In the training profile the threshold cp was 271 so I changed it to what my cycling power profile has for cp which is 204 and the 30:30s are realistic targets now.
3 questions
Thanks
Short update, so after I manually changed CP yesterday, it suggested a new threshold this morning of 270; it was previously 271 set by the software and reduced it manually yesterday. I accepted it, and the threshold CP has changed in the training profile to 270 (which seems right I have my FTP at 240). Iāll update later what this did to the 30:30 intervals when the plan finishes refreshing.
In my cycling power profile, these are the trends
|Year |VT1 222W, CP 284W, MAP328W, MPP1142W|
|3 months |VT1 159W, CP 204W, MAP 275W, MPP 873W|
|6 weeks|VT1 158 W CP 203 W MAP 275 W MPP 873 W|
14/07/24 athletica VO2 - 3 x 9 x 1 30/30 intervals 286 watts fitness level combined 104 bike 68
23/09/24 athletica VO2 - 3 x 9 x 1 30/30 intervals 286 watts fitness level combined 65 bike 40
29/11/24 athletica VO2 - 3 x 9 x 1 23/09/24 athletica VO2 - 3 x 9 x 1 30/30 intervals 286 watts fitness level combined 65 bike 40
29/11/24 athletica VO2 - 3 x 9 x 1 30/30 intervals 371 watts fitness level combined 60 bike 37
fitness level combined 60 bike 37
Sorry for all the questions, Paul you know I love this platform and what ye guyās are building I just need validation of the plans
why the VO2 - 3 x 9 x 1 30/30 intervalsāincreased significantly to 371 watts when fitness has decreased?
I put my FTP at 240 at the moment, but the cycling power profile CP is 204.
Folks, this has been going on too long now, and Iām not impressed. Iām losing confidence in the platformās ability to detect basic training and health metrics and prescribe the correct workouts.
here is another good example clearly showing the targets are too high after detraining for 5 months now.
For an aerobic session, Athletica prescribes 174 watts. If I do an Endurance session in on Wahoo SYSTM for the same target duration, it prescribes 143 watts. For the session, the Athletica target HR is 130 and the result of doing the lower intensity Wahoo session is 139???
I have tried triggering a new plan and changing around settings below and no change:
Let our Smart Coach choose the best option
Prescribe workouts by heart rate
Last week, there was a recommended Threshold of 270 to 250 changes that I accepted, which marginally reduced the VO2 targets but not the other sessions.
At the moment, Iām training unstructured and plan structured training again in January. I would appreciate this being resolved before then and having confidence in the prescribed workouts.
@Pablo_ie I understand your frustration regarding the platformās inability to accurately detect your thresholds, and I empathise with your concerns. I would feel the same way if my thresholds werenāt appropriately set.
I feel like the primary reason behind these thresholds being inaccurately calibrated is likely due to how we currently track maximum efforts over extended periods. This approach doesnāt always account for different training phases, such as detraining, as well as it should. This is certainly a limitation, and I agree that itās disappointing.
We are actively working to address this and provide better solutions. Specifically, we are planning to introduce functionality that will allow you to select a time range for evaluating efforts, which will significantly improve threshold accuracy.
Our current modelling choices are designed to reflect average user behaviour, aiming to work well on average. However, adapting to individual scenarios like yours can sometimes be challenging, and in this case, itās clear the model isnāt performing as expected. I sincerely apologise for this.
I would like to prioritise these improvements, and I hope theyāll be ready before you come back to structured training in January. I truly appreciate your patience and understanding as we work towards making the platform more robust and personalised for all.
OK I can work with that, thanks
This conversation is a powerful reminder of the remarkable capabilities of the human mind.
Obviously, I declined. Baffling how the system is recommending increases.
Zone change detected. Our Smart Coach suggests updating the following bike zones:
So the same situation this year detraining since mid August and efforts are not adjusting accordingly. True AI should be well able to detect adjustments are needed without a user needing to fail a workout. Last years situation was why I didnāt use the app this year, I left the work I was doing syncing to see if things had improved.
@Marjaana @Prof just like target efforts in workouts why hasnāt my power history reduced since July give reduced training frequency and load since then? Power history hasnāt updated since July?
Have you done a test week when you restarted training? I took a decent break after my main event this year, and did a test week when restarting training Tests included MAP and well as FTP etc. The zones and thresholds have been working just fine.
Hi Phil, I didnāt stop training I reduced load and intensity. Workouts from trainerroad and weekend club spin are still syncing. If itās AI driven it should be able to know and notify you of recommended reduced power target ranges like it does when the ranges increase. The app has years of my training data and models build on large data sets. If the pitch is AI the intelligence should be just that and not require athletes to test. Cheers
My power profile says last 3 months map is 313 and last 6 weeks 253. If I schedule a 40/20 vo2 session the target intervals are 392. When they should be I would estimate be between 260 to 290. The 392 target hasnāt changed well beyond 3 months.
