A workout design question because I'm losing it with Forecast AI

I’m finding that AI generated workouts are impossible, as are recommended workouts. As an example, today I’m following a day of two silver breakthroughs, have a status of tired… my prescribed training:

  • High Intensity
  • Interval target 380 w
  • Keep recovery between intervals ‘very easy’
  • Make sure you are well rested beforehand (LOL after two breakthroughs yesterday, and tired status)

AI generated workout has:

  • 447w intervals - see 380 advised above - 21x interevals (OMG)
  • My best 60s power in the last year is 437, so 447 will kill me, absolutely impossible in ERG mode on the Neo trainer
  • recoveries of 68% FTP - see ‘very easy’ advised above

Now I’m just ignoring the above workout, it’s crazy, too hard and I’m tired before I start. So, I thought “can I create a workout with free ride intervals for the high intensity instead of ERG mode”. I thought of basing the target MPA for each interval on the AI generated workout. This is where I need advise/coaching on workout design.

See below, my workout design objective was to do a %TP target and grade for the freeride interval, then a fixed power ERG recovery for recoveries. I’m assuming that the grade% will set a difficulty on the trainer to simulate the resistance of a 5% incline?

Workout intervals I had setup:

  • Power: 150% TP at 5% grade
  • Duration: 950 W MPA (~96% MPA)
  • Rest-in-between Power 100 W (I expected this to re-engage ERG mode but it didn’t)
  • Rest-in-between Duration: 2:00 minutes
  • Interval Count: 10

I tried the workout, the free ride interval worked, but in the recovery I expected ERG mode to come back and set at 100W but it didn’t, it stayed in 5% free ride mode… does this sound correct? Is there a way to mix free ride and ERG in the same interval sets.

The idea is to create workouts with mixed mode, then I can incorporate the target power and durations as per the daily training advice. I feel the combo of ERG and free ride may be a way to follow the advised training sessions for indoor workouts… I’m trying to find ways to make Xert work in light of the prescribed workouts which I cannot face, nevermind finish.

I suggest taking a look at example “mixedmode” workouts in the library and note how they are defined. This one uses XSSR for RIBs which is a common choice.
You can also create both the work and RIB intervals in Slope mode especially if short intervals and RIBs where trainer response time is a factor.

As to your main issue, overtraining experienced on an XFAI plan, I suggest listening to your body more and not the math, then adjust the math to match (as you are doing now). That means tweaking the knobs/dials/sliders and recasting the plan with a lower goal if necessary. In some cases, Adapt Forecast is enough to reshuffle the deck and provide some relief when needed.
My general advice is don’t follow the plan blindly if you’re not feeling up to it.

I’m following an XFAI plan with a conservative goal and while it’s tougher than the 120 Day Program I’m used to, so far I’m completing weeks intact. :slight_smile: I tweaked the options and experimented to get the plan looking how I wanted.

Thanks @ridgerider2 , so the workout you reference doesn’t use repeated intervals but separate individual blocks. I assume then that if you use a single line interval with repeats, that both the on and off intervals are slope mode? In which case you would just use your own gears and cadence to control the power? Not unreasonable and still may be decent option.

I was trying to create a workout with a mixedmode repeated interval set like below. The idea being that it would be quick to update/change:

This actually looks correct for what you are trying to achieve, and should automatically switch slope to erg and back. In the app, did you ride in slope or auto mode? Needs to be latter

More broadly on XFAI, I agree with @ridgerider2. Listen to your gut and if it feels too hard it is.

You should bear in mind it’s still beta and it’s possible for someone with your very high TL it is just not giving good recommendations.

Some of the behavior you describe just sounds like it’s not working e.g. above, it should not recommend hard high intensity workouts when you are tired (yellow status). Perhaps a case of needing to ‘adapt’? I would report to support directly, and manually choose easy endurance in this case.

You can also play around with polarization and recovery needs to adjust how often you get recommended intensity


@wescaine You are correct, using the repeating intervals does work, switching between slope mode for the hard and then ERG for the easy intervals. I tried it using slope mode for fixed duration and tarket %PP durations, both work great.

I feel like I’ve found something new and exciting.

This may be useful when I can’t face ERG mode for hard intervals. Or to do intervals in slope mode, targettig the prescribed power, or as close as possible… the benefit is that in the recovery phase you can relax and use easy erg mode.

This may sounds like a small feature, but could make a big, great difference for some workouts.

1 Like

The other problem, which no current model can take into consideration, is stress accumulated outside of the forecast.

You might be doing strength work, or there might be job or other life stress. I suspect the need to rearrange workouts, or to insert “easier” workouts, and reoptimize the plan will happen frequently.

I agree, some days I just ignore the advised workout, do whatever I feel is appropriate for how I feel and then run the forecaster.

