I think it must be using FTP to pick the workouts, since users report getting well tailored workouts with manually input FTPs (i.e. longer duration in a given zone at lower manual FTP). The AI can only pick workouts from the library and those workouts all scale off your FTP.
They’ve also said the goal is to start you at around Threshold level 3 to give you a good selection of workouts and progression options. So the AI may not know or care what your FTP is when prescribing workouts but there’s an algorithm in there somewhere that does! Maybe it’s reverse engineered i.e. the AI simply prescribes watts and then it has to figure out what your FTP would need to be for a level 3 Threshold workout to give you those watt targets.
I personally don’t care whether or not TR AI is using FTP or raw watts, progression is progression. But here’s where I’m genuinely confused… if the new AI doesn’t even know what FTP is (or however Nate worded it), then how is it selecting workouts when literally every interval in every workout is prescribed as a percentage of my FTP and not just raw watts?
For example, McGregor isn’t “3x20 minutes at 229 watts”, it’s “3x20 minutes at 88% FTP” and the wattage of those intervals is determined by… my FTP, that the system supposedly doesn’t know or care about?
My best hour outside 275w. TR estimate is 279w. I use 275w.
I would also suggest to use the TR FTP as intended as a training tool. In the past I have had slightly different numbers suggested as my FTP from AI, Garmin and Intervals. As well with enough time you get to know what you are capable of and can naturally feel where your threshold is although I have found this can also change as you get more trained holding it for longer durations.
Over the last two seasons with the prior model I have hit personal PBs across both sprint and endurance time frames and I am at an age where it’s more likely my best days are behind me.
I am going to give this new framework a shot. It looks like it helps with predicting yellow and red days as well as making sure that endurance rides don’t impact the hard days. That’s all a big win from my perspective.
I think one of the most illumanating things that Nate and Co have said is that from the models they have run they have found that progressively higher watts trumps lesser watts for progressively longer period for growing FTP. That’s quite insightful and something to consider as we all work to get faster and train for our specific events. (I am hoping that once I get to specialty phase the push to grow FTP will change and I’ll be more focused on growing time inzone for my targeted event)
If you consider training zones to be defined by percentages of FTP, then AIFTP would give AI zones which may or may not be what you’re used to, especially if you do any riding outside of TR
Its certainly confusing, try doing 4 hours spins in zone 2 with an additional 10% watts and see how you feel after a few months.
Its a value that may work within the confines of the TR app.
Are the % movements in the AI FTP even comparable with the ramp test methodology.
I found my old AI FTP to be accurate and usable outside the app, so a big loss to me and probably won’t subaribe outside the dark winter months.
Other apps like Join have their own internal fitness score and eftp both of which are rubbish, TR was the only auto FTP I trusted with no max efforts and that alone was somtimes worth a sub.
This is true short term. But what I’ve found is that you burn hot for about 6-8 weeks and then you are fighting to hold and slow the decline til the next season break. Nothing wrong with it but you need to time it well IMO. It will be interesting to see long term results in a few years to see if this bucks past experiences. I hope it does.
Agree, the proof is in the pudding long term. In theory the tools are there to help manage this especially if you are very consistent in your RPE. Consistency is key as we know and if this leads to burn out we’ll end up being slower. I remember the old days where there plans were static, now that definitely led to burnout if you didn’t know enough to adjust.
I think the key challenge is the temptation to gamify FTP gains but then that’s kind of always been there hasn’t it. What I have found empowering is doing less and still progressing. That’s been quite liberating.
They addressed this on the podcast, watt targets for dynamic endurance rides will adjust. E.g. I can see I’ve got a 3 hour ride coming up where it’s telling me to target 0.64 IF, the following week I’ve got a shorter 2 hour ride with a slightly higher 0.67 target. Zone 2 is plenty big enough to enable you to be able to stay in zone without having to ramp the watts up 10%. And the dynamic endurance rides are a huge improvement on the previous model where it just tried to progress your Endurance levels by increasing intensity until you were doing all your endurance right at the top of the zone.
This is my fear with the new system. I’m worried that the goal is to get as many of the people in the middle of the bell curve to just get quick FTP gains because that looks good for TR. As someone who trains more than average and wants to increase how long I can push hard, FTP gains are a goal but not THE goal.
Do you have a target event that you are training for or just trying to be as strong as you can for as long as you can? Unfortunately it will take time I reckon to see if the new systems does provide gains. I have been guilty in the past of doing too much (volume or intensity or both) but in the last year have really focused on keeping endurance low and my two interval sessions following what’s prescribed and as I said I have gotten faster across everything from 5s power to 5hr power.
I can’t really see how the new system (assuming it works as described and sounds like we have been using it behind the scenes or sometime now) will not enable all of us to be faster for longer and help with the specific requirements of your target race.
We have seen historically how these sorts of AI modesl have lead to innovative improvements in other fields (look up the discovery of Halicin or how New Zealand one the America’s cup a few years ago or how the chess AI taught humans moves they had never thought of). Time will tell if this system provides similar improvements in performance for the masses who can use it effectively.
the F in FTP stands for Functional as in it’s a Functional metric to base your training zones off of and progress over time. This is how I think about it in my mind. Could i hold it for an hour, maybe, do I care, no! I’d rather use this to base my training off to get faster.
After a short break at the end of summer I started back into training. I did a ramp test to see where I was at. I made decent gains over the next 3 to 4 months. I do 4 sessions per week.
The new AI dropped my FTP by 6% (back to where the ramp test was 3 months earlier). I have trained with the new AI for a week and despite being on the bike for ~7 hours a week it predicts in 4 weeks I will gain 1 FTP point. At this rate it will take 1 year to achieve what I thought the FTP was a little over a week ago.
I think disappointment is kinda relative.
Are the workouts you’re being served up easy or hard?
There’s a lot of discussion lately on both sides of the coin for this one. I was in the camp of “aiFTP is way too hard”, and decided to adjust down manually to what I know my FTP to be. The workouts are much better now.
I have traditionally struggled on the threshold rides TR recommends and more frequently than not I blow up. I think the AI is adjusting to a training schedule that is more reasonable… so, I do think it’s most likely more accurate.
However, what I don’t understand is why the ramp test taken 4 months ago was so inaccurate. I am certain I am stronger now than I was then. In fact, I feel I’m close to my peak from last summer though the FTP numbers show a very different story.
I do very well on ramp tests because I’m quite strong anaerobically, but then really struggle with the threshold workouts that follow.
Frankly, if you feel you’re at the point you were during your peak over summer, and you got there by using TR, then something is working. If the other zones feel good but threshold is tough, continue to rate them appropriately and I’m sure the workouts will converge to a good spot. You could also manually adjust the intensity of threshold to something a bit more moderate and see what the workouts look like.
There is no disputing TR works. (Before AI and with AI.) Ultimately, the FTP number does not matter. It’s how the training equates to performance in the race. Probably the FTP was wrong for years and it’s right now.
However, that is not to say that the new AI is not disappointing in my case. 1. I did have my heart set on a w/kg number that would be necessary to complete a bucket list race and I now realize that it will probably be impossible to achieve. (maybe better to know?) 2. At least when the FTP was not known for weeks I could hold on to a bit more optimism/motivation to keep pressing ahead.
Hopefully they will release some form of power curve estimation in the near future which should help with the whole ftp vs. ftp thing too!
