 1. I hear: “well TrainerRoad now is better choosing the right workout. So new AIFTP is good. But the AI doesn’t need the number anyway…”
→ personally I am very satisfied with the workout it chooses. I just don’t get the logic of the argument.
Regarding 2. All things said by TrainerRoad staff sounds to me like: “Well if you look at the literature and Keegan and .. there is no 100% definition of FTP so it doesn’t matter that ours is completely incomparable with anything in the bike community” (sorry if I sound mean. It is not meant like that. I just want to point out how it can be understood by clients like me)
I also read and hear the comments about pacing in the podcast and in this thread. They go like this: “you can just manually adjust your ftp until it predicts a certain workout to fail and then you might estimated the pacing from this value…” I mean..okay 
I guess even a 20min ramp up test would be faster to do than the math and simulation you ask me to do
PS: again, I want to highlight that I really do like the program and how it suggests trainings. To me it’s Just that the communication on how it wants to help me to compare and predict my current level to the outside world could be improved
8 Likes
Language and terminology is important. The new AIFTP is NOT FTP. That has been made clear from TR but is confusing for its customers and is not a reference against what the community understands as FTP.
A proposal:
Rename the current AI FTP to something like Training Reference Power (TRP) or AI TRP this is then clear that it is a power level used to define training levels.
Generate various models of FTP with the known athlete data using the various methods in the industry so they can then be compared to what exists elsewhere and display them on maybe the Power Records screen along with the curve, maybe rename that screen to Power Records & FTP
2 Likes
What you’ve written was raised umpteen times and more both during the beta program and since, but TR chose to stick to their guns and did not make that change, so all the indications are that that is not going to happen else it already would’ve happened…..
5 Likes
Nate has shown on the forum that they have the ability to predict 20min and 60min power for an athlete, and of his intention to roll that out as a feature. No ETA as yet though.
4 Likes
Just outstanding improvements in the app. So pleased. Just perfect now and exciting for the future. Like other posts have said. I was concerned with the predictions but having listed to the podcasts and used it, it’s brilliant!
3 Likes
This is the way
1 Like
And if you have a bad testing day for some reason now you lost the testing day AND probably the 1-3 days of rest you took in anticipation of the test.
You shouldn’t need to take any rest before doing a FTP test. I do Kolie Moore style FTP tests - they don’t require you rest beforehand and they’re a decent workout in and of themselves. Working exactly at FTP isn’t too sensitive to fatigue in my experience (unlike working above FTP).
Here’s my cal from the last time I took a test - no rest beforehand.
2 Likes
I am pretty sure that TR has said that aiFTP is an estimation of FTP?
2 Likes