Honestly…. Worrying about a 4.6% FTP increase in 28days because it’s not a 5.6% increase is the biggest vindication of @LacticacidMCB ‘s earlier point I could imagine.
5 Likes
2 Likes
I think the underlying issue is that the forum is littered with people who don’t “trust” the AI and/or who hate the redefinition of “FTP”, so every little thing causes them to go “see, it was off in this example, so it’s probably wrong constantly!” It becomes really clear to me as soon as the post says something like “I performed the workout exactly as prescribed and followed the plan perfectly and the AI did X”. Then you look and the workout/plan was almost always NOT done “exactly”. It’s a never ending challenge to “prove the AI wrong” and I’m not sure I think it will ever go away for that personality type.
12 Likes
I don’t think you read my post correctly. It’s not about the changes in predicted ftp, it’s that the initial prediction always seems far to optimistic.
100%. In 5 years of trying, it’s the first time I’ve cracked 4 w/kg. I couldn’t be happier with how it’s gone.
This post was just a suggestion of how things could be improved a bit.
3 Likes
Exactly for this well known behaviour, I’d suggest that predicted ftp lean more towards conservative.
Yeah I’d have to concur with that… especially it seems for those on a “balanced” plan.
2 Likes
At first I had FTP Prediction turned off because I felt it was driving me crazy but when I had to move some days around because of life I started to see what a valuable tool it can be for arranging workouts.
I’ve done a few workouts where I’ve had to rate them Very Hard vs Hard what the Ai predict it would be, excepted the FTP prediction hit and moved on, only to get some of those lost predicted watts back at the end of the training block.
After knowing this I’m looking at the watt more than FTP Prediction and the tool for that has been helpful.
I can only imagine what will happen when I start riding outdoors as the weather gets better. That thing will drive you mad if you pay too much attention to it.
3 Likes
Not sure if this is how it works but in my TR AI experience it’s like this example:
- Let’s say FTP is set 280 and based on next 4 workouts it’s predicted 292.
- The RPE prediction for all workouts is like 60% hard 40% very hard.
- If I set RPE for all following workouts to Hard, prediction stays on track at 292.
- If I set RPE for all following workouts to Very Hard, prediction will decrease to something like 282.
- So 10 watt difference in prediction between rating all workouts Hard vs Very Hard.
Now here is my ideal solution @eddie
- But since the estimated prediction was 60% hard vs 40% very hard…my favored approach would be to reduce the “ideal” 10 watt gain of the above example by that 60/40 relation…so we have 10 x 0.6 = 6 watt.
- So show me 282 (see number 4) + 6 = 288 watt as estimated FTP.
- Then reduce by 1.5 watt if I rate a workout very hard (so after 4x very hard I’m at 282)
- Or add 1 watt if I rate hard (so after 4x hard I’m at 292).
- Seems more realistic and less fluctuating.
Hope this makes sense.
4 Likes
This is what I thought a first but I think there is something more.
These last three weeks I had The Owl +5 and The Priest +5 both which I rated Very Hard.
Now after The Owl +5 it dropped my FTP Prediction and gave me and easier Threshold workout, I think is was the Chimney but then after I did my Sweetspot workout, Tallac +1 a Hard rating, whatever data it saw made TR change the Chimney back to The Priest +5.
Then the same thing happened the following week.
To my surprise in the end my AIFTP prediction went from 248 to 244 up to 246, this is all starting from 239.
Hell I was cool with 244, that Owl and Priest left me sore. 
So I feel that besides RPE there’s something more that TR looking at.
Sounds like a pub
4 Likes
Oh man I didn’t think of that 
1 Like
Anyone in Seattle?
Owl N’ Thistle Pub https://owlnthistle.com/
Unfortunately it doesn’t look like TR has a workout called Thistle.
1 Like
Colorado finally? 
1 Like
In my head, I’m expecting some visuals will come at some point showing a range. Something to cover the 4 week window and something that creates and displays an AI FTP history.
I’m sure some additions could include, e.g. different colours or warnings for new athletes with insufficient data.
Before I get any requests, charts mocked up quickly in Gemini, not real data, etc.
7 Likes
Surely < 1% variation is pretty much identical? Especially as most FTP measurements will have a greater margin of error.
Perhaps every FTP, everywhere, should be presented as n1 +/- n2, where n1 is the headline FTP and n2 the potential variance. I guess that n2 would vary depending upon the source of the measurement - ramp, 20 minutes, etc. - which would give us an insight into how accurate TR believe their AI derived figures to be.
My experience is that ftp prediction gets lower and lower the closer you get. Very wierd.
Also i wish it wasnt called ftp but “training power” or something else because its not ftp by the common definition. Call it training power and just add predicted 60min power and predicted 20min power.
There is no way I could pace an event with this “FTP”.
Also, everytime i pick “very hard” on threshold workouts it wont let me progress, it then gives me an easier workout next week and on top on that it lowers my Endurance work on Sunday to a 30min Endurance ride, 30min isnt Endurance, not for anyone.
Feels like this plan isnt by itself make me ready for my event, like its only objective is to increase a simulated fake ftp number.
1 Like
Don’t get caught up in the specific example (which had a 1% variance after 1 workout).
What I suggest as a change is the initial FTP prediction. From what I can tell, most people are getting a much higher prediction than what actually materialises. This is to be expected as it’s unlikely to complete the plan perfectly.
Therefore, imo, the predicted FTP should be more pessimistic/conservative. This takes into account the psychology that people prefer to be pleasantly surprised rather than mildly disappointed.
For my last 2 ftp updates, I’ve gone up, but not as much as predicted. The overall feeling I have is very happy with my progression, but left wondering where I went wrong that I didn’t go up as much as predicted.
3 Likes
And TR said cycling was best for improving cycling. What do they know.
2 Likes
If your FTP is too high, and the threshold work is too hard - isn’t it right to get easier rides next week?
2 Likes