Yesterday predicted ftp was 305 or so
Just did a vo2 max workout in crazy heat and lost almost 20!!! watts lol
And yes I’m just doing my own program and progressing but thought I’d keep aiftp for giggles
I’m not sure what is more obnoxious, the software, (with a known bug), dropping predicted FTP down 20 watts, or, expecting a 22 watt gain in a month is realistic.
Remember that while the predicted FTP obviously has its flaws for some people, in some scenarios , it is a feature in a BETA release state.
“A beta phase generally begins when the software is feature-complete but likely to contain several known or unknown bugs”
If you know there is an extenuating circumstance that the AI couldn’t possibly track (crazy heat is a perfect example of this) then click the pencil icon to edit the workout and uncheck the TR AI setting. This will remove the workout from AI FTP calculations and you should be back where you were. You can’t really blame the AI for not knowing you did the workout in crazy heat.
I can definitely think of situations where a 22 watt gain is feasible (well at least one, going from a detrained state to trained)
There is at least one other thread with super weird ftp predictions and it may be a bug. Definitely contact support, it’ll help them make it better
Joe
A beta is supposed to be an opt in early access for users that want bleeding edge and are ok with bugs. If you release a product to all of your customers as the only way they can experience the product, it’s not a beta anymore. That’s just your production release.
If you then call it beta, you’re just gaslighting your customer base.
It’s a common tactic, because it allows studios to lower their code quality standards, and then deflect criticism “because beta.”
It’s been clearly labeled beta from the start. But I agree that they should turn off prediction for everyone and then whoever wants can opt back into beta (like the old days) and they can selectively add users into beta.
I think the system works way better than most people complaining thing. They are just focused on the prediction.
I worked for one of the largest software companies in the world. We would only ever release beta to a carefully selected group of customers who would accept the risks involved. The idea that you might just chuck it out to all customers and hope most of them don’t hate it is bizarre.
Did you follow the structure of the prescribed workout? 
Also, how did you rate the effort?
I think I can see what’s going on here. Can I share those details here in this thread?
I rated it max effort.
That’s what I always do if I can’t finish prescribed sets.
I’m just doing my own workouts, I don’t follow a program, I only made a program just to see what happens with the ftp estimate.
Sure you can share whatever ideas you have
For the last month or so I do 5x4’ vo2 max Tuesday and have been able to up the watts. Thursday my version of hard starts. Sat group ride Sunday long ride.
I don’t care about the aiftp except for interest. When I made the plan it gave me 285>308 I think. Throughout the four weeks it fluctuated in the low 300s and then boom lol.
I think part of the problem is that you did a VO2 Max workout, but it’s associated with a Sweet Spot workout. Where did this Eichorn +1 workout come from, and what TR workout (if any) were you doing?
If you’re not following a plan and simply hand-picking workouts, that’s likely going to make your FTP predictions fluctuate more as well, and there’s likely a lot of moving parts in your plan from one day to the next.
Regardless, if you couldn’t finish the sets of the workout, that will likely have an effect on your FTP prediction.
Let me know if this helps.
It was original a select Beta, then a public Beta.
I think the problems with it were unfortunately a surprise to TR following the select Beta. Perhaps it should have been pulled back to a select Beta, but the ship has sailed?
My personal experience is that the current TRAI is producing better results than the previous iterations of the software. The FTP detection is also as accurate as the previous detection method (aligns with Strava Premium, Garmin and Intervals.icu).
But the BETA FTP prediction is generally an unrealistic 15-20w gain each cycle, but I achieve a 5-10w gain if I stick closely to the plan.
I still like the prediction as I can see how changing my training effects the likely outcome at the end of the cycle. I have just learnt to mentally halve the number it gives me.
A+ for the Training and FTP detection but C- for the prediction.