Progression Levels, whether or not you use AT, are a huge leap forward for TR. Effectively, they link your power curve to their workouts which allows an athlete to strategize training around the needs of their events or weaknesses. The more I think about this, the more I come to the conclusion that beyond a baseline assessment at the beginning of training (or after a protracted period of detraining) and 90 days+ of good training data, it actually would be counterproductive to continue to do assessments at all.
This will, of course, require TR to model in something similar to mFTP, but it would likely have far more efficacy and accuracy than an FTP assessment (this assumes that a rider trains regularly and with some semblance of structure. After all, why would you test if you weren’t doing so). The ramp test, 20 minute test, and 8 minute tests all present the same vulnerability, they are built around a standard distribution around the mean surrounding power curves, but in reality, a large subset of riders do not conform to a “standard” power profile rendering the tests pretty unreliable for a substantial number of athletes.
All three tests are susceptible, to varying degrees, to overrepresenting FTP for people with a very anaerobic dominant power curve. They simply don’t exhaust the anaerobic capacity of those riders, whose aerobic capacity could be quite weak making workouts in zones 3/4 mismatched to their desired benefit. Conversely, a rider (often smaller riders) with a steep fall-off at the anaerobic end of their power duration curve will test low and won’t be presented with workouts that will generate the stimulus to push their aerobic threshold higher.
There is enough anecdotal evidence in this forum alone to indicate that people who don’t fall within the “normal” distribution routinely fail assessments or fail workouts soon after an increase. They are typically met with a blanket HTFU or “let go of your ego” response which is at best, short-sighted. Even back to the original Training and Racing With a Power Meter, Allen and Coggan laid out different test durations based on phenotypes. Modern modeling has allowed us access to the power duration curve so defaulting to the “old standard” 20-minute test was satisfactory. Training experts continued to peel back that onion with the understanding that athletes, especially newer athletes, found pacing and sustained efforts difficult and we ended it with abridged versions that were more “achievable” (i.e. Ramp and 8 minute) but perhaps drops off athletes at the tails of the bell curve phenotypically with some regularity.
This isn’t an “FTP is dead” argument, but rather the way we establish it as a one-size-fits-all approach probably should be. I’m not entirely convinced that a 4DP test is a superior replacement either. My thought is with a well-structured plan with variation commensurate with establishing a well-rounded power curve, “estimated” FTP could supersede assessments pretty seamlessly. The limitation is clearly surrounding those training without diversity in their training, riders with extended periods without training, and perhaps lower volume athletes (I think this circles back to variation and populating the curve). If FTP tests aren’t eliminated, at minimum, I think there is a potential for right-sizing an assessment protocol against progression levels, an undertaking that I won’t understate the statistical modeling involved in creating those variants.
I suspect FTP estimation is the vision of the TR team and hinted at in the podcasts to be sure. I am curious as to how TR gets around the problem of not having enough historical rider data to populate that model or control for the period of time after which the tests can be dropped. We see this now in beta. Many of us have seen our introductory levels set at a baseline of 1. We have had to manipulate and cherry-pick workouts to dial in those levels, and that simply takes time. Not having an FTP appraisal during the “learning” period wouldn’t give AT or Train Now much to work with. This would definitely be a deal-breaker for athletes early in their training careers. Those of us who have trained for some time can probably guess within 10 watts what our FTP is and might have a better sense check on this.
Curious about people’s thoughts here. Clearly, my premise has vulnerabilities of its own. Do you think estimated FTP with solid progression levels is satisfactory for training? Would you rather leverage that training day to testing after each meso-cycle? Who is already doing this with WKO and what are your intuitions about your mFTP values?
Edit: I will no longer be responding to this thread 6/30/21
In the TR AT universe you can do away with FTP altogether. Every PL has an equivalent at a different FTP, so the predictive power of FTP no longer has any meaning. Whether or not you are a ‘well-rounded’ cyclist is irrelevant. I love a d*ck measuring contest though.
Regarding the starting point, a questionnaire regarding age, sex, weight and fitness level would give enough accuracy to start a beginner on their power training journey. The effort survey should do the rest, even for experienced riders once the data has been collected and incorporated.
I find myself in agreement with a lot of things in your post, and certainly believe that Adaptive Training is a big step forward in dealing with the fact that not everyone’s power-duration curve matches the standard shape.
