Any downside to manual FTP adjustment with the new AI system?

[eddie]

I would give that next workout a shot. If it’s really unmanageable, turn down the intensity until it feels right and answer the survey honestly. I’ll take a look if that’s the case, so just let me know.

Regarding your recent SS workout, I’d definitely classify that as “Very Hard” since you were really on your limit. That makes a difference in how we prescribe your upcoming workouts and FTP.

For reference, here is what the very hard survey response looks like. How would another 6 minutes of intervals have felt had I asked you 60 seconds before the end of the last one?

---

Very Hard
This ride was very difficult to complete. This ride tested you. If there would have been one more interval, you wouldn’t have been able to do it.

[DLanzone]

Thank you - I’ll update my response and we will see how things go this weekend!

[Rizzi]

@eddie

I’m curious about the Wattbike situation, though. How much of your training history has skewed power data?

@robmca

I reckon the Wattbike may have been over reading up to 40watts, but I’m not sure for how long, I think it was a gradual increase over time.

I’d hope that one parameter of the user’s individual ML model is the vendor of the power source and that you could erase a vendor from your individual data and then retrain the model. That would be a fix for that situation (and BTW could also help with the often addressed wish for different FTP by power source or outdoor/indoor).

Would like to know if this is the case and if not…why not ?

[kevistraining]

@Nate_Pearson

This was a bug, it’s been fixed now.

This makes sense for me.

I think the prescription of workouts that are too hard to complete (wrong RPE) is a bug.

I think the rest of this thread is a simple intensity versus time in zone - potentionally even just a personnel preference thing? - like @eddie pointed out earlier in the thread

[joguar]

Short sweet spot intervals are a really weird thing. Sweet spot is meant to be able to beheld for a decent duration and I know I’m weak at these long sub threshold efforts and strong at short high power efforts (intervals eFTP has been between 260-280W and old TR AIFTP was 227-234). I stack up at top decile or higher on TR short term power rankings and bottom 40% for long term power. Now AIFTP has me at 243W with P 257ish and both feel super threshold.

Bought a new big fan so hopefully with adequate cooling I can complete workouts but my HR definitely went to LTHR on SS workouts before (234W) and now above LTHR after the AIFTP bump (243).

[robmca]

I suppose I could go through all the activities and delete them one by one!!

[FlatFylde]

Does where you set your FTP to outside TrainerRoad matter if you want to execute the best possible training within TrainerRoad?

[LacticacidMCB]

@robmca

I suppose I could go through all the activities and delete them one by one!!

Or mark them as not to be included.

[grwoolf]

@joguar

Short sweet spot intervals are a really weird thing. Sweet spot is meant to be able to beheld for a decent duration and I know I’m weak at these long sub threshold efforts and strong at short high power efforts (intervals eFTP has been between 260-280W and old TR AIFTP was 227-234). I stack up at top decile or higher on TR short term power rankings and bottom 40% for long term power. Now AIFTP has me at 243W with P 257ish and both feel super threshold.

Yeah, I’m not a fan/believer is short sweet spot work. Maybe do one or 2 workouts like that if getting back into training to get the body/mind ready to start pushing decent watts again, but you might as well be doing threshold workout if you have limited time. The advantage of sweet spot is that you can push a lot of it without completely destroying yourself. If you aren’t doing a bunch of time in zone, there really isn’t any upside compared to threshold.

And your intervals.icu eFTP difference just highlights how different some of these metrics can be. I have the exact opposite experience with intervals.icu since I focus on longer stuff and weak on short efforts. It has my eFTP at 274 while TR AIFTP has me at 318w. And my “real” number (as I define it) sits at 302w. My intervals eFTP is always dramatically low because it only looks at short intervals to project FTP. I could do 5 hours at 260 watts and it would still keep my estimated FTP at 274 because it doesn’t have the logic to look at durations over 1 hour (which is really dumb). But if it sees a maximum effort 5 minute interval (which I’d never do this time of year), it’s more than happy to project a higher FTP based on that (really, really dumb IMO since it has a big anerobic contribution that varies a lot by individual).

[DLanzone]

Hi @eddie,

I completed the threshold workout today (Crested). I decided to rate this one an All Out effort. It wasn’t “drop to the floor dead” all out, but there really wasn’t any “recovery” during the unders, just a steady build of lactate throughout the entire interval. The unders were at 245W, basically my threshold, while the overs were at 271W. So, really just three sets of At/Overs.

I guess I could have rated it “very hard”, but I don’t think pushing into higher intensities from here would be very productive.

Seems like everything is getting adjusted down accordingly now, but let me know if you think a manual adjustment down of the FTP is a more productive approach.

Pretty tough first week back on the platform, ha.

[kevistraining]

Just checking in with my “FTP too high” support group

My AIFTP is currently 334w but my hour power is 317w (old AIFTP was 325w).

So far it seems to be working as intended if the intention is to keep me at the point where level 4 threshold workouts are hard/very hard. I did cloudripper -2 last night and that’s exactly what the RPE of the ride as a whole was.

I’m happy to see that in 28 days it’s scheduled me a level 5.4 threshold workout - if I’m at the point where I can do a workout like that at my current “high” FTP I’ll be extremely happy.

