I endlessly tinker with the app, and find it wildly entertaining. When TR rolled out their update, AIFTP dropped me from 254 to 243, but my prediction had me around 270. I found that when I complete a workout outdoors, it recalculates all my future workouts, sometimes making them not recommended. I will play around with those workouts a bit and usually “find” the lost watts those Not Reccomended workouts create. It will be interesting to see what the system detects in 11 days. I’m stoked if I go from 243 > 270.
3 Likes
This is my superpower.
4 Likes
Do you have the workouts pinned? Otherwise it sounds like a bug.
The AI did this for me word for word. Blackcap vs Three Sisters.
I gave Blackcap a Very Hard rating due to the Overs being a little too hard where I thought I was going to fail.
AI came back with Three Sisters for the next workout which was a Hard rating. After that Cloudripper -2 another Hard rating.
2 Likes
My FTP is now 255 and at the beginning of the plan it had a prediction of 281, low volume base phase.
Whenever I rated the workout as per the TR estimation, the prediction stayed at the initial figure, maybe even budge upward slightly.
But after just 2 workouts I rated very hard (one of the being the final one in the block before the recovery week), it significantly decreased the prediction.
Now the prediction for tomorrow is 261. I feel a bit cheated 
but whatever, hope that it’s correct though.
1 Like
Being realistic though, 255 > 261 feels more beliveable.
For me 255 > 281 would be a very large jump suggesting inaccuracy elsewhere to begin with ?
1 Like
Absolutely it’s more believable and the 26w jump was way too big.
But why give such a high prediction to begin with?
And it continues after the FTP detection today as well, from 261 actual to 292 predicted. I’ve been training consistently with TR for 5 years and my highest was 280.
1 Like
I had something similar with the first AI prediction being 238 > 250, before it eventually settling with 245.
It then immediately suggested 255 in 28 days time, before re-evaluating down to 251 30min later without a single change being made.
Feels like the initial proposals are often over the top.
1 Like
I suspect the predictions will settle down once we’ve gone through the first cycle.
I ended up +5 and my new prediction is + 6.
I’m in the last week of a detection cycle.
Last night I was predicted 239->249 I think I started the cycle predicting 254.
I added in some outdoor endurance rides to the week and it elevated to 251.
I completed the ride, but outdoor execution in London is variable:
And now I’m down to 247!
Wtf?
That’s even less than not riding, when a solid hour added watts!
Ps also wanted to show off my fatigue mgmt 
——-edit
I’ve faffed about with all the rides to see if anything makes it better;
- Any increase in quality rides (extending, adding, converting from endurance) decreased predicted ftp
- Any change to the Sunday ride makes no difference at all
- Removing my Weds and Fri 1h outdoor endurance increased ftp
- Inserting Weds 30min Easy Solo increased ftp
- Inserting Fri 30min Easy Solo affected nothing. Longer was detrimental.
- Accepting fatigue management recommendations decreased predicted ftp
- Adding a 30min swim tonight affected nothing
So I’m back to a predicted 251.
3 Likes
That is to be expected. The initial prediction is if everything goes perfectly and is often a bit high.
1 Like