Accuracy / meaning of the predicted FTP

TrainerRoad Software
1

I’m wondering if there’s any beta users who could comment on the accuracy of the predicted FTP with the new update.

I am a new TR user and previously I was estimating my FTP with Kolie Moore Baseline and I also did a TT race in Zwift recently so I am pretty confident in my ability to ride at 260 ish watts for about 45+ min. Adaptive TR (pre-update) had me at 262 and the new AI FTP prediction has me at 260. OK, seems close enough.

Now the FTP predictor has me at 276W in 25 days and 13 workouts. This is kind of unbelievable.

I’m wondering if TR will bump up my FTP to 275, but bring my levels down and have me doing only slightly harder workouts. Or is this prediction saying my “real” FTP is going to go up 15W and I could theoretically go do a Kolie Moore Baseline or a Zwift TT race at 275W?

Is there anyone from the beta program or TR who could comment and help to set my expectations?

2

I think it’s saying you’re going to be able to do Threshold workouts that are level 3 for 275W FTP, e.g.

  • Kololo, 3x6-minute intervals that range from 105-106% of FTP (289W - 292W) with 2-minutes of recovery in between
  • Yazghil, 2x10-minute intervals that range from 98-99% of FTP (270W - 272W) with 3-minutes of recovery in between

I made a Kolie Moore Baseline workout in Workout Creator (more or less) and it’s rated “Threshold 8.5”. Since an 8.5 Theshold workout is significantly harder than a level 3, it’s unlikely you’d be able to do a Kolie Moore Baseline at 275W.

4 Likes
3

25 days out it’s not extremely accurate, it will adjust as you get closer and become more accurate. I use it as motivation to keep doing workouts in the extreme heat we are having on the east coast of Australia right now. Currently, considering getting out the ice vest to train today.

2 Likes
4

My opinion so far is not very accurate. Its pretty optomistic at the outset and any deviation from the predicted difficulty results in wayts being chipped away.

If you are seeing wattage increases that appear to be too good to true then its very unlikwly that will be a reality come detection day. But thats using the old ramp test FTP.

Who knows with the new threshold 3.0 model. Are increases via the new model even comparable to old? I don’t know as I dont quite undwrstand the new model.

1 Like
5

Unfortunately the predicted AIFTP was not available to the large beta tester group until fairly recently.

I don’t think anyone will be able to answer that question until the 6th Feb.

I was predicted what I thought was an unrealistic rise and now I’m half way there the prediction is far more realistic.

Although to be absolutely fair to TR…. I’ve not 100% stuck to the planned workouts…. I’ve added a few runs and commutes etc

7

… and rank them “hard”. Not that you won’t be able to do harder ones, but as they get harder you are more likely to rank them very hard/all out/fail.

8

Not exactly, from my surmise the system is calibrated so that if your FTP is set correctly then you will most likely rank a PL 3.0 threshold workout as “hard” (i.e., the middle of the 5 level ranking).

I don’t think that means FTP is being predicted from the PL of the workout.

I think there is an open question about how TR’s FTP correlates to other measures of FTP - mine looks a bit on the high side, but not crazy amounts - but that’s another debate which is currently raging in about 5 other threads so maybe not worth opening here :smiley:

9

If you can’t do level 3 threshold workouts on your old FTP then your AIFTP is gonna drop compared to what you had before.

This seems right in my opinion… level 3 threshold workouts don’t spend an awful long time at threshold….

4 Likes
10

I’ve been on the beta, and in terms of selecting the right workouts, the algorithms were always bang-on. The workouts were always in the right zone and had the right difficulty. When I struggled, it was for reasons TR AI couldn’t know (me becoming sick and kids being rowdy).

If I were you, I wouldn’t worry about TR’s FTP algorithms. If the name FTP bothers you, give it a nick name that’s different.

1 Like
11

As of about a month ago I could do 3x8 at about 280-285. I did 280, 280, 287. Zwift - 8 Minute Intervals | Strava If I had to guess I’d have probably rated that very hard. So that 3x6 at 276 is probably moderate-hard at my current fitness. With some fitness gains in the next month I’d guess that a 3x6 at 276 might even be easy.

In any case, if TR is going to reassess in a month and normalize me to threshold 3, I guess I don’t understand why TR didn’t originally onboard me at a normalized threshold level of 3 with whatever aiFTP that would be. My current threshold is level is 4.8. Starting me at say 270/3 instead of 260/5 would make the baseline the same as whatever TR is going to give me in a month?

12

Hey @Xicutioner4768!

I think these two threads will help:

And as other have mentioned, the AI FTP Prediction has a butterfly effect like the weather forecast, as you get closer to it it becomes more and more accurate. You’re still 3 weeks out from it, so I would give it a bit more time :slight_smile: