I Still Use Ramp Tests

After using AI FTP detection for both myself and my two teenage XC racers, I found that AI was ramping up our FTP faster than our will to perform :slight_smile:

We love watching our numbers improve dramatically with AI FTP, and appreciate the vote of confidence that the program gives us. My gut tells me that it probably is very accurate IF we were machines.

This is why I still use Ramp Tests and encourage the kids to do the same. Ramp Tests measure your will in an indirect way. In other words, how far do I want to push it, as opposed to how far I should or could push it if it wasnt for that whiny I dont wanna voice.

After a few month of using AI FTP, I found myself and the kids burning out on TR and starting to fail or skip work-outs. Once we returned to Ramp Tests, the FTP measured down a bit, and the burn-out subsided.

We currently alternate the two for no other reason than we can do a work-out as opposed to Ramp Test. I find it interesting that since doing so, AI FTP is now aligning more with the Ramp Tests. I assume, because our work-outs are geared around the lower FTP scores.

Is this optimizing our training plan? I think so. We are still here!

1 Like

Hey there @hoge,

Thanks for the feedback! If Ramp Tests or a blend of them with AIFTPD are giving you the best results, I say stick with what works.

Ramp Tests are a snapshot of your fitness on any one given day and can be affected by external factors. If you start feeling physically fatigued, you likely won’t test as well and could end up with a score that’s slightly lower than your true potential. This is why we recommend most athletes allow AIFTP Detection to do its thing.

The same goes for your mental readiness and toughness. One thing that makes Ramp Tests so hard is that you need to go deep. We’re talking allll out in order to get a truly accurate assessment, which is really difficult. This can be especially hard for those at a young age. :zombie:

You mentioned that “whiny I don’t wanna voice,” which tells you to push only as far as you “want to push it, as opposed to how far you should or could push,” and this is exactly what I’m talking about that can cause many athletes to test poorly with Ramp Tests. It’s hard to push yourself to your limit when every part of you wants to stop pedaling.

There are inherent difficulties with any type of fitness assessment, such as Ramp Tests and 8/20 minute tests, which can be extremely taxing and create slightly inaccurate results, and this is exactly why we created AIFTP Detection. Our goal here is to analyze your workout data and serve you the workouts that we think you can do based on that.

Of course, feeding the software with accurate information through post-workout surveys helps drive these decisions, so being honest during those times is crucial. Again, we want you to think of this tool as you would a coach. They crunch the numbers and write your workouts, but you need to tell them how you’re feeling after each workout and take the liberty to skip or switch workouts using Workout Alternates when you feel the need.

Also, don’t forget: As always, your time off the bike is equally important as your time in the saddle. Rest, recovery, and nutrition are key to feeling ready for your next workout and consistently producing physiological adaptations. :muscle:

My experience mirrors yours in that AIFTP sent me into a death spiral. As soon as I took a ramp test, I dropped my FTP with my corresponding ramp test result and now my training actually feels doable and in line with my capabilities again


@hoge and @3dflyer87 you guys don’t by any chance have a written record of what your AI FTP was set at when you decided to do your “first” ramp test to adjust it, and what it was adjusted to?

I specify written record, as wild guesstimates from past would be less helpful! :slight_smile: Maybe you could find it in your FTP History on TR ? I was manually recording mine for about 3 months before I found where that was!! LOL

Mine was also steadily moving up. At first, everything felt great. Fri Jun 30 208 → 215. Yaaaaaaaaaaay!! Go me!! TdF in no time, baby!!

Terrible, terrible two weeks. RPE on every WO was 11/10, everything VH or AO = Technical DNF = reduced difficulty % or bailed entirely.

I used Smith -2 to dial it in. Manual 215 → 210, much better.

A event, life, and then B event lead to very little training for a few weeks. Got prompted to run AI, ran it just to see, recommended 207, happily accepted. Nearly irrelevant difference, but ya, it’s prob / def come down a bit.

After lots of learning, which I shouldn’t have taken me this long [I’m pretty thick sometimes] I now know that WOs should be ranked Mod or Hard. One VH / wk is acceptable, but not idea. >1 VH, or even 1 AO, something is wrong, and you need to correct it. Wish I had known this when I started. Thought firmly that I should be “pushing”, and every WO should be VH, and that meant I was “really doing the most I could do”. :roll_eyes:

I’ve had the opposite experience. Ramp tests always over estimated, and I haven’t done one since AIFTP launched. Garmin and TR estimates are usually within a couple % of each other if I’ve done a hard ride outside/max effort (rare) for Garmin to look at.

1 Like

Same here. Ramp tests used to derail my training once I was starting to peak. Previously if I’d just done a VO2 block I could really kick out a huge ramp test but then would start to fail threshold/SS with longer intervals.

Now, AI FTP seems to ramp my FTP at a more manageable rate, even keeping it the same at times or with only a 1-2 watt increase later in the season, which I’ve found really helps to consolidate where I’m at.

