No what I mean is that xert see TR SS workouts as LOW stress and thus endurance.
Yeah, anyone using blockchain in their workout routine?
2 Likes
Nate has said in prior podcasts that the most fit people (presumably FTP or wkg@ftp) tend to ride the most. Not a shocker considering cycling is an aerobic sport and building aerobic fitness requires time. I would guess they have at least one data scientists trying to suss out some patterns. Of course, not all might be useful, we could learn that being a 5wkg rider could be strongly correlated with owning a red bike. lol.
Just speculating here, but I would guess that the best (negative) correlation with FTP increases would be how new one is to structured training / TR and how long one has been riding. I don’t mean this as a put down, but basically newbie gains. One is going to make quick gains at the beginning from just riding and being new to riding and at some point from adding structure if they are not at the same time. After time the gains become smaller and smaller as one approaches their potential or what they can achieve given their cycling time and recovery constraints.
One of the problems with big data, is just that there is so much data. There are people who heavily modify TR plans or use TR but design their own plans. Identifying those people and making comparisons to other TR users is probably not very easy.
Another couple of things I thought of re: adaptive plans:
- Predicting workout ‘failure’ (i hate that phrase, but whatevs…) based on other recent rides and adjusting schedule. Basically if it sees you did a 300TSS ride on Saturday, maybe move that Tuesday interval session to Wednesday.
- Predicting plan ‘compliance’ and adjusting plans to get better compliance. Really tough problem, but it would be interesting to see what links they could identify (say, overestimating FTP) with people quitting plans and either adjusting parameters such as FTP or adjusting plans to add more rest.
1 Like
True. What has worked for me (N=1) is that I follow the TR workouts recommendations in SSB LV 1 and 2 and follow XERT recommendations for intensity as well: that is, color of star. If it recommends a LTP level workout then I will not do a TR SS workout if scheduled but opt for a LTP type in XERT.
Important: But this hasn’t been an issue following LV plans. ( I burn out eventually with mid volume plans, and have never tried a high volume plan because of this.) In fact, I have found that doing a LV SS plan and then riding on the weekend at endurance pace has been POLARIZED in nature.
Cheers.
2 Likes
Thats a really good way to do it in my mind 
Have you tried letting XERT recommend you a specific workout?
I was just looking at their site and am unsure how I feel about that. On one hand, the adaptive nature seems appealing, on the other there is usually some mental and fueling preparation that happens before a workout. My mental attitude is a little different going into a 3x30 SS workout vs 6x3min VO2max. I definitely would not be in the best mood if I was surprised by a VO2max workout. lol.
1 Like
The workouts are never surprising as you know the stage of base, build, prep that your in and your freshness status plus you are offered 5 workouts to choose from with the option of choosing others. So you have the final say.
1 Like
New TR feature that sort of belongs in this thread (or not):
A ride ending flat on Saturday turned a 3.5 endurance ride into a 1 hour ride, so yesterday (Sunday) I tried this beta feature and no endurance recommendations at 90+ minutes:
Its a beta feature. And only in the desktop beta at the moment IIUC.
2 Likes
Just want to address the point in the opening post: Xert is not an artificially intelligent system - it is an algorithm based system that you could fairly easily reproduce in a spreadsheet if you wanted.
It doesn’t learn anything that changes the recommendations it gives bar perhaps the amount of time you have to train on a given day of the week, and even then I think it just looks back at the last six weeks and takes an average.
Mike
2 Likes
What the problem, lol? If you hopped on the bike expecting Pettit but the “AI” loads and says: We see you’ve recovered well, your workout today is instead…
6 Likes
??? TrainerRoad (Bluetooth connection) allows me to set the following modes on my Kickr 2017 (wheel off):
- Erg
- Resistance
- Standard/Level
The only mode not supported by TR is Sim mode.
1 Like
Full TR article, but it misses the Standard/Level option (that is only an option for Wahoo trainers on BLE, AFAIK).
1 Like
All three modes described here:
but not in the Kickr support article.
1 Like
Oh, cool! I have not seen that specific one. 
lol. I wake up at 4:50am and try to be riding by 5:05-5:10am. Not the kind of surprise I want at that time, for sure.
3 Likes
Very cool. I had not seen that before.
I wished TR let you build mixed mode workouts. For example:
- ERG for warm-up and recovery period
- Standard at a set level for intervals
Mixed mode would be great for 30-30s, “all out”, etc. To be really marvelous, TR would need a “Standard” level calculator:
- Select trainer
- Select target power - either absolute or % of FTP
- Select target cadence
- Select target gearing
From the above, TR would spit-out the recommended standard level for your trainer. So for mixed mode, you would set the standard level, and ideally when the interval started, you wouldn’t need to shift.
3 Likes
Not a real answer, but an option. TR uses the ‘T’ keyboard shortcut for swapping between trainer modes quickly. Mobile is a royal pain, and there’s an open Feature Request to make it easier. So depending on your app device, you may have something to try.
1 Like
Yes, I’m aware of this shortcut, but I’m not aware of a way to set the resistance level. Nor a way to calculate what gear I should be in at what cadence with what gearing. To really make this useful, I’d want to have all of this. So if I’m doing something like 30-30s, I’m not hunting and pecking for gears. I’m all set.
1 Like
It’s called the “Gartner Hype Cycle” 
As for the question about AI/ML. I kinda doubt that AI/ML will replace coaches anytime soon. At best, I think it would be more like “Augmented intelligence” where the machine supports the decision making process of humans. But I think one important question that AI/ML could help answer is to which degree one responds better to intensity or volume (so that we move away from generic training advice to more personalized).
Having dabbled in AI/ML professionally, one issue I see is that data are messy and most people probably wont spend the time clean up data properly to enable correct prediction. I for example, have data where PM or HR monitor are clearly showing wrong data. Also, I would probably need data from more sensors, such as sleep, resting HR, weight etc. to make valid predictions.
2 Likes