If Iām using an app like this I want to be able to load a work and do it in erg mode and believe itās the right targets to drive adaptation based on the recent data. Yeah some people would say donāt do vo2 sessions on erg and thatās their preference. In TrainerRoad it will adjust as recent as the previous days work out.
This has prompted me to look at my MAP values
Last 6 weeks has me at 247W MAP
Last 12 months (year) has me at 279W MAP.
My VO2 30/30s give a range of 253-304W. 253 is 26 less than 279, and 304 is 25 more. Giving the strong evidence the VO2 range is based on the last year value.
A comment on the intervals. I did a total of 36 intervals over the sets in this weekās VO2 session. As I do them outside I just go with what feels right within the range. Intervals.icu gives me an average of 283W over all the VO2 intervals this week. Thus it seems my MAP is actually closer to my year high than the last 6 weeks indicates.
I think the problem for me is that I havenāt done a maximal 5 minute effort in the last 6 weeks and Athletica doesnāt have any in my scheduled workouts. I did the Maffetone test and FTP but not power intervals in between. Since VO2 looks to be anchored to my max mean 5 minute power (definition of MAP in web app) , sounds like I need to fit in an effort of that duration. I suspect that will reset the 5 min value for last 6 weeks back to around 279W or slightly higher.
To be fair the AI coach does suggest this on my power profile chart.
Have you logged any near maximal effort between 3 and 5 minutes? If you have already logged these efforts recently, you can ignore this.
But using the mobile app mostly, I donāt look at my power profile chart (only available on web version) as often as I used to. Itād be useful to get that feedback in the mobile app .
Going back to your situation. If I hover on my. VO2 planned workouts, the target is 278.50W, which is what a smart trainer would use. If you look at your power profile for last 12 months, does it show 392W as MAP?
Hey Phil itās 353 for the year. TrainerRoad has my 30:30 at 278 Iād say 290āis about right for my level of fitness in the vo2 range at the moment.
Cheers
Hey @Pablo_ie
I understand your frustration. You are frustrated that AI Coach is not keeping up with your detraining, and is getting you back on the bike with to high of targets. You know that you can always manually override the thresholds, decline automatic updates, and decline automatic updates.
We can have the best models and yet fail to meet the expectations. Human bodies donāt always fit models. What Iāve learned with my own training is that we can lose the high power/ sprint ability quickly, but also regain it quickly. We are also made up of different combinations of fiber types, and for example keeping up with some strength training and sprint training can produce quick improvement or even prevent high power to drop when we donāt train as much volume as we used to. AND when we do a good block of aerobic base volume, we can absorb that higher end work much more and can see big improvements.
Which leads to changes that are not always visual just looking at the numbers. What do I mean by that?
When you are approaching a 40/20, 30/30, or 60/60 workout, it is good to know a target. I think we all agree on that. But some days, we just donāt have it. Too much stress, too little sleep etc etc. Other days we can really knock the ball out of the park. @Prof taught me this: Know your target but donāt limit yourself with the upper target number. If you donāt ever let yourself go above your targets, how will you ever improve?
Following this advice, Iāve seen myself go from VO2max target range of 280-320w to over 400w and even on over 500w on a good day. Had I always limited myself to modeled numbers (correct or incorrect) Iād still be targeting 300w unknowing I can crank out over 500w on a series of 30/30s.
MAP and definitely CP may be a little slower to adapt (at least physiologically), and in our model weāve decided 6 month is a good cut off of detecting meaningful changes up or down, unless my memory has completely failed me.
Physiology is not always that straight forward. Remember you have full control of automatic detection or not. Keep listening to your body. And keep keeping on pushing us forward 
MJ
Hi @Marjaana
The query I have is purely about the AI capabilities being positioned. Iām fully OK with my declining ability to hit Vo2 max targets. I expect it and know how to manage sessions. I havenāt used athletica for training this year due to the same lack of deload reaction that happened last year, my training is still syncing thou. In the post above from Dec 2024, @Andrea mentions improvements were planned; I donāt see them. Also, you mention the ability to decline update recommendations there have been none to decline.
These are the pitches by athletica.ai
Personalized Adaptive Training Plans: The AI develops dynamic, individualized training plans that respond automatically to workout results, daily health data (like HRV and resting heart rate), and subjective feedback. Plans are updated in real-time to ensure progressive adaptation and readiness for races or key events
AI-Driven Performance Analysis*:** After every workout, the platform analyzes performance metricsāpower, pace, heart rate, RPE, and user commentsāto refine profiles, update training targets, and flag trends like overreaching or fatigue early.*
Automated Threshold Detection: The platform calculates and continuously updates aerobic and critical power/speed thresholds without clinical lab testing, ensuring all training zones reflect current ability
My power graph hasnāt reacted to deload and havenāt moved since July 19th when deload started. 30:30 are still at a target of 392 after 3 months of deload.
