In the next version of AI Workout selection, when the AI hits a yellow day and recommends a substitute on a planned hard day, it should work backwards and reduce the intensity of workouts leading up to that planned hard day. It should sacrifice Friday for Saturday, rather then ask what is the best ride for Friday, schedule Friday, then ask what is the best ride for Saturday. It wouldn’t need to do that for every workout, just when the monopoly card shows up.
Having some issues with AI FTP Prediction - in my Specialty phase.
Context:
Started A “demanding” 12 week plan in December when the Beta first launched. (standard road building FTP, base / build / specialty).
When the beta went public my AI FTP (I’m aware this isn’t my actual ftp, I’ve read the AI FTP Prediction FAQ thread and listened to the 2 podcasts with Johnathan & Nate discussing everything) jumped from 290 → 314, have been training with this “number” since today, where my new AI FTP detection brought me to 328.
I clicked the “check volume” button, which recommended a Demanding approach, but when I accepted it duplicated all my workouts. I had to delete the plan, then re-make the plan with the backed-out start date in December. Looks like my Specialty phase is correctly lined up, but my new Predicted AI FTP is -0.5% (329 → 328) at the end of the Speciality Phase. I referred to the FAQ, and my case doesn’t fall into any of the common scenarios bins. No matter which new training approach I choose (Conservative all the way to Aggressive), my prediction still declines. I noticed the Balanced/Demanding/Aggresive all look identical? I guess Dynamic will scale? Only the 2 most conservative change the plan.
Please analyze my Cal if that helps, I have detailed notes for each workout in addition to survey responses - the workouts in the 314W era seemed to be the “goldilocks” zone of difficulty - I only flat-out failed one, and a few Threshold efforts had the occasional rep where I would lose my cadence due to the ERG spiral of death, but was able to increase torque and ramp it back up quickly after coming to a brief stop.
Is this decline in my Specialty phase an over-correction from my initial 314W prediction back down? But then why did it detect me higher Today? Should I be taking a full rest week before I embark on this specialty phase? Any suggestion would be greatly appreciated, please let me know what other information I could provide to better assist. I am really enjoying the new format so far, it’s keeping me motivated to train (when I see an increase in my future, at least, haha), and the dynamic nature has allowed me to easily adjust to off-bike activities with friends.
A note on the cal - a bit bloated as I record my bike commuting around the city. A new Plan is set to start 2 weeks after this specialty phase ends.
and future weeks with different training approaches:
Use the TR -designated RPE for survey results
Could well do to be fair
It feels like the AI only ever recommends I do Pioneer-3 (other than sunday when I do longer rides)
I changed yesterdays to a 45min but today still remains Yellow unfortunately
First impression is that you’re trying to do a lot.
I would update your forward calendar with non-TR riding you expect to do, even if it could change; add in commutes as Solo rides, Zwift races, and so on as it all adds up. That would give AI a better view of your actual load, it might look less than previous months currently.
I’d analyse whether you really have rest days and recovery weeks.
Then I’d play with the plan, considering that less can be more.
2 Likes
This from @eddie a few months back:
Something else you could do is check out the bottom end of Tempo zone:
Leavitt -4 also carries a WL of 3.9 for Endurance.
I personally would prefer to increase duration. But you seem to have maxxed out 1½h as well, given that it’s having you do James on Sunday.
But to answer your original question:
… it’s probably expecting you to fail the Threshold workout because of fatigue.
It looks like your typical week is one Zwift race, one threshold workout, & few fairly short Z2 workouts. This is not a lot of productive intensity, unless you’re going to treat the Zwift race like another interval workout. Then you have this chunk of time from the 8th to the 11th where you’re not doing any intensity or anything especially long. Possibly some missed gains there. I see you have a movable time-off annotation making scheduling a bit tricky. Can you make Sunday longer? Perhaps 2-3h?
If you can only do one structured interval workout per week, it’s important to prioritise that. Investigate rearranging your week. This might involve moving Thursday’s moderate Z2 to Monday so that you’re fresher on Friday for Threshold. Or swapping your Thursday & Tuesday’s Z2 workouts. That should also leave you fresher for Threshold. Don’t be afraid to skip something before it if it means you can hit that workout hard.
Also, subject to traffic, I’d make the commutes easier, below .6 IF, ideally around .5 IF, & allow it to take five minutes longer. Rough guess on the numbers from them it looks like you’re riding about .85 IF. Commutes this short will always be junk miles, but at lower intensity they at least won’t be fatiguing junk miles. Unless of course if you can weave a pair of traditional 5-minute VO2 intervals into those routes, but it’ll probably be an extra level of unpleasant to do that with a backpack.
1 Like
hmm, the zwift races were baked into the plan (replacing a VO2 workout with the race), the commutes are all <30 minutes, usually 5-10 TSS. Assumed since there’s ~2 years of commuting data backed into the log it would be able to factor in that minimal addition. I tried backing down the plan all the way to “moderate”, but that resulted in an ever greater decrease than balanced/demanding.
My Monday rest day I never do anything demanding on (at most 2 commutes to the grocery store, ~10 minutes, HR in recovery zone, no PM on shitty commuter but from feel <50% ftp).
On the recovery weeks I just follow what TR prescribes, try to keep off-bike activity lowkey. Eg. this past recovery block I went for a gentle hike, learned how to XC ski, neither particularly intense/long days.
In my early 20’s with a non-demanding white collar job, few obligations to distract from training / recovering properly.
