I updated my body weight (increased) in the app and the AI FTP prediction went down a watt. Then as a test I decreased my weight and AI FTP recalculated and went back up a watt. I know it’s just a watt but why would it adjust.
Does anyone know how TR uses your body weight? I always thought it was just to calculate the w/kg.
I updated my weight, down 1.5 lbs, and AIFTP recalculated but ended up with the same number. Not sure what’s going on here as I don’t recall TR mentioning weight and/or w/kg being a factor in AIFTP.
I just checked again and it does the same. This time I change my weight to 100 lbs and AI gave me 244 watts. At 192 lbs I get 143 watts and 196 lbs I get 142 watts. Maybe I should use my AI predicted weight of 100 lbs going forward.
Yea totally get it’s just 1W. But would be good for all to know from TR if weight is a factor or not.
A watt loss here and there starts to add up.
@eddie - is this a bug?
I don’t know if they do include user weight in their model, but it would make sense that they did. a 400W FTP for someone who weighs 100kg might mean they still have a bit of room for improvement, but if someone weighs 60kg with a 400W FTP, then they have reached their genetic limit for sure.
That’s one part, but changes in weight - especially quick and significant ones - could also hint at fueling issues or increased fatigue, which in turn impact FTP.
As long as it’s just 1 or 2 Watts, it might just as well be a rounding error in a formula somewhere and very insignificant in the real world and long term.
One watt could just be a 0.1 watt rounding difference.
If this behavior is a byproduct of AI, that often means there is no “formula” or algorithm in play. There doesn’t have to be a “why” behind it (at least not programed into the system), it can just be a reflection of what the data/pattern predicts. We don’t know the inner workings of the TR system and all we can do is speculate, but there is often no explanation for the outcomes in AI-driven systems unless that part is specifically developed and exposed (which can be much harder than simply predicting the outcome). The change in weight might have associated that user with a different pattern that predicts a slightly different outcome, no logic/reason provided.
Nope. No bug here.
Weight is one of the many things fed into the model.
I can’t really say more about how that affects things, but keeping your weight accurate and up to date will likely help the model become even more accurate. 
Eddie - thanks for clarifying. I don’t think weight accuracy is mentioned in the new TR AI rollout or any other posts here. Might be good to let the masses know.
That being the case, do y’all have any plans to sync weight from Garmin Connect, etc.? I update it in Garmin pretty much every day but I’m too lazy to update it manually elsewhere; if it synced automatically it’d be sweet.
Not that I know of, but I’ve just shared that suggestion with the team.
I use another platform that pulls my weight data from Apple, so maybe we could start pulling health data from others as well…
Agree. If its important being able to have my scale automatically update it would be great.
Maybe TR should not link weight to AI. I don’t see anything that AI is reading that can determine what type of weight I’m gaining/ losing and where. I could have gain 5 lbs of leg muscle or 5 lbs of fat in my gut and AI won’t know.
…because it’s a highly nonlinear “black box” and we don’t really know 
I say this, jokingly, as someone who works day to day building ML models