I’m just coming to the end of the build phase of the low volume century plan and am thinking about doing a big ride at the end of the rest week (168km, 3350m climbing). I’m guessing this might look like overtraining to the AI and so it might reduce my estimated FTP. On the other hand maybe I’ll get bonus credit for the longer ride. I find seeing my estimated ftp go up motivating so it would be nice to get a bigger number when i press the button (although i realise its only an estimate, fitness goes beyond just your ftp, and there will be error in the estimate, etc). Do you think it would it be better to do “the test” (press the ftp detect button) before i do the ride or afterwards on the tuesday when it’s scheduled? Would be interested in people’s thoughts
Best to contact email@example.com related to this question. We don’t have enough info about the hidden innerworkings of the AI FTP Detection to make anything more than wild guesses or possible anecdotes of questionable relevance to your specific state.
My thought is that if it really is AI, then the more data you give it the better. It should know if you are “overtrained” or when you bonk, otherwise it’s a lousy estimator of fitness
Thanks- had thought of support as only being there for IT type issues. Will get in touch with them and see what they say
Sounds like the next big feature request: A sneak preview button showing the new FTP estimate before you lock it in, and a secondary AI model that estimates the optimal moment to lock in the new FTP estimate maximizing the FTP increase!
Nate has definitely spoken on the podcast about hopes to have the ai automatically detect when your ftp had gone up without the need for scheduling a test and had also said they hope to take your goals into account when deciding when best to do this (eg potentially keeping ftp the same so that you can move higher up threshold progression levels and start doing longer steady state work if you have a time trial coming up)
Sounds like a bad idea to me. The point isn’t to get the highest FTP possible and would definitely lead to people trying to game the model to artificially inflate it.
Definitely depends on which AI algorithm is used but if I were a bettin man (and I am) I would bet that they account for outliers that are outside of confidence intervals. This is to say, one anomalous result does not skew your overall trend.
Think of it like a walking path, if you take the same footpath over and over it becomes worn in and one day of walking on the grass doesn’t make much of a difference.
Also keep in mind that the AIFTPD takes a minimum of 12 workouts/rides and apparently looks at a wide range of history if present (1-3 months IIRC), so it’s probably not likely that a single workout/event will skew the detection in a significant way.
I’m almost certain he was joking
I’m a betting man too & you’d have taken my money!
Support came back to me to say it shouldn’t matter whether the test is done before or after the ride
Funny thing happened to me this week. I am on a plan and I was suggested a Ramp test yesterday (Sunday) which is strange given that mine always land on Tuesdays. Another strange thing was that I am smack in the middle of a block. Some context, I am doing a MTB series on alternating weeks. I hot my workouts on “Series Off” weeks and on the “On” weeks I have been doing the V02 work/ Race/ Long Rides instead of the threshold workouts.
When I saw the test suggestion i thought for sure my FTP estimate was going to drop because I had skipped a couple threshold workouts but instead it went up! My ego was thrilled but my legs were concerned.
is AI just celebrating my efforts on my long rides/ races? The efforts HAVE been good, I just didnt think it was doing this yet.
Not sure it’s the issue in your case, but there have been quite a few people reporting Ramp Tests getting set at odd times. Those seem to be bugs from what I have seen.
I suggest contacting firstname.lastname@example.org anytime you get something that seems out of place like that. Might be intended, but if it’s a bug, they need to review to be sure.
awwww. Here I was popping champaign recovery drinks.
Could be legit. But TR generally rolls features like this out with at least some level of notification vs slipping in features unannounced. And considering the other issues I’ve seen, it’s certainly worth the time to email them.
I reached out and spoke to another AWESOME support staff. They mentioned that it could be a bug and would report to “engineering”. i havent figured out how to break it to my ego yet.
At the risk of guessing, I won’t be surprised if your FTP gain is “real”. I just think the automatic application of the Ramp Test mid plan is a mistake. So, don’t lose hope yet as I think it’s quite likely your gains are legit.