Note how your status is blue at the start of each of those meaning that the optimization was successful in ensuring you’ll be fresh enough at the start of each of them, i.e. they aren’t so hard as to require many days of recovery.
You can increase the recovery between high intensity activities by sliding the Recovery Demands slider to the right and re-running the forecast. This can still give multiple days of high intensity if that’s what fits best into your schedule but the amount of intensity is less overall. Ultimately, it’s not just the number of high intensity days in a row but the amount of intensity in the training prescribed and whether you’ll recover sufficiently between them.
OK, I’m going to have faith in the system and try to complete as prescribed… perhaps I underestimate myself.
If going above/below daily target, then how often should I hit ‘adapt forecast’. I ask because the ‘adapt forecast’ button always has a red dot, even if I’ve just run it.
Slide the Reoovery Demands to the right if you find it’s too much.
You can run Adapt once a day if you don’t mind seeing things change. If you prefer not to see things change and are good at keeping on plan, you can run it every once it a while, maybe once a week.
My target date is set to September - the reason for this is I am going all in on CX racing this year. Perhaps not unsurprisingly, reforecasting seems to take quite a while. I’m wondering what your take is on ideal target dates? Currently I’m considering splitting into 2/3 and using the signatures from those points in time as targets, but am concerned how that will mess with the overall planning (i.e. is there a base phase, does it ramp near the end, etc)
The other thing I noticed is that having reforecasted a few times, the forecast has plateaued towards the end whereas it was ramping all the way till the end initially (Weekly hours actual trends down…) . It seems that the target has been met prematurely and rather than relax the plan to time it perfectly, or continue ramping to exceed initial target, it’s settled on this. Feels like there could be three options: 1) Optimal progression to target, 2) Stretch target if possible, 3) Achieve target earlier if possible.
I think 9 months into the future is too distant to forecast a single plan.
Perhaps this will be handled better when multiple-event functionality is added.
In the meantime, you could plan interim goals one at a time.
For example, I don’t want to peak until August so I set my first goal for March 31 with a conservative target.
In April I’ll float a bit in Continuous mode as I switch to riding mostly outdoors.
In May I’ll set a target event or goal date for August for a more structured approach.
Those who compete likely need some type of ABC event planning. Perhaps the Race option will offer this function. The idea being your A event is more important than anything else along the way. In this scenario the progression chart could be wavy or stepped as you peak for your main event. That’s what I plan to do manually.
Question: I have been using the Simulated Weight Workout to represent that activity. I’m doing these a couple of times a week at this point and up till now have been adding them to the planner as planned activities. Curious - will the AI planner reduce the frequency of / intensity of high intensity bike workouts as it sees that I am doing some yellow and red work via these - presumably it can’t tell they are non-bike…?
If so, I am not sure this is desirable - yes I want to be fresh for the high intensity bike workouts and these can make my legs feel a bit tired but at the same time I know we are not really comparing like for like and worry it may be overcompensating?
Yes, the entry affects form calculation/recovery load assuming the leg work you are doing is the type that carries over to the next day.
When added to the Planner watch the before/after effect on the gradient.
If you want to lower the high/peak strain amount you can copy/edit the workout in Workout Designer and raise MMP by a decimal pt or more.
You could alternately create a placeholder workout with low strain if all you want to do is retain an entry on Planner to indicate gym work.
Can anyone help me understand why Xert has such a poor prognosis for my fitness?
I’ve offered it 3 months of 12 hours a week, and it expects me to get less fit. Last year I peaked on Trainerroad at about a 350-360 eFTP, and could push 500w for 4 mins, on an average of 7 hours a week. I also managed to do 310 normalised for 4h30 hours in a race (73-74kg). Before I put this plan into place my average hours of CV/week was about 8, with two intensity sessions and about 5-8hrs of mixed running and riding in z2.
Similarly it keeps scheduling really weird hours. I’ve told it I can do up to 12 hours - but half of the time it’s only asking me to do 5 hours for weeks on end.
Yet when I look at my training program I can’t get the engine to increase my prescription of intensity - it’s only offering me 1 or 2 small intensity sessions a week. The only thing I can do to enable it to offer me any positive prognosis is to schedule into next year - if i offer it 2 extra months it suggests no change in any fitness parameters.
I’ve set a goal event rather than a goal power; my presumption is that the engine would try and optimise my fitness in the domain I set to be as good as it could be at the date of the event.
The whole “no change in training load” argument is problematic. If you are using AI to predict gains in fitness you are unlikely to see as substantial changes in “training load”, because “training load” is normalised to fitness. I could be doing 80 TSS at 200W FTP, you could be doing 80 TSS at 400W ftp.
So is the constantly decaying TP on a daily basis. Why is Xert’s AI predictor set up on a model that predicts that without a maximal effort your fitness is literally devolving on a daily basis - despite training at, above or around that power. We know from studies into detraining that this isn’t how it works.
I just don’t know how the engine can look at someone who has pushed 310 for 5 hours, 360 for 40 and 500w for 4 mins (on not much more than 8hrs a week on average); decide they actually only have a 320W ftp (that is then dropping by like 1w every day or two), and that if they spend another 4 months training 12 hours a week they are literally unable to improve their fitness?
