FTP prediction after training phase

I could be mistaken but I vaguely remember Nate talking about or showing a feature that would give a rough FTP estimate or prediction for completing a certain training plan or phase. Did I dream this up or is this indeed something that was at least considered or trialed somewhere?

2 Likes

Ah thank you! I spend a good few minutes searching the forum for something like this but never saw that one.

1 Like

Pretty sure it does exist as an internal TR staff-only feature at the moment based on Nate’s instagram (he put a screenshot of it on his story…). So hopefully coming down the tracks at some point soon, but like you say not around yet.

FTP prediction will still requires max efforts. There’s no free lunch.

2 Likes

Ah I didn’t mean adaptive FTP, I meant a little thing saying something like, Current FTP: 250 if you completed this training phase completely and without issue your FTP will be around 260.

1 Like

That would be a special brand of magic if they could predict adaptations. Or, maybe it would be more along the lines of the average gain for a certain block of a certain plan.

1 Like

Yeah so that’s what Nate posted a screenshot of. Unfortunately it was on his story a fair while ago now and I don’t remember precisely what the screen looked like. I should have made a screenshot :smiley:

Yeah more like the second part I think, they could get that from the stats on gains of other people who’ve completed the same blocks. If you presented it well, with error bars (I guess at a nice wide interval, 95% of users or something) it would be interesting. Although not sure really the practical use.

I think it would be incredibly motivating. “Do this and it’ll gain you x watts” would deffo see me stick with a program longer. The problems are, 1, to be able to give some sort of prediction, and b, what happens if you don’t reach it. Could be a lot of negative publicity if the prediction is off.

I actually think this is something ML could potentially tease out of the database. It’s exactly the kind of statistics that are hard to do without it, due to the large amount of contributing factors. Not all of them are under TR’s influence (or even im their database) though - what happens if the best predictor of increased FTP is sleep duration or weight gain, for example?

Check out Banister’s Impulse-response model. It can model CP changes due to training. Then you can do periodic performance tests and re-fit the model data for closer predictions. If I were to input my planned workouts for the next 12 weeks, the model would predict what the future CP could be.

This is a great paper for basics of testing and modeling.
Skiba and Clarke

I think it would be motivating if (pretty big if!) it worked out for you.

Obviously any forecast like that is going to be on a bell curve with pretty wide confidence intervals. I think its also fair to say the majority of TR users are fairly competitive…

So it sounds great but I think a possible unintended consequence could be people getting disilusioned if they fell in the bottom half, or towards the bottom of the range etc… That could be pretty damaging both to the individuals training and to TRs business in the long run, if it happened like that for lots of people (not saying it definitely would, but we already have a lot of threads where people are disappointed in their progress etc and this kind of function would make it really clear).

So I think its reasonable TR would look pretty hard at the benefits of giving that kind of information, or looking pretty hard at how they communicate it before releasing. It’s clear the functionality exists on some level internally.

1 Like

i’m curious about the feature, but I’m a bit skeptical. I’ve done all base-build-specialty with AT and I have not increased my FTP during this period. On one hand, I don’t do testing, so maybe it’s shifted a little, but by this point after a few years of consistent training with TR I’ve got a good feel about different workouts and where my power/HR align on certain efforts and I haven’t seen anything I’d consider an FTP shift, and I’ve been doing HV stuff really consistently without fail.

1 Like

Yeah, I’m in that group too. What if the ML search digs out that TR doesn’t work for a large number of people? :grimacing:

1 Like

I wouldn’t say AT is not working for me, I am without a doubt fitter in a few ways as far as doing really hard efforts with repeatability. At 290w for my ftp (a bit over 4w/kg depending on my weight) I may be kind of maxed out on the ftp side of things

1 Like

Would be neat if it worked that way. But it doesn’t.

1 Like

Yes, I said previously that periodic efforts would be needed to model FTP. But someone thought it would predict FTP before a block was undertaken - like SSB will give you 20 watts. I said that that would be a special brand of magic. There’s no way to determine how an individual will respond to a certain training block to the watt.

I think it would be great if TR modeled FTP and automatically inserted efforts into the workouts to maintain the model.

Yeah so I think that the only thing you could do would be look at average gain with an error band either side (maybe the 95% interval or something to remove wild outliers).

Someone on another thread has built a spreadsheet model of gains from plans based on TR users self-reported data - it just kicks out average (mean) gain for a certain duration plan. This is from a pretty limited dataset but with the amount of data TR has you could obviously do a lot more personalised estimates based on age, volume, plan compliance etc.

As above based on Nate’s Instagram such a feature does exist internally in TR to some extent, but it’s not been released and we can only speculate why (I do a bit of speculation in that linked thread…)

Well, the model can predict how someone will adapt, but you will always have to individualize the model, and that’s where max efforts are required. Unfortunately, there is no way around that part. Also, it’s unreasonable to expect to be able to predict thresholds/performance to the watt. Our bodies don’t work that way and thresholds are not static.

All I was tying to say is that if you want to get an idea of how your training block could go, you can use the IR model to get an idea and iterate the model as you progress along the training block. It’s one numerical method of prediction. It’s better than just guessing, in my opinion.

Here’s how I learned to use the IR model. I found the accompanying spreadsheet that makes it easy to plug in TSS/Bike score and model the training block. The spreadsheet also makes interpreting CP tests super easy.

I guess general questions I have are: Do people think a projected FTP increase would be motivating or demotivating as far as executing the block of training? And how motivated or demotivated would people be after completing a block and see how their progression lined up with the prediction? How would peoples motivation be affected over time as their training experience increases and presumably those projected increases decrease in magnitude?

Personally, I’m at a point in my cycling and training progression where seeing a projected 1% increase for a 4-6 week block and achieving it would be hugely motivating. I don’t think that would be the case for most people here. Heck, I’d be happy to bust my butt for 6 months to get 1% and thrilled if I could stack up a couple of those back to back.