However, I’m not sure that AT does away with the need for FTP testing. There is only so far AT can go in adjusting things if the athlete’s FTP setting is way off. AT can make intervals longer, recoveries shorter, and some small changes in intensity, but if the FTP is way off, then the athlete can end up doing longer harder efforts in completely the wrong energy system. AT can do the fine adjustment, but for that to work, you need to be in at least the right ballpark through gross adjustment with FTP testing.
An athlete may be riding within their limits, finishing all their workouts, and modelled FTP may end up underestimating, because they are always riding conservatively (perhaps because that is what they believe their limit to be). At some point, only testing (such as a ramp test) will reveal that their true limit was further out than both the model and the athlete believed.
But I agree with your fundamental point that with AT, frequent testing is probably much less important than traditionally, especially with more experienced who are likely to be making only incremental gains. For athletes newer to structured training, the gains they could be making might blow their established training zones out of the water in just a few weeks.
Any sort of modeling I’ve seen falls down without max efforts at various points along the power curve, over rolling periods of time. Erg?
A lot of people just want to simply follow a plan. And not overtrain. AT on paper appears to be a good fit for that scenario.
If your FTP is way off, your either going to have very inadequate time in zone or be working the wrong zone entirely.
“Who needs FTP assessment?? I haven’t assessed in years, but all I have to do is find a 1hr sweet spot workout at a level of 25 and I’ll be good.”
Totally agree. I’ve already ditched the thought of future testing and plan on simply adding x number of watts to my FTP when progression levels are high enough to justify.
I tend to agree but I think the same argument has been made a number of times on this forum in various guises irrespective of AT. It always boils down to: If you train consistently and hard, you already know more about where you are than any FTP test can tell you.
*coughcough Xert *coughcough
Does exactly that (scales workouts to your power curve and even recovery time! this is even superior to 4DP which basically is just fine tuning your vo2max targets) and with the “no decay-training load matched” you theoretically don’t even need max efforts (surprisingly accurate, I retested last week the first time since January and it was only 8 watts off)
In my opinion (!) Xert’s algorithm (yes, I know no machine learning, but at least I know roughly what it does) and TSS tracking is way superior to the current state of AT (I mean you can select fitting intensity and duration since forever which is now a huge thing in TR :-P)
BUT the idea of using a long term plan and the huge workout library are way superior in TR …
I generally agree. I have a very good subjective view of my fitness, but nothing can replace objective. I’ve gone long periods without assessing, and it’s easy to lose track, get a bit lazy and avoid having to truly dig deep. An objective assessment can snap you back to reality. In training, you should rarely push yourself to the absolute max, and that leaves us pretending that we know what that absolute max is.
You only have to be a little bit wrong (4%-6%) to be working an entirely different zone than you are targeting.
I’m really looking forward to AT predicting FTP, but it will still largely rely on our very granular (1-5) and completely subjective assessments of how hard a workout was. What’s “hard” to one user is “very hard” to another.
Lighten up a bit! I was just using an extreme example. One of the big advantages of AT is that it can take the guessing out of training. But if your FTP number isn’t very accurate, your target zones and entire training plan are off. Someone who can barely complete a 1.5 level threshold workout is not working the same zone as someone who can complete a 9.0 level threshold workout, but they both think they’re doing threshold.
“Use AT to take the guesswork out of training. All you have to do is guess your FTP correctly!”
I think the combination of of objective assessment and AT is crucial.
AT might eventually mean no need to test FTP to set training sessions and intensities, but surely if you’re serious about tracking progress then you need objective and comparable testing of some sort. Doesn’t always need to be an FTP test but thats probably one of the most significant markers for many/most cycling disciplines.
I’d argue that it’s also likely to under-/overasses by 5% on a particular good/Bad day not to mention that your ftp easily varies +/- 5-10 watts on any given day
I’d even argue that I don’t need to know my absolute max if I closely observe how any given intensity feels.
I might not be the most experienced cyclist (in fact even kind of unexperienced with just 3 years under my belt), but I am able to tell if I’m working low endurance, high endurance, tempo, low threshold, high threshold, vo2max even without looking at any numbers.