I just a priori disagree that you are unable to increase your fitness on a static number of hours - but it seems XERT’s AI is locked to this?
Otherwise Tadej Pogacar would have to be doing 100h a week by now - except no - pro cyclists’ weekly volume remains reasonably consistent, year on year. What improves is their total load as they get fitter.
I hesitate to add here that I don’t mind what XERT thinks my FTP is, or my power is. I don’t have an ego attached to it. What I care about is that it seems to not be willing to suggest any improvement is possible.
Do not use Event, use Goal instead. As far as I understand, event will make sure you can complete the event, but not win. So basically handle the XSS you prescribe in that day. And Xert believes you already can do that, so it is just maintaining your fitness.
For Goal you will set a power target for a focus duration. E.g. 350 W for 20 min. The higher target, the higher will the final training load be. You can see how the training volume will increase week by week. Xert will try to ramp up your fitness in the most efficient way (I guess this part can be discussed, as it assumes the Xert fitness model is perfect).
I believe the Xert model assumes you need to increase your training load to improve your fitness. That being said, there is a difference between acute vs criminal training load. But if you keep doing 100 XSS/day then both will eventually converge to 100 XSS/day and Xert will assume your fitness has plateaued.
That’s weird, why wouldn’t giving it a goal race make it optimise to win that race.
Is it clever enough to understand that you could be doing 100 xss a day, but improving in power output?
Back to my point - if the XERT model is right why isn’t pogacar doing 150 hours a week?
And when I change it to a power based goal, give it 4 months - to June - it tells me I am unable to improve any of my power profile on 10-12 hours a wek.
Hi @josephhlbusby . I was just trying to explain how (based on my experience) Xert Forecast AI works. This is also the title of the thread (which I started btw).
It sounds like you want to complain about how it works (or not work). That is a totally different thing, and I suggest you target that energy towards Xert and not users on the forum.
If you want to learn how it works, there are many on the forum (regular users) who can help you out.
As an example I notice is that you assume the hours available is that it is a time constraint. Forecast AI doesn’t schedule workouts for each day, it only assigns XSS (low, high, and peak) for each day. It is therefore really a XSS constraint, and you might increase your training load slightly by doing harder workouts within the same time. Forecast AI assumes around 60 low XSS/hour (easy endurance), if you ride at tempo or sweet spot you will produce more. Or use less time.
So instead of setting this value to 10 hours, I suggest you increase it slightly and look at the plan it produce. Try to match some of the days with a workout and see how much time they will actually take.
Sorry hans - my sincere apologies. I didn’t mean to come across as critical to you - I confess my frustration levels have been high since trying xert; and right now I can’t see a way out or a way to find some value from the engine - because it seems to offer me very little in terms of supporting planning how to train, when to rest or when to conduct intensity.
Can I ask - why does Forecast AI assume .6 IF throughout? Shouldn’t it be telling me what intensity I need to ride at to achieve the adaptations I’m looking for?
My misunderstanding is that I can offer Forecast AI a set of available hours (e.g. 1.5hrs weekdays, 3 hours each weekend), and it will optimise the problem within those parameters.
I personally run Xert in conjunction with TrainerRoad. Unfortunately I have not found the perfect single app to coach me.
TrainerRoad plans are adaptive in certain way, but not enough. If you do something that was not on the plan (or skip something), the plan is not really updated. So if you skip a VO2Max session it will not be moved and you will still have your zone 2 ride the day after.
Xert fixes that, but I do not think it is good at telling me which intensity I should now do my VO2Max at if that is on the plan. TrainerRoad will help you with a natural progression. So if you now successfully did a VO2Max workout, your next will be slightly more difficult. If you haven’t done it for some time or failed the last time you will get an easier. In Xert you have to manage that yourself.
Each workout has a difficulty score. The first one 7.3, second 8.0, and the last one 8.5.
I rated today’s workout as hard. If I rated it as easy or moderate I would get a harder. If I rated it as very hard I would probably get only a slightly harder workout next time.
I can also easily swap it out for a shorter or longer with the same difficulty. In that case it will probably have a bit higher intensity or shorter rests (it is easier to complete 2 x 20.5 min than 3 x 20.5 min).
Okay, I see you aren’t riding Xert SMART workouts.
This is a bit off-topic, but an important distinction so here goes —
Xert’s signature-scaled workouts are a good reason not to mix platforms.
Watch this 5-minute snippet showing how VO2max is scaled – https://youtu.be/bpceJ62Sn-o?t=379
Instead of hundreds of similar workouts anchored on %FTP, one workout scales to each individual’s signature based on SMART intervals (TP, LTP, PP, MPA, MMP, XSSR, etc.).
As a result, one workout adapts to serve the needs of many based on their unique requirements.
As you get fitter the workout scales with you.
This feature allows you to see how your current signature applies to any workout in Xert.
You know what to expect (MPA drawdown and difficulty) before you ride it.
This also works for workouts imported from elsewhere which can optionally be SMARTified during import or tweaked afterwards according to your signature.
Cool stuff IMO. YMMV
