From the folks at Trainerroad, who seem to be willing to share a bit more of their data nowadays than they were previously.
If a long ride has TSS >100% of your average weekly TSS from the last 6 weeks, it can disrupt your training for the next 3 weeks.
Even if it has TSS >50% of that average weekly TSS, you are statistically 54% more likely to miss your next hard workout.
Since TSS != TSS, I think this is a gross over simplification (meaning I am sure there are lots of exceptions) but on the flip side I actually agree with the conceptā¦ Too much is too much. Again likely 18 year olds have more tolerance and recover faster then 60 year oldsā¦ But people should look at the core idea and realize that a long ride that pushes you too far is potentially not taking you closer to your goals.
I am sore for last 3 days after hiking down some mountainā¦ 
for 3 hoursā¦ carrying a 20kg packā¦ Taking the gondola up did not seem to require much recovery though.
I heard this on the podcast episode and was intrigued. So I went back through my logs to see if this ever happened in 2025. I rarely ever exceeded 50% and my long rides were generally 4-5 hours long at my highest volume last year.
What Iād be curious to know is since most recreational cyclists probably get their greatest TSS on their weekends (back to back days), if that has any effect. This was the case for me as Iād routinely go out and do my long Zone 2 rides on back to back days with my work week being time crunched 60-90 minute max. Most people probably do a long ride and a group ride, which I assume to be spicy and carry a higher TSS. For some, that group ride might even be their long ride for the week.
I think a deeper dive would be warranted on this one.
Iād suggest they are giving one example of ignoring the principal of progressive overload.
If your weekly ride time is 3 hours/wk then going out for a 3+ hour ride on Saturday is probably going to wipe you out. I think any reputable coach would recommend you gradually increase the long ride length.
The 3 hour/wk guy should gradually build his weekly hour total and the long ride along side it.
Dave
Yes your point brings up part of the problem here. Which is there are so many variants to this story that creates exceptions. Only doing 1 hour rides and going for 3 is also bad for most yet you might fit within the acceptable 100% of TSS rules. There are many cases that you could do > 100% of TSS and also be fine. If you slowly build up to doing 3 hour rides, you probably could do that 1 ride a week and be fine doing 3:15 or similar variants.
On the flip side I still believe they are bringing up a valid point and another way of looking at and realizing that over doing it does not usually produce positive gains. As I said I am a product of that right nowā¦ My MTB vacation turned into a hiking downhill vacationā¦
Is generalising from a large dataset in the way that TR has done here (and I expect will do more of in future) any better / worse, than the results of a scientific study involving 15 participants?
If you think estimating your max HR from 220-age is reasonable then yes generalizing across a population is fine. That Ā± is usually the killer, which one other person pointed out this same problem in studies.
The bigger problem in this āstudyā for me is that TSS is a very flawed metric so trying to use it to generalize results is risky. But if you forget about precise math and focus on the bigger picture of the message itās right on. So kudos to them for this.
Also I think on an individual level using TSS as a metric here and finding your own limits within the way you ride yours might be 120% or might be 80% and this can change depending on how you are currently training, eating and sleepingā¦. When you take TSS and apply it to the individual and self/coach/system analysis it can be useful.
The other interesting idea and question that is not easily answered is. I donāt have time to train for 4 daysā¦ have to work out of the city. Does over doing it today by doing XYZ have such a negative impact in this case? Also how fun is XYZ
For an enthusiast in an established training regime, taking the six week average TSS as the marker likely makes it safe enough for a ārule of thumbā such as the one proposed.
Anyone who is a serious medal contender for the olympics should probably avoid such generalisations when planning their training.
Could be a good question to ask TR to have a look at.
Should one go into an enforced ārestā period via a big blowout workout?
I think it is good information, the issue in my opinion comes from the idea that these are all TR users and they share a similar worldview.
Specifically, you are only getting data from people that 1) are heavily inclined towards indoors training as either a majority or the only cycling workout, 2) believe in and execute plans with 3X hard to very hard intensity sessions per week, 3) are so heavily time crunched that they canāt get much volume in or they do not find value in volume.
