How important is heart rate to the TR AI?

Only semi-related, but has anyone has success using an Apple Watch as a cycling HRM via the HRM app? ‎HRM Heart Rate Monitor on the App Store

This app claims to connect your watch to your iPhone, then use the phone to broadcast HR data to a cycling computer via BLE. I was traveling recently and had a bike+computer but no HR strap. I tried using this app but was not able to get it to connect reliably. It’s a cool idea but seems perhaps underbaked or limited by BLE functionality at the moment.

1 Like

Nope. I tried a half dozen apps that claimed to do this and couldn’t get it to work. I bought a polar hrm and have been very happy with it.

1 Like

I would think going by the survey answers is too subjective and too granular for determining FTP with any degree of accuracy. And I get way more power fluctuations using my power pedals for power than if I use my smart trainer for power, irrespective of ftp. At best, either would be only very loosely correlated with FTP IMO.

I don’t follow your logic I’m afraid.

Ten workouts is how it gets enough power data and each subjective survey is how it knows how you felt each workout was.

I mean we don’t know how the tr magic box is determining this at all.

If I do 10 endurance workouts will it know my ftp with any accuracy?

I would hope it uses some subjective data to figure out the individuals strengths/weaknesses and fill out a power duration curve that the ai can use to approximate FTP.

Hopefully it has some logic to deal with the large variances in ability >FTP but we don’t know :man_shrugging:.

In erg mode, the power data would essentially be the same for those 10 workouts irrespective of how hard or easy they were for the rider, so no difference to enable detecting FTP. If I answer the the survey saying all were easy does that mean my FTP is way higher than if I answered that all were hard? That’s subjective and non quantitative.

Ah. No, that’s a flaw in your thinking, this is an erg mode workout:


And so is this.

Which one do you think I found harder?

1 Like

I’m particularly interested in this as I have a heart condition that makes my heart rate play up (seeing a cardiologist etc. before anyone freaks). I wonder if I should just not use a HR monitor cause it confuses the AI… Saying that when my heart races it makes the effort feel hard anyway so what’s showing up in HR is somewhat reflected in power anyway.

1 Like

Have you looked at https://crickles.casa/ ?

Oh no! That’s frustrating :cry:.

We do consider HR when we have it, but it’s not necessary for an accurate AI FTP detection or for adaptations. An accurate power meter is far more important and reliable. If you have access to it, it’s extra data which can only be a good thing! :wink:

2 Likes

We do consider HR when we have it, but it’s not necessary for an accurate AI FTP detection or for adaptations. An accurate power meter is far more important and reliable. If you have access to it, it’s extra data which can only be a good thing!

Nice, ok good to know. I figured it wasn’t necessary b/c I worked out without HR for a while before finding the broadcast feature on my watch, I just wanted to make sure I wasn’t throwing the algorithm off by not providing the data all of a sudden.

Yeah I’m just going to broadcast from my watch. I figure it’s not going to be as accurate as a strap connected to my torso, but it’s better than nothing and if it’s not critical to AIFTP or adaptive training (and if the less accurate data isn’t going to cause an issue) I think it will be enough.

Incidentally I took @gally24’s advice and emailed garmin to tell them mine is draining battery (I confirmed it by testing the voltage on a new battery before popping it in and after the computer stopped connecting 3 weeks later and the battery definitely drained to 2.4v in that time :expressionless: ), so hopefully they’ll send me that pro strap. If not :person_shrugging: I have the watch :slight_smile:

2 Likes

Do you see it in the heart rate data?

Might be worth looking at the individual data points than graphs