The issue though, and this may just be me, is mental. If I wake-up wanting to train, but the prescribed workout is crazy hard or just too hard for my tiredness, then mentally I can’t refocus to do something else. So instread of shuffling workouts ( I can’t see hwo to do this) I tend to just waste that day and do what equates to ‘nothing’… I guess at least we have the forecast adapter. However, when following a prescribed training plan I find it important for my motivation that I follow and excute as per the plan. Missing days, ignoring the daily workout, etc lead me to feel I’m not following the plan, won’t see the benefit, so what’s the point.

The challenge for something like Xert over coached plans is that the athlete themselves need to take ownership of deciding what you can do and when (and when not to). It’d be ideal if automated platforms could pull the right info to better predict tireness/readiness at an individual level. I’m not sure if some kind of pre or post workout questionaire could be introduced, eg. each day it asks “how do you feel today” out of 10 and then adapts according to your answer… I know Garmin and TR do this post workout, which I then assume feeds into some kind of algorithm for the days or future training load prediction.

Just do another workout, and rerun the advisor.
I cannot speak for TR, but my Garmin has never asked me a question post workout.

We’ve been working on both the AI workouts and the forecast optimization algorithm to address some of these concerns. We have seen this show up in various forecasts and it is something we had to dive deep into in order to understand why and we are in the midst of evaluating various solutions. What we believe is happening is that too much intensity is being recommended in part due to the polarization demands.

An approach to solving this we are seeing is to spread out the high intensity across multiple days by adding smaller high intensity workouts. This breaks the polarization schedule and, in effect, makes the training more and more pyramidal in structure as the training demands increase. We’re seeing that the polarization method can only go so far as it will generate workouts that will become too hard and/or lengthy to complete in order to sustain the improvement needed to reach the target.

Stay tuned for updates and improvements along these lines. If you think you’ve reached a situation where the training/workouts are getting too hard, let support@xertonline.com know and we can look to generate a new forecast for you using some of the tools we have in development.

A few things you can try to reduce the total amount of high intensity training you’ll need to do would be:

  1. Choose a longer focus duration athlete like Climber or Sprint Time Traillist.
  2. Choose Polar specificity.
  3. Reduce Polarization Level from 4:1 to 3:1 or less.

We’ll be releasing a smaller update tomorrow that you should use to forecast. It’s not a full fix to your issue but will improve things.

1 Like

I have reduced the Polarization level down to 2%, was at 2.5%, will see what happens. I’ll try one thing at a time so I understand what impact the settings have individually.

This is where I get confused, I am sticking with the beta, I am trying to make it work.
Yesterday was ZRL TTT on zwift, this morning I did an hour tempo. On the ‘training’ tab it has me as very tired.
I run the adaptive forecaster and it’s still expecting me to do another bunch of high (27) and peak (7) training today… then high intensity again tomorrow with another 25 high and 6 peak. Holey Moley.

If I go into the planner and hit ‘+’ to add another workout today, then it’s lists recovery workouts, which makes sense, but is different to the training tab. Likewise if I use the ‘planner’ tab to add a workout for tomorrow it proposes recovery-tempo workouts. Not high stress as recommended in the ‘training’ tab.

Which is correct, the training or planner tab advice?

I am not sure if this is your case but I had a similar issue where the suggested values in the planner were crazily high but every time I was checking the training tab it was suggesting very low intensity rides and reloading the page would [edit:] not make them agree.

Turns out that I run the “Adopt Forecast” and it updated the planner with a low intensity suggesting, the exact one the training tab was offering. My conclusion was that the training tab has already done the maths and was suggesting the right thing and the planner is depending on the “Adopt forecast” button.

Instructions to lower intensity were to –

  1. Choose a longer focus duration athlete like Climber or Sprint Time Traillist.
  2. Choose Polar specificity.
  3. Reduce Polarization Level from 4:1 to 3:1 or less.

Did you try doing all three then run Forecast AI afterwards? (not Adapt Forecast)
2.5% to 2% indicates you nudged Periodization, not Polarization.

If you really want to see what differs in a forecast, slam the FAI sliders right or left, not nudge them.
Once you understand the differences you can fine tune things to fit your needs and physiological response.

OK, so let me check the steps required:

  1. On the My Fitness page, go to Program
  2. Set a “Target Date” and “Focus” at minimum
  3. Set a “Target Focus Power” - this is the target power ability for the given Focus, eg “Rouler” = max acheivable 6 min power… not relevant right now, but what if the ‘focus’ is a w/kg, do you put in the w/kg, or calculate the watts, I assume you calculate the watts target?
  4. Go to the “Settings” tab on the Program widow and set the sliders for various settings.
  5. On the Program windows, Program tab, run the "Run Forecast AI’ to regenerate a new plan.

Question - If I want to change any of the settings like focus, periodization etc. Do I need to do it from the ‘Program’ page and retun the “Run Foecast AI”. I assume this is correct.

This means that the “Adapt Forecast” button is only for updating the forecast/plan based upon the training you have/haven’t done it keeps all the same settings? Correct?

I’m hoping that others find my clarifications/questions useful and I’m not the only one learning slowly.

Yes, that is what I have been doing plus ignoring the red dot for up to a week if I like what I see for the week. :slight_smile: