This topic is inspired a bit by the latest podcast. It seemed that TR is building all kinds of stuff to ‘omit’ facing reality and testing and/or going all out. Nate et all. talk about the ability to ‘express your FTP’ and not having to do that in a future feature that ‘estimates’ your FTP based on completing workouts or other types of efforts.
Sure this has its value I guess since probably a large base of TR users just want to ‘improve’ and lack the willingness to do for example a ramp test.
Some critical thinking:
Coming from a process background one of the main things I find is that improvement can only be measured against a standardized measurement protocol. If you remove this, what is your base measurement going to look like? And how do you keep your data quality high?
For me, I value the ramp test because I get a good sense of improvement if I continue to use the same measurement (ramp test), equipment (my trainer and the bike Power meter), and the situation (indoors, the same amount of cooling).
I used to use Xert which bases the FTP estimate on max efforts and matches this with a general power profile like Coggans power chart. I really hope we do not go that direction, because it motivated me to do max efforts all the time instead of investing in good training. I used to chase FTP improvements ‘on paper’.
The FTP that Xert gave me was an expression of my strongest point. In my case 5-20 sec sprint after some 5min anaerobic interval. I am very good at those. However, the FTP that came out of these estimates i could not hold for even 8 minutes… And most importantly it overestimated my 20min value so much that my power targets were just way too high and therefore training totally unproductive.
But a ramp test does something quite similar, by taking your best 1 minute power.
A ramp test over estimates my FTP. 17 watts difference between my Ramp Test result and longer threshold testing done 2 weeks ago.
If Trainineroad can estimate FTP from various duration of power (not just the VO2 and upwards end of it), i can only think of this as a good thing and get rid of the Ramp Test!!
The point of my argument is that you need to do the same test to tell the difference between points A and B. Whenever that is over or underestimating does not make a difference. You cannot compare your FTP to another person’s anyway because it says nothing about your ability to express it.
If you have a dataset that has a regular reference, it has good quality. If you remove the reference you don’t have a confidence measure anymore. I’m curious how TR is going to solve this. Of course, there are ways to do so, which I think will not be publicly shared. But it depends on enough people continue to keep testing and or not changing the workouts anymore.
You’re assuming people value these same things. For me, I just want fitness gains. The FTP number is meaningless to me. When I do race, it’s on a tri bike in a sprint distance race. Doesn’t matter what the power readings are, I’m going at whatever race pace will allow that day.
I know a lot of runners and swimmers. They don’t constantly assess fitness. They just increase pace targets or interval length in workouts. With TR, the system has to adjust an FTP number to keep your workouts in the closest to “correct” zones for you, but with adaptive training, it can have a fudge factor that seems to correct for any zone that you’re not within the bell curve.
Don’t get me wrong, I’m not disagreeing with what you’re saying just that not everyone cares what their test results are if their real world results don’t measure up or better yet exceed their expectations.
There are other ways to do the same process. FTP really needs max efforts from my perspective. I know what your saying about Xert as that did mess with my head. I also use Intervals.icu which does give an estimated FTP. For me it works as well as a ramp test. The two sets of data tend to be within a few watts of each other. They do have the data to extrapolate our changes to a power curve and get a new FTP.
I really cant see how there can be any meaningful ‘estimate’ of FTP (or any other measure of power) unless people do MAX efforts of some kind. If you dont have max effort data then you cant possibly model what someone is capable of.
The ramp test was supposed to be a proxy because people wouldnt do the 20 min test, which was a proxy for the 60 min test etc etc. Doesnt matter what type of max effort test you do, it still has to be done somewhere. Everything else is guesses based on estimates of random data.
I’m very much looking forward to this feature so long as:
I can “trigger” the FTP estimation whenever I want
You don’t have to be on a TR plan for this to work
I’m working with a coach, so I just use TR to control workouts. Right now to estimate FTP my coach has me doing max efforts - a 20 minutes plus shorter durations - about every 4 weeks. So interested in seeing how the TR estimate would along or not to the WKO5 model
I think you have some assumptions baked into your thoughts here that are coloring your conclusions in ways you may have missed.
The ramp test works.
This is a known false assumption. It works well for many (it works pretty well for me, maybe a bit high) but it flat out does not work for many others. High anaerobic capacity breaks it, outside stress breaks it, many other things break it. Its an estimate tool that works pretty well for the middle of a bell curve of people, not a test.
The ‘Same Situation’ exists.
You are getting a cold and will get sick tomorrow but dont know it yet. You slept 35 minutes less on average for the last 4 days. You had coffee/didnt. There are millions of things. I can take the same test a few days apart and get an 8% different result under ‘identical’ conditions.
You need to do a max effort to to compare data.
You just dont. If we know what your ‘all out’ looks like, what your ‘hard’ looks like etc. We know those things for a few million other sets of data. Many of those other sets of data also have real max hour efforts. You can combine them pretty well, its just hard. The ‘old’ way to do stuff like this was to make a model (like the coggans chart). The more modern (insert big data/ML/Buzz word here) way is not to look at a sample or average but look at ALL the data. Billions of data points and take your best guess from there. Use all that data to predict your ftp last month, use this months data to see if our prediction was right and tweak it. Use that best guess to predict forward, then use actual forward results to test the prediction and tweak it. Machine learning is a funny phrase cause it is totally not learning, it is looking at lots of data and getting lots and lots of wrong answers really really fast till you find the least wrong one. You dont need a ‘base measurement’ you use all the measurements. If doing x makes 90% of people faster that is great but a quality ML will also have data for the 10% and do something better for them as well once it has enough data and the logic is sound (its still early days here for TR, they are in the fun part now of troubleshooting a magic black box). A ramp test works for 30-60% and screws everyone else forever.
TR cares that the ftp they get is really your ftp.
They dance around this but I think you will be happier thinking of the number they give you as your functional training power. They dont need the number that you could in theory on a perfect day do for an hour. They need a number that sets workout levels. It would have been easier for them to call it the flux capacity setting and make it 0-10000 and avoid the confusion but marketing requires it be the ftp concept people know.
At best the ramp is one possible measure of fitness gain. However, it can easily miss fitness gain.
It’s only rock solid use is as an educated guess as a training zone estimator.
If you want to know if you are gaining fitness, then track your PL. it’s as simple as that. The best measure of what you can do is what you have done recently, and PLs (or other similar systems) give a simple quantifiable metric of that.
I don’t actually care whether my FTP is “accurate” in the sense that it’s within a Watt or so. I just need it to be accurate enough that my workouts are doable, target the right zones and I don’t burn out during the duration of the plan.
It’s really only around threshold that FTP makes any sense/difference - VO2max efforts are just going hard for the duration, not much more to them than that; Endurance/sub LT1 are just riding at an effort you can hold a conversation. Back to threshold, over-unders will tell you pretty quickly if your FTP (genuine/tested/guestimated/wet finger in the air) is right or not. But even here, 1 or 2% either way isn’t going to be noticeable, 5% more than likely is.
Also as above - rest of life can have a big effect on your “real time” FTP. You could quite easily have a 5% range from day to day depending on sleep deprivation, stress, illness, etc.
I’ll chuck it another thing I like about “auto” FTP estimation, whether it’s Xert or another system. I don’t want to be training based on an FTP that’s gone stale because I haven’t formally tested in 4-6 weeks. I don’t even want to test at all ideally, I just want to train.
The big part for me is the functional training protocol. If I can squat 315 for however many on a perfect day, rested, fed and tapered that means that is my max. I would NEVER use that max to base my % off of for day to day training. It would be to hard and I would make no progress. Training max is the accepted idea in strength training.
People seem to fight that idea HARD in cycling. (I know my numbers are low here so hopefully that wont just result in the ideas being discarded). I have my TR FTP set to 214. I get workouts that are hard and that I can progress on. Perfect. Last week I did a zwift race with a day off before it, no beers the night before and good diet. I averaged over my TR ftp for an hour and honestly if the group had gone faster I had quite a bit left in me. Garmin says my ftp is 232, which honestly I think is correct. Should I change my TR ftp to what all the data suggests is my real ftp? To my actual race day hour power? No, I 100% should not. My training quality would be worse, my progress worse, etc.
AT wants to help you set THAT number, not your ftp. It wants to find the level you should train around for that moment, not your magic self worth number.
Maybe I should have said “TR FTP” I certainly wasn’t meaning a lab tested physiologically accurate one.
I think testing (in the TR or similar training world) is very useful for those who are just starting out and get “beginner” gains so their training levels are constantly changing. But after a year/eighteen months or so those gains level off and you are generally pushing other parts of your performance envelope.
Thanks for the feedback. Just to clarify, my point is about TR making systems that avoid going ‘all out’. Not about my ability to test.
Don’t get me wrong, I like that they are building this. But my question is how is this going to help people get faster? The confidence or accuracy of the derived FTP of a ‘non-all-out’ effort is a lot lower. Or in other words, if you don’t go all out, the accuracy of your data is very low.
Example 1: I’m going all out, on a ramp test, or a race, or whatever. The estimated FTP is correct. My training is in going to be in (ballpark) the right zone with a relatively high accuracy.
Example 2: I’m never going all out, not on training (that hurts too much) and not in racing. My max efforts are 90% of my actual capability. Or 85%, or 95%? How can you tell?
I think this is not a true statement. What is all out? I did an ftp test and fell over dead and it said 205 but then the next week i did an hour long race for over my ‘ftp’ and had plenty left in the tank and would not have called it all out at all. Was the all out data accurate? Which was all out? Did my ftp go up 20 points in a week?
I totally get what you are saying, but if the goal is success and not optimal progress it still works. If I have a few hundred million rides form people who go hard and easy (most) and I have data from someone who never goes past ‘hard’ into all out or very hard but mark the levels of effort has easy, moderate or hard I can make some pretty good guesses what the harder data might look like. It will be wrong because going hard is a skill in and of itself but they will be directionaly correct in helping them pick a number to pick their next workouts. The real kicker here is I can measure to see if their moderate efforts got better based on what I told them, times many many thousands.
If i have data, action, feedback I can change action until feedback moves in the right direction. I can back test it a hundred million times. The quality of the data does not matter when you have enough of it.
Will that ever work as well for people who push less hard? No, it cant, but then again the act of pushing hard is half the battle so why would you expect it to.
I think the best thing TR could do is get rid of the most important thing to many cyclists who use the software. The all important FTP number that allows some to base their whole self worth as a cycling upon.
The whole point of the system is to get correct training targets to drive adaptation as quickly as possible. If they can do that, they’ll have an edge on the rest of the market place.
I personally wonder if the whole notion of needing to do max efforts is even relevant if you were to have the huge database that TR has. If the team were to develop a suitable way to survey the results of a workout they should be able to ‘goal seek’ the individuals training requirements. If the rider is able to do harder and harder workouts, eventually the software should just increase the power targets.
I certainly wasn’t meaning a lab tested physiologically accurate one.