I’m still not sure that “over-under” threshold workouts “feel” right though - cloudripper -2 predicted RPE was spot on but the “unders” were at my hour-power watts - the workout as a whole was fine because each of the 3 sets was only 9 mins long.

I was planning on manually reducing my FTP for my next block - but t I think I’m going to stick with my high AIFTP for a while longer.

[DLanzone]

Great job nailing that workout!

For me, unders at my actual 1 hour power would probably feel ok. Mine were at my 30 minute power, which for me is very tough as I was basically in race mode when I set it.

The next threshold workout has my unders at my 1 hour power, so it seems like it is reaching a similar conclusion for my profile. So initial indication is that it’ll reach that equilibrium eventually.

Unfortunately, my SS workouts are still hanging out at (~99%) and around my true FTP. So I am still more curious about the more macro adjustment of changing FTP manually. SS workouts are my favorite, and I’d like them to feel manageable again more quickly. I’m quite battered after this last week of overestimation. I’m supposed to do my first dynamic endurance ride today but the legs are pretty smoked. It’s only the “start of plan” phase.

Tl;dr - threshold workouts are adjusting sensibly after very hard and all out SS and threshold workouts. SS workouts are somewhat more rigid and still look like threshold workouts over the next week. More pain needed to get all zones in line.

[kevistraining]

@DLanzone

indication is that it’ll reach that equilibrium eventually

I’m coming to the same conclusion.

I think that the correction/normalisation to get everyone doing the right kind of workouts was needed - but I think there are a few cases where there has definitely been an over-correction and it’s gone too far the other way.

There is obviously evidence of people getting legitimately too low and too high FTPs - but it seems to be working as expected for the majority.

[BCM]

Remember - while it is an estimate, hour power isn’t necessarily your FTP. But that’s actually especially true after you take an FTP Bump and start training at that higher FTP. For me - TTE resets pretty low, and you’re probably in that 40-50 minute mark, and you start having to push it back up again and “Bed In” the new FTP. (Basically another way of saying your progression levels reset really low and you have to work them back up)

So with that in mind, 334 / 317 doesn’t sound totally out of whack.

[clemclem]

This update has been surprising and at times, frustrating. I’ve spent a significant amount of time trying to understand the behavior of the new system.

I’ve been testing the calendar since beta switching between my actual FTP and the inflated value to compare the resulting workout prescriptions. With my real FTP (in WKO5, incl. TTE. Previous AI FTP from TR was bang on), the workouts are well aligned. Appropriate O/U targets, baseline is 3×10+ at FTP, or 2×20+ at SST. With the higher FTP, however, I’m assigned level 1-2 FTP workouts such as 5×5 at (physiologically) 110%+ FTP, which places the effort in max aerobic territory and doesn’t match the intended training stimulus.

The system appears to struggle in a few specific areas:

• Endurance rides above LT1 (workaround: manually setting conservative endurance rides)
• O/U workouts being at FTP/over (workaround: avoiding TR for these and using Intervals.icu values and the workout builder)
• Low-level FTP workouts (levels 1–2) being prescribed at physiologically suprathreshold intensities (e.g. 5×5 min) (workaround: manual FTP adjustment to avoid these cases)

That said, my predicted FTP is now trending back toward my WKO5 values, which makes me cautiously optimistic that this was a calibration issue specific to our use case, i.e: trained cyclists with high workout levels in the old system.

I’ll continue using TR for now, in the hope that the system stabilizes. Despite the criticism above I still think it’s a strong tool that mainly needs some fine-tuning.

If I were to suggest a change, it would be to calibrate AI FTP using a level-1 baseline such as 3×10 at FTP, rather than the current level-3 effort. This would likely reduce FTP inflation for many TR users (even if I understand it wouldn’t be popular with everyone), while aligning more closely with physiological markers and remaining consistent with TR’s system.

Feedback / feature request: it would be very valuable to have a slider allowing users to bias the system between intensity and TTE. I understand the core demographic includes newer cyclists, but for athletes past the rapid noob phase, a TTE-biased approach would resolve many of the issues described above.

[DLanzone]

Just want to give a strong endorsement to your feature request.

[brenth]

I agree with your comments. I think my starting point for FTP is off. I changed FTP to my old FTP to get back into zone. Workouts prescribed are definitely a different level of RPE. I do wonder if we are going to end up at the same place eventually. I just wasn’t up for a few weeks of very hard rides continually being adjusted.

[kevistraining]

@BCM

So with that in mind, 334 / 317 doesn’t sound totally out of whack.

Agree - for context the old AIFTP that I was completely satisfied with was 325w

[Kaj_Johnson]

Has the “calibrating everyone to a Threshold PL 3.0” changed? I just completed a 20 min test then ran AIFTPD. The values were only 10w apart so decided to accept TR’s estimation and get on board with trying to feed the system data and let it figure me out. I threw in a TR AI Threshold workout just to see what it would be recommending and the workout was a PL 4.0. I thought it was supposed to be a 3.0 after accepting an AIFTPD?

[jsiegrist]

I only ever had one right after the first ftp assessment, and it was a 2.8. Everything since has been over 4 even after my most recent ftp update that went up.