You are correct about the missing data on that one workout. This should have little to no impact on the interval from 21-Jan up to that date since all of those have data. Am I missing something? AI prediction was 268 watts on 21-Jan and actually was 240 on 18-Feb.
I agree with how it starts with a high prediction e.g. I’m 240 Predicts 250 when I start my 1 month block and ‘if‘ I go with their How Did It Feel rating it still goes down but if I say e.g. Easy is predicted and I put Moderate (like I did today to test my theory) Predicted now plumets to 247!
My situation is a 27-watt drop with what seems to be reasonable compliance and completion! Definitely not perfect
Just a tip for future weeks. I’m pretty sure that doing hard workouts on yellow days doesn’t constitute reasonable compliance.
Apologies, rather than looking at the past period, I’d been focussing on the upcoming period and what we can do to get you out of this cycle rather than the past months.
Looking back the harsh reality is that you’ve only done 2 non endurance rides between the 18th and today, one of which has no power data - everything else has been missed, or endurance, and the highest TSS endurance ride on the 27th has no power data so I’m not surprised your prediction has taken a nosedive.
I’m guessing the lack of data is also making the model conservative for your upcoming workouts.
Taking a step back I think you have to answer a question for your self. Does it bother you that the predicted FTP keeps dropping or that your fitness isn’t increasing the way you would like it to? Even though it might seem as the same thing I do believe that with the new Ai your approach going forward might be different based on your answer.
To be fair, that is exactly the plan prescribed. I do see that there are more problems with the current cycle with missing data, etc. I still have a concern that either I am missing someting, or the model is, for that period between 21-Jan and 18-Feb. That is my main concern and I am just seeing the same pattern with the current prediction cycle.
As I stated before, I am not concerned about the number, but the direction. If the direction that the FTP prediction is indicative of my fitness that is concerning. But also, if the FTP prediction is not good quality, as a software engineer, is my training recommendation going to be good quality? If so, then I would be concerned that my fitness will not improve which is my ultimate concern. It is making me concerned that I am spending money on something that may not give me the proper fitness improvements I am seeking.
If it is all on my (lack of) completing the workouts I am willing to accept that. I just look at the period from 21-Jan to 18-Feb and I can’t make sense of it. The current cycle is in-progress so that is less my concern since I see what has been pointed out about missing data, etc. Sorry to belabor the previous cycle but that has not been addressed to my understanding
I’ve tried, but can give it another shot. Look at your first week of February. If I tell you that the Ai doesn’t like when you do sweetspot and threshold on yellow days, considers all logged workout sets as sets taken to failure for strength training and for some reason doesn’t really like multiple activities on the same day, even though they are only five seconds. Do you still think that you did your previous cycle according to what the Ai prescribed?
Brain fail on my part - I forgot we’re in March 
Torneng has valid points, but looking further back I do see some weirdness too e.g. from what I see, all rides seem to get full credit on all zones (even though done outside - I can’t achieve this in Erg mode indoors!) and there are odd workouts that appear to be paired with the wrong workouts - e.g. workout on the 7th of Feb is 1:22 , gets 3.3/3.3 endurance score, but is paired with a 45 minute endurance workout which doesn’t seem right
I’d be tempted to raise a support ticket to see if someone can look into if there’s been an issue
Thank you very much for the time you have put into this.
That is intended behaviour. Outside workouts are pass/fail, so you will gett full credit in all zones not matter what.
OP,
It looks like you had no riding / time off from Nov 23- Dec 28 (with the exception of a ride here or there) . I am guessing the system (and most coaches would too) started you off easy to rebuild and get back into it. It seems your upcoming FTP prediction is getting dialed in as you have done more workouts and now have some training and data. In the meantime if the current workouts feel too easy I would just select a harder alternate.
Your points are valid, and I will admit that if I did not mark a strength workout as completed, I assumed that the AI would not consider it. Listening to the podcast, I got the impression that working out on yellow days is not “bad” just to be done with caution. I will say that my execution was not perfect, but completing seems like it would be counted as positive, but maybe if I was overextending it may treat that as digging a hole, not doing more is better. Thanks for your input, I do appreciate it, especially calling out what may be my BS 
.
This is interesting. If you have power, why would it not be possible to grade the workout in a similar way as indoors? It would seem like during the nice weather when most people would not enjoy riding inside they would get much less effective training.
I don’t do outside workouts so wasn’t aware of this.
I guess it’s because of how hard it would be to get a good compliance so it will reduce people’s frustration with being ablt to get a “perfect score” due to outside factors
Week 1 Actual 2h26m vs Planned 5h30m
Week 2 5h14m vs 5h0m
Week 3 8h6m vs 6h45m
Week 4 6h41 vs 5h30m
Not bad overall, but getting behind at the start of the 4 week period means that the effect of that lasts almost throughout the whole period. Catching up later in the 4 week period can be good, but the benefits won’t really accrue until March/April.
Your scheduling is a bit lumpy. Periods where you’re off the bike for 4 days out of 5, and then training 9 days in a row including intervals on yellow days. If you could even those out it would help, but obviously we all have to work around everything else in our lives.
What we can’t really tell is this: the prediction you got on 21st Jan assumed (amongst other things) that you’d do 5 specific SS (2) and Threshold (3) workouts over the next 4 weeks. We can see that you did 2 SS and 3 Threshold workouts, but we have no idea how they compare to what was originally planned. My guess is that you ended up doing easier workouts than was originally planned and that’s a key driver as to why you ended up not getting the FTP increase that was predicted. If so, the question is “Why?”
ETA: Note to self. When you get your next AIFTP Detection, immediately screenshot the next 4 weeks in your calendar! 
That would be workout level v2, which is something that was in the works for 4-5 years before TR scrapped it because they couldn’t get it to work perfectly for all scenarios.
I’m pretty sure they got downgraded because the Ai felt he was doing too much.
Maybe. Also, that SS 2.4 to kick off the block looks very conservative, but looking back to the previous two interval sessions that maybe makes sense.