Comparing actual FTP Test to the AI FTP Detection Feature

AI predicted mine > 10% lower, and I expect its due to an odd situation + the model being overtrained.

I was at FTP of 254 and crushing workouts. All of a sudden my allergies kicked in and I failed a stretch I thought I would be able to hit (this was prior to realizing my allergies kicked in), and then I failed an achievable workout that I thought would be easy. At this point I realized my allergies were exacerbating my asthma, took a rest week and upped my antihistamines.

After the rest week AI predicted me to be at 210. Technically I didn’t do an FTP test, but since the rest week my legs have felt the best they have felt all year. I am nailing workouts based off me estimating my FTP at 265

The weird situation was failing workouts due to allergies and immediately doing a rest week. The reason why I think the model is overtrained is because a properly fit model wouldn’t predict a 20% decrease in fitness when someone was nailing workouts prior to failing two and then taking a rest week. Especially since I still did 6 or 7 hours of endurance riding during my rest week.

I have also noticed something interesting along the way. There seems to be a ‘bug’ in the app not letting you use FTP prediction within 2 weeks of previous prediction.

I couldn’t remember how long it had been since I used the FTP prediction so I tried again too early (at day 10) and it said ‘you can only used FTP prediction once every 14 days…’ or whatever the message is. Then it asked if I want to update to my previous prediction of 221. I thought that was odd since my previous prediction was 210. I tried it again the next day, it asked if I want to update to my previous prediction of 240 something. I did it again at day 13 and it asked if I want to update to my previous prediction of 260. Anyway, it says you can’t use it but it looks like the underlying mechanism is still tracking your fitness (this might have been fixed since I updated the app this morning). I have like 5 hours until I get to use the prediction again. Even if it stays at 260, it is a lot closer to how I feel than the previous prediction of 210.

2 Likes

As Chad has mentioned multiple times, even in the last podcast I believe, our FTP does not change daily but our ability to express our FTP does :grin:

1 Like

The situation you describe in your last paragraph certainly seems quite odd. Maybe send that over to TR support as it may help them identify some underlying issue. Glad your FTP is going up so fast though…

Look at the definition. It definitely does, so he is wrong here. FTP is FTP at a given point, there is no expressing it. Oh your FTP is the same but you cant express it on a certain day. Sorry just plan wrong… however it is not important.

2 Likes

I don’t see where you’re getting that info. If your FTP actually changed that much then training zones would be irrelevant and we’d have ramp tests before every ride. FTP is something that builds and drops slowly. You can sleep poorly one night and be off for a workout and need to drop the next day’s workout intensity, but that doesn’t mean your overall fitness and FTP have dropped-just your ability to express it that day. In other words, the physiological adaptations in your muscles, heart, mitochondria etc that brought you to your FTP do not suddenly change in one night, but your ability to reach that potential does.

2 Likes

No because the ‘functional’

I have to say that TR would likely be rather pleased with the results of the poll in the OP. Reading through various threads over the last few months I had gotten the distinct impression that folks felt that, overall, the AI FTP detection had seemed a bit aggressive in its calculation. However, with 52 responses thus far, approximately 76% have seen their actual FTP tests come in either “spot on” (+/- 1%) or they actually outperformed the AI Detection by (1.1% to 5%), which in-and-of-itself is rather marginal. Clearly this particular selection group has not seen an overestimation. Kudos to the AI FTP Detection team :+1: .

2 Likes

An update on my situation and some thoughts.

I was encouraged to see if I could complete a level 2.2 threshold over-under workout at the AI FTP of 310 watts. I completed the workout but I would say it was a 9/10 workout; it felt like a VO2 workout (ragged breathing during the last half of each interval, HR up around 95% of max) and I had trouble holding power during the 2 minute under intervals (about 5% under the prescribed watts). My survey response was “all out”.

I know I completed the workout but should a level 2.2 threshold workout be that difficult?

1 Like
  • What was your actual Threshold Progression Level at the start of that workout? This matters to have a basic expectation of how “hard” it might feel.

  • PL to Workout Level gives “Difficulty”, where Achievable will likely feel “easy” while a “Breakthrough” might crush you.

  • Prior to knowing that PL info, which is super important, no… that doesn’t sound right to me (unless you were Breakthrough or high Stretch). What you describe seems like VO2 work vs the burn & “not enough relief” of how I think a typical O-U workout feels.

  • What rating did you assign to the workout?
  • If you are following a TR training plan with AT enabled, what changes (if any) did AT give for following Threshold workouts?

Interesting discussion here. Do you think your FTP is different at the start of the ride vs the end? At the beginning of most rides I would be able to output my FTP power for ~1hr, but after a 3hr race I’m fairly certain I could not. Does that mean my FTP is now lower at that point?

If so, do you know of any software I could utilize that has a variable FTP during a ride? For example, my FTP at the start of the ride is 300W, and it decreases based on workload throughout.

