Forecast AI almost no goal attainable

I haven’t had time to look at the Forecast AI until now, just been winging it with the training. I’m in pretty good shape right now but certainly not peak. However, when using the Forecast AI goal setting and trying to set some goals it basically says I can’t make but a few watts improvement for any athlete type, why is that? I’ve played with available hours, how much specificity, target date etc. It seems to think I’m at~ 98% of my max potential for every rider type.

Example, it says I can only increase my 4 minute power from 407 to 424 in 2+ months with 18 hours of training. I haven’t trained this in a long time but use to be able to do 400+ for over 5 minutes. What am I doing wrong? Do I just set a goal that it says is unattainable then when I reach it (hopefully) set another higher goal?

Ok so you are basically in the same position I was in with Xert about 3 months ago. If you search my posts you’ll find me arguing with Armando etc about this over threads and threads. Fwiw - my profile at last BT was 375/1450/295 and working 4min with a goal of 515.

Some things that helped.

  1. go and look at the Xert advanced settings for ‘training responsiveness’. That is the model’s understanding of how much fitness you get for a given fitness input. For some reason it had me modelled at the bottom (possibly off the bottom) of the scale. This meant that it needed a long time to make any improvements (eg I couldn’t even sustain the same 4 min power unless my plan duration was over 6 months).

  2. Make sure you aren’t syncing any activities that arent directly improving your cycling fitness. E.g. I had circa 4 hours of lifting a week synced to Xert on top of my 8 houes of riding. I then went to forecast AI and said ‘ok so I’ll do 12 hours a week not 8’, but it was like ‘you’re already doing 12 that isn’t an improvement you won’t get fitter’

  3. Upload more data - as much as you can - and look back on your training for the last 4 years for breakthroughs. Really be critical - how much have I ACTUALLY improved my watts @ x minutes every year. Eg, 10w on your tp year in year out would be pretty damned good, but if you’re then asking ‘i want 10w on my tp and here is 4 months Xert go’ you can see it might say ‘no chance’. Sometimes tho why it is saying no is frustrating and unintuitive.

  4. As much as Armando has promised me it isn’t, I still think that there is an element of it that is normalised to a curve of ‘how long goes it takes the average person to get to a 450w 4min power’. When I was asking it to get me to 500w for 4mins, it assumed I had to be tadej pogacar. It works better if you are in the middle of the bell curve.

  5. In the end it doesn’t matter. Set it to focus on the power duration you want e.g. 4 minutes. Set the hours honestly, and do the work. I think I got caught up in thinking that if I could only set it to 30w improvement not 15w improvement I’d get a completely different plan - but it’s still mostly going to depend on your periodization and polarisation settings (aka how soon in your plan to you want to be doing intensity and how much / little / hard to you want the intensity to be). The point is if you do ronnestads in slope mode you have every opportunity to prove the prediction wrong :-).

It should all come down to the responsiveness settings. Could you show yours?

sorry for the slow response. How do we know what to set the responsiveness to? I upload plenty of data, I ride on average about 15 hours a week. I don’t input any other type of workouts or data, just cycling power data. My numbers aren’t even at the peak of what I’ve been in the last 12 months, so it’s weird that the AI is telling me I can’t improve any of the numbers by more than 5-10 watts when my FTP alone is about 25-30 watts lower than it was 12 months ago, my max power is down about 80 watts from what it was 6 weeks ago. Some of this is due to not doing consistent break through efforts but I’ve also been busy and haven’t been able to map out an actual training plan and a target event. I’d like to get that going this next week however.

My current settings

0.4 is probably on the very lower end for threshold responsiveness and regarding to what you said about your FTP now versus 12 months ago while the AI plan doesn’t believe in you that might be the issue.
It basically says that using a standard ramp rate of 2 you could increase your TP only by 3w over a whole month of training.

You can verify the number yourself:

  1. note your peak TP 12 months ago (TP_peak)
  2. note your low training load the same time (LTL_peak)
  3. note your current TP (TP_curr)
  4. note your current low training load (LTL_curr)
  5. TP_diff = TP_peak - TP_curr
  6. LTL_diff = LTL_peak - LTL_curr
  7. Threshold_responsiveness = TP_diff / LTL_diff [w/xss]

Threshold_responsiveness should be around that 0.4, else your current setting might me off.

Can you post these 7 steps and the result?

I’m not understanding all your formulas, are those available somewhere I can just copy/paste? I’ve finally had some time to sit down this morning and try to work out a schedule for a road race on August 4th. I can’t get Xert to cooperate. When I use the Forecast AI and input all the info for the race, it says I’m already ready and puts together a plan that will have my numbers LOWER than they are today.

Make it make sense. My TP is down 30 watts, my Peak power is down 45 watts from numbers I’ve seen fairly recently, and those are numbers I achieved without any real structured training.

When I try to use the old method of planning for an event, the planner says it can’t compute the taper week and doesn’t complete the plan.