I am trying to get along with Forecast AI and have a plan a for an event in May that I am reasonably happy with. However, on my high intensity days, the workouts prescribed by the AI Generator seem absurd or impossible and I think some work is needed on constraining them with some common sense parameters.
First example is a 7hr 42 minute workout recommended for a day when I had fixed availablity of 3.5 hours
Now, I have this one where there is an absurdly small 2 watt difference between work and recovery intervals (come on, it’s really just 3 x 15.x minute blocks), and the first block has me breakthroughing for about half of its duration, which is ridiculous.
I appreciate that this is still beta, but it’s almost unusable in its current state. Surely you can apply some common sense constraints on the AI generated WOs such as: no forcing (repeated) (near) BTs, no WOs longer than availability, clear distinction between work and recovery intervals, etc.
The second workout appears to have an error. What were your targets? When did you generate this?
The first makes sense since you’re asking for 264xss as a mixed puncheur. AIG doesn’t consider your availability since it may generate workouts that would be too hard if it tried to condense 264 mixed puncheur into 3.5 hours. Note how you have the first 7 intervals going pretty deep.
You likely won’t find a workout that meets these requirements in our library (or any workout library quite frankly). Those are big XSS numbers that you’d likely only do on a longer outdoor hard group ride/race. When I see these forecasted, I will likely try and plan an outdoor hard ride for them as that’s likely what XFAI used to base the recommendation on and what will be needed to keep the progressive overload principle.
So yeah, 7 hours on a trainer. No way. But a 4 hour hard outdoor ride, sign me up.
The second workout was generated a couple of hours ago from the planner for a high intensity (Pure Climber) day this coming Sunday.
The first one was generated in a similar manner maybe a week ago. Let me stress that I am not asking for anything specific: it is all being prescribed by Xert! Forecast AI is assigning 260 XSS on a Saturday where I have 4+ hours to play with, and it’s not me who’s saying make it mixed Puncheur. I can perhaps squeeze out 260 XSS in 5 hours, but yeah there’s no way I can do all those mega deep intervals, so why is Xert prescribing them? (I prefer doing workouts out on the road using my Garmin, so I am using the AI generator to try to get a flavour of what I need to do to achieve the high and peak strain prescribed. Will we be getting a way of seeing these live?)
Appreciate you being attentive and keen to understand why things are the way the are.
It appears that you have a latent breakthrough perhaps from your Decay setting. You may want to change it and run with less of a decay for your training. Optimal and Aggressive decay demand greater frequency of fully explanatory breakthrough efforts to keep the signature on track.
Your signature is a bit of an edge case given how low HIE is so I can see how Autogen workouts might cough up something unreasonable. We’d have to spend time with your specific signature and specific autogen request to see why it didn’t behave as expected. You can send that into support so we can do some troubleshooting.
Thanks Armando. I’ll get something to support on this.
I have toggled between Optimal in spring and summer and No Decay during winter over the years, but over the last 6-7 months I had left it on No Decay as my signature from tracking TL was usually spot on, and I’ve never been one to chase BTs.
However, coming into this year I switched back to Optimal and Xert had my TP way over where I would expect it to be at this time of year (~280w instead of 250w - after months of detraining so I know 280 w is unrealistically high). Historical breakthroughs also had my TP at 313 w when in reality, under the old model, I was never much more than 290 w at my best. (I have always dreamed of a 300w TP, and I know I have never really been there!).
My PP has always been very low (I don’t think it has ever been over 800 w) and it is only getting lower as I get older and lose muscle, but I think my real HIE is in the high teens rather than 10 or 11.
Switching to aggressive decay last week and doing a couple of all out sprints on Zwift and some hard climbs has gotten my sig somewhere near reality now. I am in still in the early stages of my build towards 110-120 TL come June, so there may be a small latent BT in there, but nothing all that significant I’d suggest.
I’ll try to get out this weekend and do a proper BT attempt on a long climb where I can keep pushing on as some of the shorter BT climbs I do force you to back off the effort at the top, which may contribute to the low HIE estimation.
Unfortunately, I’ve cleared my plan and now Forecast AI will not create a new one with the parameters I set for target TL or event XSS (both reasonable and achieveable based on past performance, improvement rates and availabilty). FAI also seems to be stuck with my previous Aggressive decay setting even though I’ve changed it to Optimum (can see this from jagged decay of TL during week).
I have given up for now, but there is definitely a bug in relation to setting availablity to >4 hours because when I choose this option for Saturday and Sunday, FAI allocates no training on these days, which is the opposite of what is supposed to happen. Try it for yourself and you’ll see the huge difference between choosing >4 hours and =4 hours.
Edit: I eventually managed to coax a viable plan from FAI and I can now see that AI generated workout is much more reasonable, thanks. The one I now have for Sunday is very hard but it’s within time I have and not 7.5 hours as before!
Please have a look at the >4 hour thing because I want to tell FAI that I can do big rides on Sat/Sun but I seem to be constrained to a max of 4 the way things are, which severely limits my max weekly hours artificially.