I think if someone like Garmin used data only collected from their 5xx, 8xx, and 10xx Edge devices youād get a much better cross sectional area of how people actually ride their bikes than the data TR has.
In the spirit of full disclosure, I admit Iām heavily biased against the TR model.
Dave
Yes, Garmin, Wahoo and Strava have been gathering data for a long time now.
Maybe TRās example might nudge them towards sharing some of their insights one day.
Yes donāt increase by more than 10% weekly is similar such advice. For a more data geeks type this TSS advice is good as well, especially for people highly consistent and middle age. For 16 year old or 80 year old probably not the best adviceā¦ 30-40, consistent every week, primarily a cyclist, likely doing a fair amount of intensity itās pretty reasonable advice, but this obviously represents a very small minority of cyclists in the world but a reasonable percent of TRs cyclists. Tens of thousands of them/you.
Overall more guard rails for over achivers is not a bad thing
Yes I agree, itās especially slanted toward the TR type of cyclist. I am not anti-TR. I used to be more so when their was such a focus on high volume sweet spot, but they have toned back to much more reasonable advice. Not my favorite but a million times better than it was.
Imo, no better.
They were selective with the cohort used in the dataset, and it fitted their narrative.
What was the result of other cohorts /datasets if they looked at any? We can guarantee that if the results didnāt fit their narative they will not have share it.
Anyway, nothing we didnāt already know, do a hero weekend you might miss a bit of training, is that a bad thing? The only difference is they put some numbers on it.
What is more interesting is how shocking low the average TR users power PRs are for various durations. See the YouTube on the subject, split in men / women and 10 year age buckets.
Is it a good idea to generalise training based on this massive datasetā¦ my analysis of quality data at work suggests probably not.
Or keep the same 3 hr during the week and build the weekend ride up by 10-15 minutes a week.
That wouldnāt break their rule.
I thought they were saying weekly TSS which includes the weekend ride.
Also I hope they excluded data from after the new AI because there load numbers now arenāt TSS as aiFTP doesnāt equal the FTP TSS was based on.
Well put.
Great point, TR are excellent story tellersā¦ They come up with data to back their storiesā¦ Some people just need a belief and trusting something and in this case matching the data to the story creates value. I should get better at this. Riders using TrainerDay increase their FTP by 30% per year on average. Oh did I forget to say thatās only the ones that started a 1w/kg
I wonder how much of your own data is needed to do something like Kristen Faulkner does?
So I had discussions with Andrea in the past and we were discussing the holy grail would be turning the 42/7 PMC model into something dynamic 50/5? Other groups in the past were working on this and called it performance modelling. How fast do I recover, how fast do I improve.
There is also very deep alternatives to TSS being worked on by some smart guys but what I find is most of this requires pro level tracking and consistency to have a chance of turning into something meaningful.
So while I am a bit anti-tss for many cases itās still very useful in specific cases and specifically because we all understand it for the most part. So ultimately what we want to know is how fast we recover from which efforts. But diet, alcohol, health, other activities, sleep, temperature, wind, other kinds of stress all likely need to be included to have a a robust enough dataset to come to conclusions more rapidly/consistently. So every idea that emerges is just a theoryā¦ Then it needs to be validated with deeper testing. The more variables you introduce (like not knowing sleep patterns) the longer it likely takes to come to clear patterns. Most people donāt have enough time or energy for this.
So generally if collect every piece of data you can maybe even a daily journal for 5 minutes for years and dump it all to ai in a few years including all training activities, you likely can find a lot of decent trends.
Even basics like (some/many can be automatic)
7/21/2026
Sleep 7:10hrs
Ate: Healthy, last meal 7pmā¦
Morning resting HR: 52bpm
Stress: Work is a bit stressful
General Daily feeling: Great, lots of energy
Workouts in Strava: Felt hard for the last 1/4 but the intervals felt 6/10 RPE
Wind low
Temp during training 23c
If you had 3 years of data like this then I would say itās very likely that machine learning and AI could tell you a lotā¦ You can just use tools like claude code or GPT codex and it will do it all now if you have the data or a lot of the data. Just say all my files are here. Use machine learning to analyze it and tell me what you seeā¦ and start asking questionsā¦