Is the bottom line that you’re asking how to set up your plan so that your FTP increases in the Specialty phase? If so, I would try to change the way you think about this phase. Specialty is meant to hone your skills towards the event that would typically occur at the end of the phase. It often means an increase in the specific intensity your event will focus on, but also decrease volume to try to keep you fresh going into the event. All that to say your FTP very well may not go up and might even drop a bit. It’s not a phase aimed at increasing FTP. I’m not sure how the new AI manages this, but historically, a lot of people without a specific event would just skip Specialty and do base/build/base/build…
6 Likes
I think this fact gets even more obscured by the lure of trying to game the FTP detection at the end of the phase!
It’s not clear if @chrishtatu has a specific event as a goal or not; even if not, I think the specialty phase still has value - given the past 8 weeks have seen a 50 watt increase it’s not a bad idea to have a few weeks consolidating at that level before you go again.
5 Likes
Cheers, thank you for the clarification. Looks like I misunderstood the purpose of Specialty. My first warm-up crits will be in early April, so a good shout to replace this specialty block with another Build Phase, push specialty until March, then start my next plan (oriented at Randonneuring vs racing). Or, as @Flashpoint51 pointed out, just stick with this to get used to more VO2 efforts at this level, then go again to build.
Thanks again for your insight & clarification y’all, appreciate it!
5 Likes
understanding your point Nate, can somehow relate to “so-called” though. What everyone names AI these days, used to be an algorithm, not very long ago. I understand that everyone jumps on the bandwagon, but still, naming something intelligent that actually isn’t. Is that excellent?
Regarding “so-called”, “bandwagon” etc, see here for some facts:
3 Likes
SkyNet has your username…
3 Likes
Thanks for the response
The calendar is fairly flexible yea (Just following what TR recommended) with the exception of the Zwift races which are either every Wed or every Sat (I prefer the Wed evenings), typically 25-35min TT’s at >0.9 IF or longer at >0.85 IF.
The TT series ends at the end of Feb, with a 4 week new series running throughout March on the same day and time, however these will be 60-120 min long, same intensity.
I do treat the races as one of the hard workouts of the week, given its entirely at Z4 or higher.
Think ill play around with the durations & maybe turn off Masters plan, although I originally put that on after 3 hard sessions a week left me aching in the mornings, before then having to commute.
Commute wise, they are short and I dont have a PM on the commuter so have to rely on HR, but HR wise I average 134-145BPM, 68% - 74% of Max HR which should have me well inside Z1/Z2. They dont feel like a .85 ride 
.
1 Like
i mean, people are skeptical about the AI because it is doing odd things. Rather than get defensive perhaps hear what people are saying and try to improve the current model. No one wants an “AI coach” that tells you you fitness will get worse if you follow a month of their “custom” workouts they made for you. I mean, we’d all fire a human coach who said that, but that is my experience with TR right now, and many others too. “In 9 days your ftp will be…..8% worse” Wow, thanks! Great training program you made me!
6 Likes
When discussing things it’s important that people stick to facts though, else chaos eventually beckons, and it’s usually much more productive when people park the emotive stuff. For those whom the new system isn’t working well, or how they’d hope it should, the channels to use for addressing that are TR support and/or reaching out on the forum for input from the many people happy to help if they can.
None of this works well when people become emotive, make sweeping generalisations about things not working for anybody, or make claims that aren’t factual. I was directly addressing that latter point in my reference to Nate’s post where he describes the technology, which was in contrast to the OP’s characterisation.
5 Likes
But, but, this totally sucks for absolutely everyone and every single decision it makes is the worst decision it could possibly make and it’s intentionally trying to make me weaker and keep my ftp as low as possible. You can 100% see this in every single post by every single person who has posted in the last 5 weeks, and if you disagree you’re clearly just a fanboi or an employee who gets paid by TR. 
10 Likes
Just came back to TR after being away for 2+ years( still cycling just not withTR). After choosing a training plan, I did a Ramp test to determine my FTP to base my training. A.I Predictor suggested over 20 watts less. I did my first 3 workouts based on the Ramp Test but changed today (more for curiosity) to what AI predicted. Mistake?
Welcome back, the forum reckons it’s been three years..!
No mistake, just go with the AIFTP is the short answer.
The longer answer is that you’ll get the same recommended power levels in your workouts anyway, and then we get into the whole complexity that’s already described in this thread and a lot of other threads.
4 Likes
I feel stupid asking this…
How does AIFTP Detection/Prediction work with indoors/outdoors and different power meters? My outdoor PM reads higher/easier RPE riding outside. Let’s say I’ve ridden lots and I know it’s about 15W difference between indoor PM and outdoor PM.
If I get an assigned workout that prescribes 250W intervals based on winters worth of riding indoors, and then I go and do that workout outside, how does this difference get handled? Two scenarios I can see are below:
- If I ride at the prescribed 250W outdoors, I’m actually doing the equivalent of 235W indoors and not getting the training benefit. But does TR know this? My HR and post-ride survey would look like it was an easy workout and would mess with future indoor workouts.
OR - If I ride at 265W to compensate, then TR thinks I’m progressing massively? It would up the difficultly of my next workout but then say it’s raining outside and I want to ride indoors but now I have this much harder workout assigned.
I just don’t understand how it compensates. I’m guessing it does somehow. Riding with my outdoor PM is not an option. Would like to understand.
Not sure what the solution is to that problem, but TR cannot know the difference and I doubt it will make any compensation for it.
1 Like