  • What was your actual Threshold Progression Level at the start of that workout? This matters to have a basic expectation of how “hard” it might feel.

2 weeks ago, AT had me at a 3.3 PL for threshold after bumping me up to 310 watts. I did a 3.4 PL threshold workout and, looking back, I probably failed it given the power ranges. The next week, AT gave me a 4.0 PL threshold workout the following week at which point I completely failed and could only do one interval.

AT then gave me a level 2.5 threshold workout for last week, due to time constraints I had to knock it down to a 1 hr workout which was a level 2.2 threshold. This is the one I just completed this weekend which I felt was a 9/10.

  • What rating did you assign to the workout?
  • If you are following a TR training plan with AT enabled, what changes (if any) did AT give for following Threshold workouts?

I assigned “all out” to the post-workout survey. Unfortunately I deleted the plan since then but IIRC AT kept the PL at 2.2 for threshold on the next workout.

What you describe seems like VO2 work vs the burn & “not enough relief” of how I think a typical O-U workout feels.

Yes, exactly. From experience, when I do O-U workouts properly I will start to feel the burning ease just a bit around the halfway mark of the “under” interval. At 95% there was no relief at all. I had to dial it back down to 90% to get just a bit of relief. Even then, it was me crashing down to about 80% FTP after the “over” interval and then trying to catch back up to 95% FTP.

1 Like

Thanks for the info. As above in the prior discussions, I think your FTP is well below that 310w value and is worth manually adjusting based upon some targeted guessing at the least, and maybe best to perform an actual test in the proper format and completion to get you a better point to build from. The current state seems unworkable to me.

1 Like

Agreed. I realize that FTP is not necessarily max power for an hour, but based on my recent tests and workouts I could not hold 310 watts for more than 10 minutes. I also believe there to be a huge anaerobic contribution to that power too.

I have banged out anaerobic workouts at what AT considered PL 7-10+ while struggling with threshold workouts at PL 5 and below; indeed those anaerobic efforts felt easier than the threshold workouts. Based on what I have heard in prior AACC podcasts, this would put me in the an outlier group.

I want to be clear that I am not faulting TR’s AT or AI FTP product, I have personally met several cyclists that have grown leaps and bounds with TR’s program; these are seasoned cyclists that started TR after decades of racing and riding. I wholeheartedly support TR’s efforts to make athletes faster and I’ve been paying for their product despite also being coached for the last 20 months. I will probably continue to pay for TR because I enjoy testing things out and I like how TR is shaking things up in this sport.

Technically yes, I say this only influenced by some high level coaches and to me it makes sense.

No. I don’t think it is important or matters. FTP is totally over-rated in terms of training, and training zones, it’s only important for generic power based training platforms. I was just commenting in the context of the thread.

1 Like

On my most recent FTP test day, AI suggested an improvement from 207-211. When I did the actual ramp test I got 222. Glad I did the actual test.

1 Like

Just adding another datapoint…
Last Sunday I did my first ever (longer) race, 100 km, 1300 m elevation gain, 2:45 h for me. Had marked this as ‘B’.
Preparation was not optimal, as I was skiing and on family visits for 16–6 days before the race. FTP from the last ramp test, 4 weeks before, was just above 300 W.
As I was forced to start a bit behind, end of block B, I spend the first 55 min after the race was opened to jump forward group by group. For this time, I rode with AP ~= FTP and NP > FTP. I also continued another >100 min averaging near that FTP value with 10 and 15 min climbs at/above FTP. Thus, that FTP number might not be so bad;-)
TR had a ramp test scheduled for the Monday after the race, which I left foreseeing to use AI FTP detection… but that wanted to put me at 290 W…
(I thankfully rejected and keep training for my A race in now 3 weeks using the “old” FTP,)

1 Like

@yokuha That sounds like a great performance… I might be misunderstanding, but if you rode the first 55 minutes at approximately your FTP of 300 watts (with a NP actually above that)… and then did the next hour and a half+ (100 min) averaging near FTP doesn’t that actually point to your FTP likely being above 300 watts now (and not by an immaterial amount)?

1 Like

You understood right.
Not sure what that says for my FTP. AI detection assumed it to be smaller than the previous/current FTP value;-) Which I also find strange, but it might make sense somehow?
And I am still bad with tests… :-o

Just now (today, like an hour ago):

  • AI FTP 292 (only outdoors ride the last month)
  • Didn’t trust it (because of only outdoors rides the last month), decided to give it a try… 20 min FTP Test: 294.

Almost the same and would not have had to undergo/suffer the 20-minute test.

Also I have a question: I have an evaluation scheduled for Tuesday because I will start a new training plan, but since I did this test it does not allow me to use the AI FTP Detection (too soon)… how can I tell TrainerRoad that I do not need a new evaluation and keep my current ftp and be able to replace the ramp test with a workout?

1 Like

I’m not sure you can. What I did was add the following weeks workout and do that in place of the ramp test (in my case it was a vo2). AT gave me a updated PL workout next week and so on down the line.