I guess the good thing is that all the options are still available, including other types of FTP testing if desired. What hasn’t changed is that it takes some time/experimentation/experience/reflection to find out what works best for an individual.


I did my first ramp of the year a in July and blew the roof off the ai-ftp estimates from the first half of the year. Now, I figured I would get a higher ftp doing the ramp, but still it was a surprise how low the ai-ftp estimate was. It’s interesting, if expected, that any 2 random users with completely opposite experiences. I will say I’m in more of a high-volume regime, whereas I’m guessing you and your kids are doing mid or low volume? Perhaps short hard stuff is given more weight than it should, and easy long stuff isn’t given enough? Hard to say… it’s all statistics, and there will always be people at the extremes.

Anyway, definitely planning to continue using the ramp test, I just won’t do as many of them as I used to.

1 Like

It’s worth mentioning the importance of training periodization here as well.

Rest and Recovery are absolutely critical to sustainably getting faster and not burning out. This is why, when following our plans, you’ll have a set number of weeks with some hard workouts followed by a week of recovery.

@hoge, after looking at your career, I’m seeing 19 weeks of hard work starting early in the year without a true recovery week and the only few days off being due to illness. This is essentially non-functional overreaching, which isn’t sustainable for anyone and is sure to lead to burnout. It looks like you tackled a ton of hard workouts and set all kinds of new personal records during this time, which is probably why your FTP jumped up so quickly for a short while around this time.

It looks like you took a considerable amount of time off and are now just getting back into training, which is great and I’m happy to see you using Plan Builder to set you up with a solid training schedule as well as to prepare you for your events next year! Following this plan closely and dedicating the prescribed recovery weeks to be exclusive to recovery should make for a much more sustainable plan.

Remember, stimulus + recovery = adaptations!

If you’re interested in reading a bit more about training periodization and supercompensation check out the links below!

@gcarver I wonder if this was partly because there was almost a six-month gap between your last FTP detection and the Ramp Test you totally destroyed in July. :thinking:

Either way, it looks like your detections have been pretty accurate since! :partying_face:


This must be very rider type specific experience and why some are happy while others are not. I remember testing late last year with a ramp test at 304 Watts. then I had a slow period of 3 months and when I was ready to resume more regular training at the beginning of January 2023, AI suggested that my FTP was at 301 level. There is no way that this was accurate as I was gassed at every exercise, failing multiple workouts and even not being able to keep prescribed watts during what was supposed to be recovery day at 200 watt average or just below. I stuck with the workouts for most of the spring muscling through them and not re-testing frequently. Later in early summer i felt much stronger on the bike and on the trainer and AI was still stuck on the same 301-302 range.

1 Like

I definitely don’t do enough all-out efforts. Ramp test is the main place I get one. Have been thinking I could/should do a bit more longform threshold rides where I just hold until it until I’m spent. Something like the kollie moore protocol. And it’d be a good hard-ride for the week to boot.

@eddiegrinwald pretty sure I ran the ai-ftp button a few times earlier in the year. It was just something trivial like 1 or 2 watts each time so I discarded it. The last time I discarded I do recall getting a feedback question on I why chose to ignore. Anyway, it’s all good. I’ve been taking the 1 or 2W bumps each period since July :woozy_face:

1 Like

I’m not sure how much I trust the AI FTP detection either.

I’ve done 11 workouts since August 1st. Two of them were 60 minutes and the rest 30 minutes. Before August 1st I was off the bike all of July and most of June. I did a handful of workouts in the late spring but that is it. Absolute no training of any kind (bike or otherwise) all fall and winter. I’m not in THAT great of shape in other words.

When I started out, I manually set my FTP to 225 based on past experience. Second week of August, after the 4th ride it raised it up to 231. Yesterday after a 30 minute ride it raised it up again to 234. Yesterday was my first ride in over two weeks!

All rides are selected using Train Now and usually it is the recommended ride that I pick. It generally is having me do Threshold workouts. Yesterday was Darwin -5 (2x10 minutes @ 95-99% w/3 minute rest between). Because they are 30 minute rides I’m able to complete but that is crazy intensity to come to after being off for over two weeks.

Aren’t these AI FTP issues really problems with Survey responses rather than AI FTP? If you “fail” a workout and rate it as such and as too intense it’ll give you a lower level workout next - if iirc if you do that for a low level workout it triggers an AI FTP reduction?

Similarly, if you’re passing workouts above a certain level, it triggers an AI FTP increase?

I have “failed” workouts and got a reduction in future workouts, (and rated moderate and got an increase), but not enough to trigger AI FTP so I may be misremembering the last update.


It looks like you hit your peak FTP last July when you were in a Specialty phase. Nice work on that!

