I agree - the TR math seems off in this case and definitely appears that full percentage values have been used instead of ratios.
Just looking at the table from Treff et al. and roughly interpolating shows that the breakdown of Z1/2/3 you calculated from SSBLVI of 14%/79%/7% is nowhere near a PI of 2.0
The debates about it seem to keep going and going. Not always so rational 
seems the same to me 
OK, I just went to the paper for the formula and description, and didn’t read the details about the zones. So zone 1 is > 50% vO2max and < LT1/VT1, zones 2/3 are what I expected.
This change would reduce the amount of z1 in my calculations, which explains my higher number. Fiddling with reducing the
If this is the case, then all values need to be adjusted down by 2, aka log10(100), which means that the highest PI is 1.7 or so.
edit - I need to read the full papers - they have the example I needed:
pi = math.log10(.68/.06*.26*100)
print pi
2.46933101029
This is line 3 from your quoted chart - and this result matches theirs in the table (2.47). I ignored the ones with zero z2, as the formula I am using doesn’t handle that case.
Which, if it is the case, significantly derails some of the major claims/points made on the podcast about the majority of TR plans
My understanding of polarized training is that is fundamentally NOT based on time in zone.
It’s based on sessions.
Wouldn’t the easiest option here simply be to go over this with Dr. Seiler? I’ve communicated with him on numerous occasions. He’s very open.
@Nate_Pearson @ambermalika Just message him on Twitter, get him involved with your polarized training plans, if you haven’t already. It would make for a fantastic podcast.
I haven’t had time to listen to the podcast yet, but I was kind of afraid of that.
I agree that this is Seiler’s classification for Polarized training.
However, the calculations that look at the Polarization Index (being discussed above) are based on the paper below that uses time in zones to calculate the PI: “Zones 1–3 refer to aggregated volume (time or distance) spent with low, mid, or high intensity training” (Treff et al., 2019)
Explains the big gap between the SSBHV plans and everything else, too. If everything is off by 2 log units then you’d keep those fully unpolarized plans at zero and then bump everything else up 2 log units from 0.x to 2.x.
There’s still a LOT of really, really good stuff in that podcast, but I’ve got a bad feeling about these numbers.
Pi is definitely irrational.
thank you!
We covered this in depth in the podcast. Researched used to criticize TrainerRoad is based on TiZ, not sessions. The literature jumps between these two definitions, even with Seiler as an author.
Just a heads up everyone, we’re having some more people double check our math. If we’re wrong we’re very very sorry and will correct ourselves.
It’s late here and we might not get someone to update until tomorrow.
I think our point that SSBHV is an outlier is still a good one.
Thanks Nate - I appreciate you guys looking into that.
I appreciate you double checking us!! That’s really cool!
I’m seriously impressed with the way you guys are handling this whole thing. It’s a tricky thing and hard to get perfect. I think the bigger point you guys were making still seems to hold.
Agreed, It’s a bit murky.
I’ve mentioned it to him personally. He seems to hold the session polarization hierarchy above the TIZ priority. At least from my communication with him. I certainly don’t speak for him though.
My gut read is, that since he’s become an indoor cyclist himself, he’s realized the greater need for tempo and sweetspot work in cycling. I’d be fascinated to know either way.
I’m into the podcast now. Great work.
He’s literally a tweet away…
My take from the podcast and the discussion here is: my IQ is way lower than yours. However, I must say that being on week 6 of SSBHV2, I am very very happy with my progress and looking forward to what’s to come.
You guys are amazing!!!
Oh damn. Our math is wrong on the PI index @toyman.
We’ve pulled the youtube video and the podcast. We are going to correct the math and republish it.
Thank you so much Toyman!! This is not what we want.
I’ll mention this on the podcast, but this is my fault. I should have had a process in place where public data like this is checked by 3 people. We have this review process for the code in our app but we didn’t do this here, and this is a failure in my leadership.
Hopefully, we can correct this from happening again. I also don’t want this to distract the hard work the team did on the study because of some wrongly clicked fields in Google Sheets.
To the point of use TiZ or sessions to define training distribution, I think you have to use TiZ because it’s not straightforward to sort a session into Z1/2/3. For example,when I do outdoor TR workouts, I tend to add in a decent amount of Z1 (in 3 zone model) before and especially after. So maybe I spent 2 hours total riding, of which 30 mins were spent doing VO2 max intervals and the rest at Z1. Different sport I know but on my easy runs, I like to add 4x30" sprints or hill reps to focus on form/muscle activation/turnover. Still an easy session but you’ll never capture that little bit of Z3 work with just going by the whole session.
Or more likely I’m just overthinking. Either way, really great podcast today, I thoroughly enjoyed all of the scientific breakdowns as a data nerd myself! I really agree with your point that it’s hard if not incomplete to draw conclusions from sample sizes of 6 or 12 given the individual variations in responses to training and physiology. Mathematically you might get statistical significance from large enough changes but that doesn’t mean you can generalize conclusion to everyone. Another problem is your study population, well train cat 3 males of a similar age- what’s to say the results from that study apply to master athlete or teens or women?
If my employer had this same mentality they’d have a lot more loyal employees. I applaud you.