The more I train and read I feel that it might actually be true that the key is to “learn your body” and trust yourself and not some numbers.
Thats not to be used as an excuse to not test every so often though
In my inexperienced view, the point of FTP testing is to allow inexperienced people to figure out their baseline without resorting to RPE (which isn’t useful until you have a body of experience to draw upon).
If AT is an RPE-based system – i.e. it adjusts your performance levels based on how the workout ‘feels’ after you finish – inexperience will tend to skew the results.
With an FTP test – especially the ramp test – there’s no doubt whether I’ve done a true max effort. Personally, I can’t say the same about some 20-minute interval session that I’ve only completed perhaps a handful of times.
(I’m not in adaptive training so I likely have it all wrong…)
Back to my original post, I clearly stated that this isn’t intended to disband FTP. This uses historical data and actual performance to build it. Instead of hoping someone is primed and tests well one day every mesocycle, your power duration curve takes your last 90 days of trend data and can build out a solid representation of what your LT actually is by using an inflection point and time to exhaustion. 90 days of good data can get far closer than testing can for the reasons I listed in the original post. The system only needs to have a variety of touchpoints on different energy systems and works in some decay and you will have everything you need. When you perform certain workouts, you are informally “testing.” This is obviously not going to work for someone who commutes or noodles around town and throws in a high intensity workout 1-2 days a week to prime the system, but that isn’t the core demographic for AT or really TR in general. If you feed the model data, it will give you a good appraisal of the aerobic contribution to your power curve. Data is based on both anaerobic and aerobic contributions, focusing on the aerobic to build your modeled FTP. If higher anaerobic contribution is present, it could lead to higher expectations. An FTP test is just an estimate. It doesn’t give you the depth of detail to provide a good benchmark for MLSS. It can get you close, but mapping anaerobic and aerobic power contribution can get you closer.
Does TR software look at the power duration curve at all?
Yes, it does seem like it would be ideal if TR was determining an mFTP/eFTP as you went along but it would also have to insert efforts to feed the model. I use WKO and I often see my mFTP decay even though I know actual FTP is higher.
My experience is Progression Levels are currently based on the highest Workout Level you’ve completed, not based on the actual power data of the work you’ve done. That may still be adequate, but I suspect that building something around the power curve would help them tremendously with the outdoor rides problems they are currently solving for. Now, whether the Workout levels are based on the hypothetical power duration curve, I don’t know, but I suspect there is some inconsistency in them as I think each system is on a slightly different scale and I think there are plenty of examples of easier workouts actually having more difficult scores than harder workouts. Either way, I find myself leaning towards fewer FTP tests and would be happy with a calculated FTP. I am currently more interested in progressing my levels more than progressing my FTP.
Yes, you’re absolutely right. That is why you would need a solid 90 days of data and workouts. TR would also need to use unstructured rides to build the power curve for FTP (not necessarily for workout levels specifically) structure is structure and a PR on a climb doesn’t mean repeatability for example. I think the levels matter less, than the curve on the back end feeding it. Admittedly, I am not entirely sure what that mathematical model is, I suspect some variation on coggans math model but it’s not accounting for all rides yet, it is focused solely on workout selection and less around aerobic vs anaerobic breakpoints. If you feed TRs AT model enough, it does create a good proxy though, your levels should line up around your phenotype pretty well This is the problem I alluded to in my premise, this model might prove inadequate for athletes who don’t have enough to “feed the beast.” I wonder if it could suggest an FTP and you can accept OR test. Might be a good compromise.
My opinion is that there aren’t enough athletes that make it a point to learn about training fundamentals, you don’t need to be a data scientist, but checking boxes on a plan misses a key element of understanding your body. A good coach will educate their athlete on these things, it doesn’t make them obsolete, it only gives them more value. However, I know it is extremely unlikely to get the watt bros or pay to play athletes to be attentive enough to their training to brake check along the way. Coach or ML this stuff isn’t a perfect science and the onus is on the athlete to reconcile their data with what they know about themselves as an athlete.
Going by forum comments, TR already has user compliance issues with power. I mean, some only have power on a trainer, or virtual power and a dumb trainer. And then there are calibration issues. Some use a bike power meter outside and a smart trainer inside. I would wonder what percentage of TR users use the same power meter indoors and out?