Thank you guys for the response, very appreciate!
What would you suggest? I do think that Antelope -2 is slightly on the easier side than what it should be. Would you swap it or not - even if swapping does negatively affect prediction?
It is not clear to me
Isn’t this the heart of the problem though? If you feel your AIFTP is too low, you should do harder workouts to teach it, but if you do harder workouts, it could LOWER your AIFTP. There’s no way to win in that scenario.
I think you might be confusing FTP Predictions and FTP Detections.
If you put a workout on your calendar that we think you’ll fail, your upcoming training and your prediction will reflect that.
If, however, you pass the workout and the RPE was appropriate, we’ve just gained a bunch of really useful data that changes how we see your fitness, and your subsequent training (and likely your prediction) will change again to respect that.
Let me know if this makes sense.
I’m in my last block before Breck Epic and I’m so tempted to update my FTP to what Garmin has for a week to see if I can do the threshold workouts at it. I’ve been trusting TR to feed me accurate workouts, I feel everything is working, I feel pretty fit, but just have a thought of ‘what if?’. Then again, I have yet to really do a 20-min full out of body effort to know what that number should be.
In the last two or three blocks, I would do my hard intervals a touch above the TR ftp/goal for them to get a taste. Never failed. The more I think about this, the more I think it’s a bad idea. I’m going to leave as is. Maybe do my next threshold workout at a higher number than what’s being fed just to see.
I completed the recommended “Form”.
No worries. I don’t agree, but I’ve given up on it anyway so I’ll bow out. I’ll let people using the system be the focus. Cheers.
OK, this is now completely broken.
I had an AI FTP detection that I had been ignoring for a while, and I thought I’d let it run the detection and see what it gave me, hoping maybe it would reset the current foolishness, and the FTP numbers would get closer to reality. I’ve been running with my FTP at 314 and the workouts have been OK if a bit on the easy side. Never too hard, and always rated honestly (mostly moderate, with a couple of hard). With all this, the robot thinks my FTP is 302! I told it where to stick that number, and declined the update.
After that, it dropped my prediction down to 304.
So now I’m thinking about doing a ramp test, just to get a number that is approaching accurate (inasmuch as a ramp test can be accurate), both for my own knowledge, and to see if it will force the robot back to reality.
I’m longing for the days when I’d semi-regularly get a workout that i wasn’t sure I could do, and then either succeed or fail, but either way gain a solid understanding of where my fitness is relative to that workout type. Now it never gives me one I have any doubt about, and i think my training is not progressing as it could be. There’s never an opportunity to stretch.
@eddie Would doing a ramp test have any positive effect on this situation?
-Tim
What happens when you choose a harder alternate?
…and are you progressing? You declined a lower FTP, and you feel like the workouts it was giving you at that FTP were relatively easy, so it seems like you think your FTP should probably be higher than the 314?
You could test if you’d like, but using Workout Alternates and/or manually bumping your FTP up a couple of watts could be a decent alternative if you don’t want to mix a taxing test into your training.
How often are you doing structured workouts? Are you filling out post-ride surveys for all of your rides, including unstructured ones?
@Helvellyn The harder choices the AI gives me are almost always way too hard. Like a full PL or more, and almost always “not recommended.” On occasion I’ve gone through the library manually and found a suitable substitute (roughly 0.5 - 1.0 PL harder, same type of intervals) and completed them as expected - the higher end of moderate to hard.
@MikeMunson I’ve been doing this long enough (15-ish years), and am old enough (57) that progression is challenging to discern in the space of one or 2 blocks. I figure my FTP is somewhere between 314 and 320, but definitely higher than the 302 it thinks I’m at, and nowhere near the 330 it was predicting a while ago. That said, I’m pretty sure I could be progressing faster.
@eddie I’m doing structured workouts as directed by the plan - typically 1 VO2, 1 Threshold each week, plus 2 easy days, and an endurance day. Those are planned, but unstructured, but I’m pretty good about sticking to the intended effort levels, with deviations for unavoidable terrain requirements. I also have a Monday group ride which varies from mildly challenging to drooling on my stem for 90 minutes. In the event of stem drooling, I usually either skip or take Tuesday’s easy day really easy. Prior to regenerating my plan about a month ago, I was doing nearly everything as TR workouts, and prior to the weather getting nice, I was doing everything inside. I always ride with a power meter and HRM. The robot should be drowning in good data.
As for workout alternates, I touched on that above. Generally the AI suggestions are useless. Below is what it gives me for alternates to Whiteleaf (scheduled for next Sat). Whiteleaf +1 is barely harder (3 TSS and 0.01 IF) and Lion Rock isn’t quite the same, but would be harder because of the longer intervals. However, ALL of its suggestions are “Not Recommended.” If they’re really not recommended, why is it even suggesting these? And why are they not recommended? The first two are only marginally more difficult than Whiteleaf, and while the others are more challenging, I doubt I’d fail any of them, though Villebois might be tough. It seems the threshold for “Not Recommended” is a very low % chance to fail, as whiteleaf +1 has only a 3.3% chance. That’s insignificant. Villebois has 11.8% - that’s worth a “Not Recommended” label.
I took a look at the workout library, and there are 21 hard start threshold workouts between 2.3 and 3.2, inclusive. I would think it would pick from those.
Oh, and I just remembered that I have my plan set to ‘demanding.’ It doesn’t feel demanding.
I’ll probably bump my FTP a touch and see what happens. One thing I would like to be able to do is block the AI from changing my future workouts without my express consent, and a preview of what it’s going to do before I give it the OK. Seeing the twinkling AI stars on every future workout after completing every workout is irritating at the moment. Sometimes nothing changes, sometimes it does, and sometimes I can’t tell because I don’t remember what it started as. And the FTP prediction/detection debacle does not inspire confidence at this point.
Sorry for the rant.
-Tim.
This seems like a decent place to ask this. Today, my detected FTP was 269W. This ride is basically the equivalent of a 1-hour test. That would put me at at least 290W. I feel like I could have held that effort for at least 5-10 more minutes, so I’d say likely even that is slightly low. Also, I feel like this makes 269W around 10% low, which I think is probably significant enough throw the zones off, particularly on threshold and VO2 workouts. So… should I manually set it to 290W? Do a ramp test? Other options? I also bumped up the workouts on the entire last block by 10%, so effectively I was doing the workouts at a 256*1.1 = 282W FTP, based on the RPE being too easy. I suppose continuing to do that is an option.
I think that effort is a better representation of your fitness than any other test will be.
I’d maybe try setting your FTP to ~290 to see how things feel. 
I get it, but are you sure? Seems reasonable to expect that the assumption would be you’ll pass workouts you’re adding to your calendar. Not passing a ‘Not Recommended’ workout would seem to be a neutral signal.
If you put a workout on your calendar that we think you’re not going to pass, we’re going to stick to that assumption until you complete the workout or we have data that tells us otherwise.
Yeah, if you fail a workout that we expected you to fail, you’re right, that we probably had a path that was close to accurate early on.
I understand what you’re saying, but it’s not entirely consistent with some other posts either. It seems the only band-aid right now is to choose a harder workout, but that could also produce the same effect. Essentially, the user base is guessing as to what workout should be done which isn’t super. I understand you guys are working on this, but predicted and estimated ftps (mine seem to eventually match) are being degraded at a pretty quick rate. My PFTP dropped from 307 to 265 after today’s successful and appropriately rated outdoor workout. @Jonathan mentioned a potential fix pathway a few weeks ago. Are there any plans to roll this out?
That seems like a big drop. What does your training look like between now and that prediction? 