Between then and your 301 AIFTP Detection in October which was a 5.3% drop, you still averaged over 7.5 hours a week on the bike. It looks like most of it was soul riding without as much structure, but there were definitely some hard efforts mixed in, as well as some racing (nice job at Race the Lake, by the way. I’ve always wanted to do that ride!) which is probably why your FTP stayed pretty consistent fluctuating only a watt or two either way after that initial drop.

It does look like you struggled with some Threshold workouts, in particular after coming back into structured training, which Adaptive Training continued to bump down in intensity for you, but you did crush a lot of VO2 and Sweet Spot workouts, which is great. The endurance workout you mentioned struggling with was before the big 5% drop in your FTP, but regardless, I wonder if there was something bigger going on around this time, as your FTP history shows that 200 watts shouldn’t have ever been much of a chore for you on a normal day.

I’d also like to mention that one of our recent AI FTP Detection updates analyzes each workout to ensure that there aren’t any unique situations inside the 28 day FTP Detection window that would suggest a change in FTP. This helps get athletes locked into the right FTP and ultimately most fitting workout difficulty even more quickly.

It looks like your riding since that block has been unstructured for the most part, and there’s a big chunk without any power data, but as your name suggests, you do ride a lot! I’d be curious to see what sticking closer to the low-volume plan that you’ve recently been on would do for you. I’m seeing a good handful of weeks where you’ve ridden every day, and I wonder what a little more recovery would bring into the equation. It’s clear that you can be really consistent!

1 Like

Hey @jjb5489!

I took a look at your Calendar, and AI FTP Detection is working as intended :slightly_smiling_face:

Your FTP is relative to where it was back when you were training consistently. One thing our data shows is athletes are able to increase their FTP relatively quickly if they are really low to where their recent peak has been, especially after that nice little block of training you did in August.

While you are not following a training plan, you are, however, successfully completing the recommended TrainNow workouts like Darwin -5, and AI FTP is taking that into account. This workout is meant to feel hard, so I am not surprised you rated it as such. Good job on crushing it after taking some time off!

I also see a pattern of consistently completing 30-minute workouts, and I am assuming this is due to time constraints. But just keep in mind that, unlike one-hour workouts, you are crunching in the benefits in 30 minutes, and the effort, while shorter in time, will still feel hard, if not harder.

I know you mention you usually pick what TrainNow recommends, but as we recommend our athletes on training plans to pick a Workout Alternate if the training for the day seems like a bit much, I’d like to encourage you to pick a workout from the Workout Library if the recommended workouts in TrainNow are not in line with how you feel. At the end of the day, you know yourself best. :slight_smile:

1 Like

@Macy, that’s correct, just because you failed a workout on a given day it doesn’t mean your FTP has decreased and likewise the other way around. What does get adjusted are subsequent workouts to keep your training on track.

Survey Responses

Adaptive Training takes your survey response into account when recommending adaptations but they won’t affect your Progression Levels. The survey responses are appropriately weighted against many other factors, including more quantitative data from your recent workouts.

How you respond to your post-workout surveys will not make or break your training. Just answer honestly in the moment, and the surveys will help fine-tune your training recommendations. If you forget to respond to a survey, Adaptive Training will look at all of the other factors that go into determining your Progression Levels, so you’ll stay on track.

AI FTP Detection

We used Machine Learning (a subset of Artificial Intelligence) to analyze millions of rides, to understand how FTP changes over time and in response to training. Our system looks at your personal biometrics and the unique details of your recent and past cycling data to detect your FTP without the need for maximal efforts or specific workouts.

Now, During the Specialty phase**, AI FTP Detection will stop prompting you to update your FTP** so that Adaptive Training can guide you to peak fitness for your priority event or goal. By not updating your FTP during the Specialty phase, your training plan can focus on fine-tuning your fitness through your Progression Levels.

However, you will get prompted FTP adjustments during the Specialty Phase if one of your Progression Levels surpasses level nine or one of your Progression Levels is lowered below level two. If a Progression Level increases beyond a level nine or lowers below a level two during Specialty, your FTP will be adjusted by roughly 2%.

1 Like

My numbers are well-aligned all around

AIFTP: 284
eFTP (intervals): 287
mFTP WKO5: 285
Real life test 1 hour: 288


@eddiegrinwald Thanks. I’ll stick with the AI recommendations then and will start the short volume plan again.

1 Like

What are your “eFTP Min Duration” settings in intervals.icu?


I’m Diesel Engine, (time-thrialer according to WKO5), something like all-rounder, so I don’t have good short efforts, but above average in long ones. In this scenario 600s looks more suitable for my characteristics.

Thank you for looking into that. I thought I remember following the plan as prescribed but who knows?…
I did take some “time off” the trainer and am just now getting back on it. I love the results, but I would be lying if I said I wasnt a little scared.
I do plan to take it slow and incorporate more outdoor workouts…only going at the speed of fun. I hope this is enough to get some long term sustainable results.