The Cycling Power Profile numbers for VT1, CP, etc. are changing shouldāt workout session targets reflect this?
Given this and that the policy is 6 month is the cut off of detecting meaningful changes up or down how can the bold items in the athletica.ai pitch be true?
Thanks
Thanks @Pablo_ie,
Iām sorry to hear about your experience with Athletica. Let me clarify a few points.
The core capabilities you mentioned are implemented. It is important to highlight that conservative safety and quality rules sometimes make the system appear slow to react. Thresholds are based on a rolling six-month window of data. They update automatically when one of your personal bests (over any relevant duration) leaves that window. This ensures that only meaningful, sustained changes in performance, not short-term fluctuations, affect your training zones.
Now, itās also important to highlight that training load and physiological thresholds cannot yet be directly linked in a reliable, causal way. This relationship remains largely unknown, so we rely on maximal efforts to estimate current capacity. Six months has proven, on average, to be a good compromise across athletes: stable enough to prevent noise-driven changes, yet reactive to genuine performance shifts. Maybe depends on some athleteās characteristics (e.g phenotype) maybe depends on the CP/MAP values themselves. Some users find it too slow, others too fast. This is why weāre working toward letting users select their own window length in the future, though even that would remain an arbitrary choice.
On a broader note, Athletica improves every day. Some improvements are visible; others happen quietly under the hood. Profiling and threshold estimation lies at the core of Athletica, but itās a face of a giant ecosystem. We continuously refine how thresholds, dynamics, and adaptive behaviours interact, but we sometimes have to prioritise other aspects of the platform first. AI is bounded by both science and implementation constraints, so we need to navigate between features or aspects that can be highly customised and aspects that needs to work well for the average.Yes, if zones changes, workout limits change as well automatically. If you look at your power profile and you select last 6 weeks, or 3 months or the rest of the year, you will notice the changes in MAP and CP thresholds:
Last 6 weeks: MAP 259 W, CP 209 W
Last 3 months: MAP 313 W, CP 225 W
Whole 2025: MAP 357 W, CP 270 W
This is the effects of previous max efforts leaving the a window. Whether this is reflective of reduced workload or absence of max efforts we (and anybody) do not know. This is the refinement loop acting behind the athletica interface, in line we what we say.
If this is not reflected in your training zones updating there might be a problem (issue, bug, sync, updateā¦) that we are not aware of. Your 30:30 are prescribed in the middle of Z6. Erg mode will impose 392 because halfway between 357 (MAP) and 428 (bottom Z7). HIIT science suggest 30:30 to be at 105% MAP, so you could consider based on this number that 374 would be already in Z6 prompting the adaptations targeted by this type of HIIT. However, as you can see, if only the last 6 weeks are accounted for, MAP declined to 259 W. Again, we need to check internally if your threshold is still the same for some tech reasons.
(my bold in quoting you above)
I canāt speak for others, but unless Athletica schedules a max 5 min effort, then I donāt do them. I notice my power profile summary recommends I do some max efforts between 3ā and 12ā mins. But itās not front and centre, and using the mobile app I am looking far less at the power profile page in the web app.
I do wonder if it would make sense for the Athletica AI engine, to substitute the default VO2 max 30/30s with a VO2 max 5 min session at least once every 6 weeks. With the instruction to go all out, which likely means just the one interval and put it in a lower load week? The 5 min sessions are in the global library, though not all out variety.
Then maybe when the athlete sets RPE, if they mark it as 10 (all out) you have a recent marker for MAP. Then take a weighted average between current MAP value and this newer maximal effort.
If we donāt know whether itās the absence of max efforts, then letās schedule them at an appropriate frequency, if they are critical to setting the VO2 max power ranges. Of course I ride outside, and just go for what I think I can repeat for the number of sets and intervals. I see the VO2 range as just a guide, but not a limiter. If you train indoors on ERG for VO2, I can see the MAP value being off, being a challenge.
thanks @Andrea
So what I interpretate is that I either have a bug or that Athletica is very good at adjusting targets when progressing (getting a new max effort over any relevant duration) by not then regressing in abilities?
6 months seems like a long time for someone of my age profile for the update automatically when one of your personal bests (over any relevant duration) leaves that window. Iām 50 and regression happens much faster than it used to.
Thanks for highlighting the power profile based in this data to be truly AI shouldnāt we be recommended a reduction in workout targets that can be accepted or not just like the increases?
Last 6 weeks: MAP 259 W, CP 209 W
Last 3 months: MAP 313 W, CP 225 W
Whole 2025: MAP 357 W, CP 270 W
The last update on the threshold / power history chart is 19/7.
